Properly

properly

Memoization is awesome. Let's abuse it. TL;DR: Don't apply premature optimization too early Problems Readability
Code Complexity
Premature Optimization
Obscured Logic Solutions Apply memoization in actual real business situations, and measure its impact through empirical benchmarks. Context Memoization can help you improve the performance of recursive functions involving redundant computations but compromise code readability and maintainability. It would help if you only used it with strong factual evidence on real business case scenarios. Sample Code Wrong memo = {}
def factorial_with_memo(n):
    if n in memo:
        return memo[n]
    if n == 0:
        return 1
    result = n * factorial_with_memo(n-1)
    memo[n] = result
    return result
  
  # This function optimizes the computation of factorials
  # by storing previously computed values,
  # reducing redundant calculations 
  # and improving performance for large inputs. Right def factorial(n):
    if n == 0:
        return 1
    return n * factorial(n-1) Detection [x]Semi-Automatic You can search for all places where you are using this technique and validate if they are worth it. Exceptions Real performance problems with strong factual evidence Tags Premature Optimization Level [x]Intermediate AI Generation Unless you explicitly ask the IAs to use this technique, they will suggest cleaner solutions. AI Detection ChatGPT, Gemini, and Claude.ai detect some problems with this technique but do not mention readability as a concern. Conclusion It would be best if you kept a balance between performance optimization and code clarity. You can consider alternatives such as iterative approaches or algorithmic optimizations since memoization significantly compromises code readability. Relations Code Smell 06 - Too Clever Programmer Code Smell 20 - Premature Optimization More Info Wikipedia Disclaimer Code Smells are my opinion. Credits Photo by Steffen Lemmerzahl on Unsplash A cache with a bad policy is another name for a memory leak. Rico Mariani Software Engineering Great Quotes This article is part of the CodeSmell Series. How to Find the Stinky Parts of your Code Memoization is awesome. Let's abuse it. Memoization is awesome. Let's abuse it. TL;DR: Don't apply premature optimization too early TL;DR: Don't apply premature optimization too early premature optimization Problems Readability Code Complexity Premature Optimization Obscured Logic Readability Code Complexity Premature Optimization Obscured Logic Solutions Apply memoization in actual real business situations, and measure its impact through empirical benchmarks. Apply memoization in actual real business situations, and measure its impact through empirical benchmarks. memoization Context Memoization can help you improve the performance of recursive functions involving redundant computations but compromise code readability and maintainability. It would help if you only used it with strong factual evidence on real business case scenarios. strong factual evidence real Sample Code Wrong memo = {}
def factorial_with_memo(n):
    if n in memo:
        return memo[n]
    if n == 0:
        return 1
    result = n * factorial_with_memo(n-1)
    memo[n] = result
    return result
  
  # This function optimizes the computation of factorials
  # by storing previously computed values,
  # reducing redundant calculations 
  # and improving performance for large inputs. memo = {}
def factorial_with_memo(n):
    if n in memo:
        return memo[n]
    if n == 0:
        return 1
    result = n * factorial_with_memo(n-1)
    memo[n] = result
    return result
  
  # This function optimizes the computation of factorials
  # by storing previously computed values,
  # reducing redundant calculations 
  # and improving performance for large inputs. Right def factorial(n):
    if n == 0:
        return 1
    return n * factorial(n-1) def factorial(n):
    if n == 0:
        return 1
    return n * factorial(n-1) Detection [x]Semi-Automatic [x] Semi-Automatic [x] You can search for all places where you are using this technique and validate if they are worth it. Exceptions Real performance problems with strong factual evidence Real performance problems with strong factual evidence Tags Premature Optimization Premature Optimization Level [x]Intermediate [x] Intermediate [x] AI Generation Unless you explicitly ask the IAs to use this technique, they will suggest cleaner solutions. AI Detection ChatGPT, Gemini, and Claude.ai detect some problems with this technique but do not mention readability as a concern. Claude.ai Conclusion It would be best if you kept a balance between performance optimization and code clarity. You can consider alternatives such as iterative approaches or algorithmic optimizations since memoization significantly compromises code readability. Relations Code Smell 06 - Too Clever Programmer Code Smell 06 - Too Clever Programmer Code Smell 20 - Premature Optimization Code Smell 20 - Premature Optimization More Info Wikipedia Wikipedia Disclaimer Code Smells are my opinion . opinion Credits Photo by Steffen Lemmerzahl on Unsplash Steffen Lemmerzahl Unsplash A cache with a bad policy is another name for a memory leak. A cache with a bad policy is another name for a memory leak. Rico Mariani Rico Mariani Software Engineering Great Quotes Software Engineering Great Quotes This article is part of the CodeSmell Series. How to Find the Stinky Parts of your Code How to Find the Stinky Parts of your Code

Premature Memoization: How to Properly Apply It - Code Smell 250

I lost hundreds of thousands of dollars on hiring ineffective sales specialists for my B2B startup. With the knowledge I have now, I could have avoided that. In 5 years of scaling my company, Competera, from zero to $28.2 billion in revenues, I’ve hit all sorts of bumps trying to develop an effective sales strategy. This experience taught me a lot. Today, my company has over 120 clients worldwide, with checks ranging from $50,000 to–300,000. Here are my four rules for managing a sales department in a B2B — in early-stage and late-stage companies. startup Always start with founder-led sales. When I started my company, I would manage sales on my own. In 9 months, I closed three deals for almost $1 million ARR. A founder’s involvement in sales at a B2B startup is unavoidable (and desirable).  I don't know of any cases of growing a B2B SaaS company any other way. If you want to expand your early-stage enterprise, you need to do sales yourself until you reach $3-5 million ARR. Hiring a sales account executive early is like having someone else between you and your child. No one will be more passionate and enthusiastic about selling your product than a founder. The most valuable thing you get by doing sales yourself is early feedback from users and the opportunity to build a long-term relationship with them. B2B sales are also called human-to-human for a reason. As a startup founder, you build valuable interactions and nurture relationships with potential customers. I always tell other founders, “In your first years, make friends – not customers.” One more reason to start with founder-led sales is an extremely high conversion rate. While the standard SaaS conversion rate is 15-20%, I’ve seen firsthand that a founder-led startup’s conversion rate can be as high as 60%. That’s important for any company, let alone a new one that is trying to stay afloat, start making money, and grow. Craft sales playbook A startup founder who leads deals and gets feedback from customers has more understanding of their company’s value proposition, goals, and its clients’ needs. All of this information has to become the foundation for salespeople who’ll join the company at later stages. The knowledge generated during those first months or years by the founder should form the basis of a sales playbook. It’s a blueprint that helps future reps navigate the sales process, train new hires, and deal with difficult sales cases. Crafting such a guide will benefit your sales effectiveness (some say by up to 50%). No competent sales manager will join a company without a playbook with all the necessary sales information. So until a playbook is completed, stick to a founder-led sales strategy, and don’t waste budget on sales reps. Here’s what to include in your first sales playbook: . Write up your background, the reason you started the company, and the vision of the future you can create with your product, and add success stories that you can boast about. The story you sell Have a detailed description of a target customer, her day in life, problems, pains, values, budget range, and other indicators that are important for you. At Competera, we describe our ICP this way: “A commercial director who is responsible for increasing the company’s profit in retail stores with more than $100 million in revenue.” Target ICP (ideal customer profile). Describe how interaction with your customers looks like — from an inbound request to rollout. Ideal sales formula. . Mention parts of your pitch and product that collect aha moments. Aha moments . What doubts do your customers have, and what arguments can you use to clear these doubts up? Key objections you face Never outsource sales at B2B startups After a year of leading sales, I decided to hire sales representatives. I chose a well-proven outsourcing sales team from the U.K. During a year of our partnership, they closed… one deal. After this experience, I never outsource sales and advise you to do the same. Here are my arguments against outsourcing sales: . Startups update the product frequently (every six weeks at my firm), and external sales reps couldn’t follow the iterations. You lose alignment between teams and reduce your chances of landing a long-term deal. An outsourced team doesn’t keep up with product changes . In B2B sales, customers estimate your startup’s energy, vision, and how engaged you are. An outsourced team can’t give you this, and it’s vital in B2B. You lose your company’s vibe in communication with clients . After an external sales manager provided me with a client’s feedback that sounded unreal to me, I decided to get to the bottom of things. So, I called a potential customer to figure out the problem. It turned out that the sales manager was so annoying that he just made up an excuse to get rid of him! Due to the outsourced team not being as involved in the sales process, I was missing important comments and the opinions of potential clients. You don’t get honest, direct feedback I believe that, for a startup with up to $10 million ARR, outsourcing sales can be devastating. Sure, there were benefits to our collaboration. The outsourcing team improved the sales structure and helped finalize my company’s playbook. But we lost a lot of time that we could have spent hammering out long-term deals. Always have at least two in-house sales managers. On average, 42.5% of sales reps take ten months or longer to become productive enough to show results. You should diversify your risks and hire at least two managers to compare performance. Check intermediate results of the new hires: conversation rates from the first positive response, from response to demo, from demo to follow-up conversation, and from follow-up conversation to “money.” Outstanding professionals are never actively searching for a new job, so you will have to headhunt. Identify companies from your industry that sell your average check, find out about their sales managers (LinkedIn can help), and convince them to work for you. I discovered the right candidate from Germany on LinkedIn and spent two years talking to him online. Then I found his former boss on social media and promised a stake in Competera if he would convince this guy to join me. Now, this sales manager brings Competera $3 million annually, and his former boss is our advisor and has a share in the company. Another way to find a strong candidate is to turn to a boutique agency that specializes in hiring sales managers. They test candidates, assess their psychological profile, and so on. We hired such an agency and received 117 candidates, from which we selected 40 to take the test. Of all those who passed it, five were invited for an interview, and only one got the job — the best one. Key points Until a startup’s ARR reaches several million dollars, its founder should do the sales themselves. Founder-led sales show a 60% conversion rate, while the average sales manager’s rate is up to 20%. Don’t hire a sales team unless you have a playbook, a guide containing key info about a startup’s vision, ideal customer, sales formula, and how to handle objections/rejections. Gathering this information will improve your sales effectiveness. Sales reps in B2B startups should always be in-house. An outsourced team won’t be 100% synced with your dev team, and they won’t talk to clients as much to get their feedback. Well-performing sales managers never actively look for a job. Headhunt specialists from your rivals or hire a boutique recruitment agency.

‘Outsourcing Sales Can Kill Your B2B Startup’: Here’s How To Manage Sales Properly

The context package effectively manages and propagates timeouts and deadlines through a chain of calls in your product architecture. However, be mindful when using it. It's essential to be aware that it can also introduce performance degradation to your service in some cases with external calls. In this article, I tell you about some of them. TLDR; or have a chance to destroy the performance of your service. Be careful. Use a good cook recipe. ctx.WithTimeout ctx.WithDeadline For this article, I use as an example of an external service we use. But the cases I’m talking about are generic and can be applied to any other service or database. Redis Let’s start with a basic example of code to make a call to Redis: import "github.com/go-redis/redis/v9"

reddisClient := redis.NewClient(...)

ctx := context.Background() 
ctx, cancel := context.WithTimeout(ctx, 200*time.Millisecond) 
defer cancel()

result, err := reddisClient.Get(ctx, "redis-key").Result() This code tells us that we expect a response for the Get request to come within 200 milliseconds. And everything looks logical. But what if the Get request exceeds the timeout of 200 milliseconds? When the context deadline is exceeded, all operations associated with this context should be terminated. Let's take a look under the hood at what happens when the context deadline is exceeded.
The Redis client: Closed the current connection as a connection is not safe to re-use it Established new connection: Performed handshakes using a new connection. These three steps basically mean that we don’t use the connection pool anymore if we have a good amount of requests with exceeded deadlines. Every time for requests that exceed the deadline, we close the current connection and open a new one. It makes each operation slower and adds risks to exceeding the deadline one more time. As a result, it can be an infinite loop, It affects the performance SLA of your service and bottleneck performance for the whole product. What can we do to avoid this problem? Context is used to propagate timeouts through a chain of calls, and each call has a random timeout that depends on the timing of previous calls. To prevent it for some calls in a chain, we can use a separate timeout for every external call: ctx := context.Background()

result1, err := reddisClient.Get(ctx.WithTimeout(ctx, time.Second), "redis-key-1").Result() 

result2, err := reddisClient.Get(ctx.WithTimeout(ctx, time.Second), "redis-key-2").Result() Don’t set up timeout too small because it can cause a problem we discuss here. In a case where we need to respect incoming context timeout and build a response within that timeout, we can also use a go-routine to perform a call to an external service: ctx, cancel := context.WithCancel(ctx) 

// Process asynchronously in a goroutine. 
go func() { 
  defer cancel() 

  result, err := reddisClient.Get(context.Background(), "redis-key").Result() 
  ch <- result 
  close(ch)
}() 

select { 
case <-ctx.Done(): 
  // error happend 
case res := <-ch: 
  // success 
} In the code above, we build a response in time, but we also give redisClient to complete the Get request in the background.

Go Context Timeouts and External Calls: Cook Them Properly

Anne of Green Gables by L. M. Montgomery, is part of the HackerNoon Books Series. You can jump to any chapter in this book here . Mrs. Rachel Lynde Is Properly Horrified CHAPTER IX. Mrs. Rachel Lynde Is Properly Horrified ANNE had been a fortnight at Green Gables before Mrs. Lynde arrived to inspect her. Mrs. Rachel, to do her justice, was not to blame for this. A severe and unseasonable attack of grippe had confined that good lady to her house ever since the occasion of her last visit to Green Gables. Mrs. Rachel was not often sick and had a well-defined contempt for people who were; but grippe, she asserted, was like no other illness on earth and could only be interpreted as one of the special visitations of Providence. As soon as her doctor allowed her to put her foot out-of-doors she hurried up to Green Gables, bursting with curiosity to see Matthew and Marilla’s orphan, concerning whom all sorts of stories and suppositions had gone abroad in Avonlea. Anne had made good use of every waking moment of that fortnight. Already she was acquainted with every tree and shrub about the place. She had discovered that a lane opened out below the apple orchard and ran up through a belt of woodland; and she had explored it to its furthest end in all its delicious vagaries of brook and bridge, fir coppice and wild cherry arch, corners thick with fern, and branching byways of maple and mountain ash. She had made friends with the spring down in the hollow—that wonderful deep, clear icy-cold spring; it was set about with smooth red sandstones and rimmed in by great palm-like clumps of water fern; and beyond it was a log bridge over the brook. That bridge led Anne’s dancing feet up over a wooded hill beyond, where perpetual twilight reigned under the straight, thick-growing firs and spruces; the only flowers there were myriads of delicate “June bells,” those shyest and sweetest of woodland blooms, and a few pale, aerial starflowers, like the spirits of last year’s blossoms. Gossamers glimmered like threads of silver among the trees and the fir boughs and tassels seemed to utter friendly speech. All these raptured voyages of exploration were made in the odd half hours which she was allowed for play, and Anne talked Matthew and Marilla half-deaf over her discoveries. Not that Matthew complained, to be sure; he listened to it all with a wordless smile of enjoyment on his face; Marilla permitted the “chatter” until she found herself becoming too interested in it, whereupon she always promptly quenched Anne by a curt command to hold her tongue. Anne was out in the orchard when Mrs. Rachel came, wandering at her own sweet will through the lush, tremulous grasses splashed with ruddy evening sunshine; so that good lady had an excellent chance to talk her illness fully over, describing every ache and pulse beat with such evident enjoyment that Marilla thought even grippe must bring its compensations. When details were exhausted Mrs. Rachel introduced the real reason of her call. “I’ve been hearing some surprising things about you and Matthew.” “I don’t suppose you are any more surprised than I am myself,” said Marilla. “I’m getting over my surprise now.” “It was too bad there was such a mistake,” said Mrs. Rachel sympathetically. “Couldn’t you have sent her back?” “I suppose we could, but we decided not to. Matthew took a fancy to her. And I must say I like her myself—although I admit she has her faults. The house seems a different place already. She’s a real bright little thing.” Marilla said more than she had intended to say when she began, for she read disapproval in Mrs. Rachel’s expression. “It’s a great responsibility you’ve taken on yourself,” said that lady gloomily, “especially when you’ve never had any experience with children. You don’t know much about her or her real disposition, I suppose, and there’s no guessing how a child like that will turn out. But I don’t want to discourage you I’m sure, Marilla.” “I’m not feeling discouraged,” was Marilla’s dry response, “when I make up my mind to do a thing it stays made up. I suppose you’d like to see Anne. I’ll call her in.” Anne came running in presently, her face sparkling with the delight of her orchard rovings; but, abashed at finding the delight herself in the unexpected presence of a stranger, she halted confusedly inside the door. She certainly was an odd-looking little creature in the short tight wincey dress she had worn from the asylum, below which her thin legs seemed ungracefully long. Her freckles were more numerous and obtrusive than ever; the wind had ruffled her hatless hair into over-brilliant disorder; it had never looked redder than at that moment. “Well, they didn’t pick you for your looks, that’s sure and certain,” was Mrs. Rachel Lynde’s emphatic comment. Mrs. Rachel was one of those delightful and popular people who pride themselves on speaking their mind without fear or favor. “She’s terrible skinny and homely, Marilla. Come here, child, and let me have a look at you. Lawful heart, did any one ever see such freckles? And hair as red as carrots! Come here, child, I say.” Anne “came there,” but not exactly as Mrs. Rachel expected. With one bound she crossed the kitchen floor and stood before Mrs. Rachel, her face scarlet with anger, her lips quivering, and her whole slender form trembling from head to foot. “I hate you,” she cried in a choked voice, stamping her foot on the floor. “I hate you—I hate you—I hate you—” a louder stamp with each assertion of hatred. “How dare you call me skinny and ugly? How dare you say I’m freckled and redheaded? You are a rude, impolite, unfeeling woman!” “Anne!” exclaimed Marilla in consternation. But Anne continued to face Mrs. Rachel undauntedly, head up, eyes blazing, hands clenched, passionate indignation exhaling from her like an atmosphere. “How dare you say such things about me?” she repeated vehemently. “How would you like to have such things said about you? How would you like to be told that you are fat and clumsy and probably hadn’t a spark of imagination in you? I don’t care if I do hurt your feelings by saying so! I hope I hurt them. You have hurt mine worse than they were ever hurt before even by Mrs. Thomas’ intoxicated husband. And I’ll never forgive you for it, never, never!” Stamp! Stamp! “Did anybody ever see such a temper!” exclaimed the horrified Mrs. Rachel. “Anne go to your room and stay there until I come up,” said Marilla, recovering her powers of speech with difficulty. Anne, bursting into tears, rushed to the hall door, slammed it until the tins on the porch wall outside rattled in sympathy, and fled through the hall and up the stairs like a whirlwind. A subdued slam above told that the door of the east gable had been shut with equal vehemence. “Well, I don’t envy you your job bringing that up, Marilla,” said Mrs. Rachel with unspeakable solemnity. Marilla opened her lips to say she knew not what of apology or deprecation. What she did say was a surprise to herself then and ever afterwards. “You shouldn’t have twitted her about her looks, Rachel.” “Marilla Cuthbert, you don’t mean to say that you are upholding her in such a terrible display of temper as we’ve just seen?” demanded Mrs. Rachel indignantly. “No,” said Marilla slowly, “I’m not trying to excuse her. She’s been very naughty and I’ll have to give her a talking to about it. But we must make allowances for her. She’s never been taught what is right. And you were too hard on her, Rachel.” Marilla could not help tacking on that last sentence, although she was again surprised at herself for doing it. Mrs. Rachel got up with an air of offended dignity. “Well, I see that I’ll have to be very careful what I say after this, Marilla, since the fine feelings of orphans, brought from goodness knows where, have to be considered before anything else. Oh, no, I’m not vexed—don’t worry yourself. I’m too sorry for you to leave any room for anger in my mind. You’ll have your own troubles with that child. But if you’ll take my advice—which I suppose you won’t do, although I’ve brought up ten children and buried two—you’ll do that ‘talking to’ you mention with a fair-sized birch switch. I should think that would be the most effective language for that kind of a child. Her temper matches her hair I guess. Well, good evening, Marilla. I hope you’ll come down to see me often as usual. But you can’t expect me to visit here again in a hurry, if I’m liable to be flown at and insulted in such a fashion. It’s something new in my experience.” Whereat Mrs. Rachel swept out and away—if a fat woman who always waddled could be said to sweep away—and Marilla with a very solemn face betook herself to the east gable. On the way upstairs she pondered uneasily as to what she ought to do. She felt no little dismay over the scene that had just been enacted. How unfortunate that Anne should have displayed such temper before Mrs. Rachel Lynde, of all people! Then Marilla suddenly became aware of an uncomfortable and rebuking consciousness that she felt more humiliation over this than sorrow over the discovery of such a serious defect in Anne’s disposition. And how was she to punish her? The amiable suggestion of the birch switch—to the efficiency of which all of Mrs. Rachel’s own children could have borne smarting testimony—did not appeal to Marilla. She did not believe she could whip a child. No, some other method of punishment must be found to bring Anne to a proper realization of the enormity of her offense. Marilla found Anne face downward on her bed, crying bitterly, quite oblivious of muddy boots on a clean counterpane. “Anne,” she said not ungently. No answer. “Anne,” with greater severity, “get off that bed this minute and listen to what I have to say to you.” Anne squirmed off the bed and sat rigidly on a chair beside it, her face swollen and tear-stained and her eyes fixed stubbornly on the floor. “This is a nice way for you to behave. Anne! Aren’t you ashamed of yourself?” “She hadn’t any right to call me ugly and redheaded,” retorted Anne, evasive and defiant. “You hadn’t any right to fly into such a fury and talk the way you did to her, Anne. I was ashamed of you—thoroughly ashamed of you. I wanted you to behave nicely to Mrs. Lynde, and instead of that you have disgraced me. I’m sure I don’t know why you should lose your temper like that just because Mrs. Lynde said you were red-haired and homely. You say it yourself often enough.” “Oh, but there’s such a difference between saying a thing yourself and hearing other people say it,” wailed Anne. “You may know a thing is so, but you can’t help hoping other people don’t quite think it is. I suppose you think I have an awful temper, but I couldn’t help it. When she said those things something just rose right up in me and choked me. I had to fly out at her.” “Well, you made a fine exhibition of yourself I must say. Mrs. Lynde will have a nice story to tell about you everywhere—and she’ll tell it, too. It was a dreadful thing for you to lose your temper like that, Anne.” “Just imagine how you would feel if somebody told you to your face that you were skinny and ugly,” pleaded Anne tearfully. An old remembrance suddenly rose up before Marilla. She had been a very small child when she had heard one aunt say of her to another, “What a pity she is such a dark, homely little thing.” Marilla was every day of fifty before the sting had gone out of that memory. “I don’t say that I think Mrs. Lynde was exactly right in saying what she did to you, Anne,” she admitted in a softer tone. “Rachel is too outspoken. But that is no excuse for such behavior on your part. She was a stranger and an elderly person and my visitor—all three very good reasons why you should have been respectful to her. You were rude and saucy and”—Marilla had a saving inspiration of punishment—“you must go to her and tell her you are very sorry for your bad temper and ask her to forgive you.” “I can never do that,” said Anne determinedly and darkly. “You can punish me in any way you like, Marilla. You can shut me up in a dark, damp dungeon inhabited by snakes and toads and feed me only on bread and water and I shall not complain. But I cannot ask Mrs. Lynde to forgive me.” “We’re not in the habit of shutting people up in dark damp dungeons,” said Marilla drily, “especially as they’re rather scarce in Avonlea. But apologize to Mrs. Lynde you must and shall and you’ll stay here in your room until you can tell me you’re willing to do it.” “I shall have to stay here forever then,” said Anne mournfully, “because I can’t tell Mrs. Lynde I’m sorry I said those things to her. How can I? I’m not sorry. I’m sorry I’ve vexed you; but I’m glad I told her just what I did. It was a great satisfaction. I can’t say I’m sorry when I’m not, can I? I can’t even imagine I’m sorry.” “Perhaps your imagination will be in better working order by the morning,” said Marilla, rising to depart. “You’ll have the night to think over your conduct in and come to a better frame of mind. You said you would try to be a very good girl if we kept you at Green Gables, but I must say it hasn’t seemed very much like it this evening.” Leaving this Parthian shaft to rankle in Anne’s stormy bosom, Marilla descended to the kitchen, grievously troubled in mind and vexed in soul. She was as angry with herself as with Anne, because, whenever she recalled Mrs. Rachel’s dumbfounded countenance her lips twitched with amusement and she felt a most reprehensible desire to laugh. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. L. M. Montgomery (2008). Anne of Green Gables. Urbana, Illinois: Project Gutenberg. Retrieved https://www.gutenberg.org/cache/epub/45/pg45-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org , located at https://www.gutenberg.org/policy/license.html . The lead image was generated using HackerNoon's Stable Disffusion Al Image Generator feature, via the prompt "Anne of Green Fables".

Mrs. Rachel Lynde Is Properly Horrified

Many people are familiar with (Model-View-Controller) and its challenges, yet only a few discuss how to address them effectively. In this article, we will delve into the vital components of : the and the . MVC MVC View Controller Typically, a is employed as both a and a , resulting in increased code within the . So, let's explore the possibility of separating the into distinct and entities. UIViewController View Controller UIViewController UIViewController View Controller Let's begin by examining the lifecycle of a . UIViewController One of the methods in this lifecycle is . According to Apple's documentation: loadView “You should never call this method directly. The view controller calls this method when its viewproperty is requested but is currently nil. This method loads or creates a view and assigns it to the view property.” This insight allows us to segregate the and by utilizing the method. First, we will define a protocol that encapsulates the functionality of our future . View Controller loadView View protocol AbstractTestView: UIView {
    func bind(_ model: String)
} Now, let's proceed to implement the protocol for our . View final class TestView: UIView, AbstractTestView {
    
    private let content: UILabel = {
        $0.numberOfLines = 0
        $0.textColor = .red
        return $0
    }(UILabel())
    
    init() {
        super.init(frame: .zero)
        layout()
    }
    
    required init?(coder: NSCoder) {
        fatalError("init(coder:) has not been implemented")
    }
    
    func bind(_ model: String) {
        content.text = model
    }
    
    private func layout() {
        NSLayoutConstraint.activate([
            content.centerXAnchor.constraint(equalTo: centerXAnchor),
            content.centerYAnchor.constraint(equalTo: centerYAnchor)
        ])
    }
} Finally, we can implement our , which will communicate with the through the protocol. Controller View final class TestVC: UIViewController {
    private let mainView: AbstractTestView
    
    init(view: AbstractTestView) {
        mainView = view
        super.init(nibName: nil, bundle: nil)
    }
    
    required init?(coder: NSCoder) {
        fatalError("init(coder:) has not been implemented")
    }
    
    override func loadView() {
        view = mainView
    }
    
    override func viewDidLoad() {
        super.viewDidLoad()
        
        mainView.bind("Hello from MVC!")
    }
} For more intricate elements like tables, the business logic will remain within our . Here's an example: Controller extension TestVC: UITableViewDelegate, UITableViewDataSource {
    func tableView(_ tableView: UITableView, numberOfRowsInSection section: Int) -> Int {
        
    }
    
    func tableView(_ tableView: UITableView, cellForRowAt indexPath: IndexPath) -> UITableViewCell {
        
    }
} Now, let's connect our with the . Controller View let vc = TestVC(view: TestView()) It is crucial to maintain a clear separation between business logic and the view. Continuously ask yourself: "Does this piece of code handle the display of elements or work with data?" Initially, this may pose a challenge, but with time, it will become second nature.

Mastering MVC: Properly Cooking the Controller in iOS

I created a simple service for to test if the STUN/TURN servers are working properly. After installing and configuring a STUN/TURN server, most of the developers that are new to this stuff will ask themselves, how do I know if it's working properly? my WebRTC project So I made a snippet :) <!DOCTYPE html>
<html>

<head>
    <title>Test Stun/Turn Servers</title>

    <meta charset="utf-8" />
    <meta http-equiv="X-UA-Compatible" content="IE=edge" />
    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
</head>

<body>
    <h1>Test Ice Servers</h1>

    <hr />

    <pre id='ice' style='overflow: auto'></pre>

    <hr />

    <p id='ip'></p>
    <p id='stun'>🔴 The STUN server is NOT reachable!</p>
    <p id='turn'>🔴 The TURN server is NOT reachable!</p>
    <p id='err'></p>

    <hr />

    <script>
        const Ice = document.getElementById('ice');
        const IP = document.getElementById('ip');
        const Stun = document.getElementById('stun');
        const Turn = document.getElementById('turn');
        const Err = document.getElementById('err');
        
        const iceServers = [
            // Test some STUN server
            {
                urls: 'stun:stun.l.google.com:19302',
            },
            // Test some TURN server
            {
                urls: 'turn:turnUrl',
                username: 'turnUsername',
                credential: 'turnPassword',
            },
        ];

        // Print iceServers config
        Ice.innerHTML = JSON.stringify(iceServers, null, 4);

        // Test the connections
        const pc = new RTCPeerConnection({
            iceServers
        });

        pc.onicecandidate = (e) => {
            if (!e.candidate) return;

            console.log(e.candidate.candidate);

            // If a srflx candidate was found, notify that the STUN server works!
            if (e.candidate.type == 'srflx' || e.candidate.candidate.includes('srflx')) {
                let ip = /\b\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}\b/;
                let address = e.candidate.address 
                    ? e.candidate.address 
                    : e.candidate.candidate.match(ip);
                IP.innerHTML = '🟢 Your Public IP Address is ' + address;
                Stun.innerHTML = '🟢 The STUN server is reachable!';
            }

            // If a relay candidate was found, notify that the TURN server works!
            if (e.candidate.type == 'relay' || e.candidate.candidate.includes('relay')) {
                Turn.innerHTML = '🟢 The TURN server is reachable!';
            }
        };

        // handle error
        pc.onicecandidateerror = (e) => {
            console.error(e);
            Err.innerHTML = '🔴 Error: ' + e.errorText;
        };

        pc.createDataChannel('test');
        pc.createOffer().then(offer => pc.setLocalDescription(offer));
    </script>
</body>

</html>

A Simple Script to Test Whether a STUN/TURN Server is Working Properly

Since the dawn of machine learning, imbalanced datasets have been plaguing the workdays of anyone working on classification problems. In truth, handling an imbalanced data set isn’t that complicated if you know what you’re doing. In case you do not, I’ll introduce you to the concept of classification problems, talk about real life examples, and share efficient ways to handle imbalanced data. What Are Classification Problems in Machine Learning? For those who are just now learning about… machine learning… I’ll include this short introduction to classification problems. It will help you understand the general concept of the article better. Classification problems are problems that classify data to multiple categories (class labels). If only two such categories exist, then it’s called . For multiple class labels, we use the term . supervised learning binary classification multinomial classification The most common example of binary classification is when determining whether an email is spam or not. A classifier will label emails based on the sample data it was given. Usually, it will deduce that emails containing certain phrases such as the are more likely to be spam than not. infamous “Nigerian Prince” A multinomial classification algorithm would do the same, but would have to be more complex and capable of labeling emails as promotional material, news, or any other category you come up with. When and Why Do Imbalanced Datasets Occur? Now that we know what classification problems are, let’s focus on imbalanced datasets. When a single class label has a significantly lower observation rate than other class labels in the same dataset, we that dataset as imbalanced. label Using our previous email example, if the classifier encountered only a couple of spam emails upon examining thousands then you have an imbalanced dataset on your hands. In practice, about . This means that the classifier was either examining a set of emails that have already been curated, or it’s simply not working properly. 50% of all emails sent are spam How to Handle Imbalanced Data? There are dozens of way to manage imbalanced datasets, but I’ll focus on explaining the most popular ones. These can be used to deal with 99.9% of data imbalance issues you may encounter. But what about the 0.01%? How do I deal with those? Most people will never encounter such specific examples of imbalanced data. If you do, then you’ll need years of experience to build a custom solution. A simple article, no matter how brilliantly written, wouldn’t help you at all. Undersampling & Oversampling Undersampling means you are reducing the size of the majority classes. It’s a fairly simple method of handling imbalanced data but works great when dealing with binary classification. The most effective way to undersample is to keep all examples from the rare class and choose examples from the abundant class. randomly Oversampling works the other way around. You keep all samples from the abundant class but add several new copies of the entire rare class to the dataset. Similar to undersampling, oversampling is most effective when dealing with binary classification problems. Both methods ensure you create a new dataset that’s balanced and easier to work with. Your classifier will be able to learn more and better predict future outcomes by going through more “bad” samples. Redistribute class_weight I am not a huge fan of this technique because I find it overly complicated for no good reason. However, many folks would want to put me in the data science jail if I didn’t introduce it here so let’s take a quick gander. A parameter called class_weight is present in almost every popular classifier model library. The default value of class_weight is 0, so the model automatically assigns weight to different classes that are inversely proportionate to abundance. By skewing class_weight parameters and applying the log loss function as the cost basis, you can “punish” your classifier much more severely when it misclassifies a rare sample compared to misclassifying an abundant one. Change Evaluation Criteria While the above two techniques have dealt with data presented to the classifier, this one modifies the classifier itself. First, you need to decide what needs to be changed and why. This is done by using evaluation metrics. Simply put, evaluation metrics are what’s used to modify the quality of a machine learning model. While many exist, the five I believe to be are: absolutely essential Accuracy, Precision, Recall F1 Score AUC/ROC Categorical Crossentropy Log Loss Let’s use another hypothetical example. Our model is trying to predict whether Oprah is ever going to become the God-Empress of our planet. Hopefully, the model says every time based on how you built it. NO is the prime evaluation criteria, so that’s the first thing we’ll use to evaluate our model. Accuracy A = (TN+TP)/(FP+TP+TN+FN) TP - True Positive TN - True Negative FP - False Positive FN - False Negative By applying the formula above, you get 99.5%. Your model is 99.5% accurate. Was this helpful? Not in the slightest. You still have an imbalanced data set on your hands and you know that it’s accurate. But it doesn’t help you predict anything. comes next. Precision P = (TP)/(FP+TP) It’s used to check how many of the predicted positives were true positives. Since we never predicted a positive, it’s useless. Essentially, it says our model is not good but doesn’t help us whatsoever. On to . recall R = (TP)/(TP+FN) We use recall to check how many of the true positives have proper class labels. Since we got no true positives, we can’t use it. Next up we have the it’s a personal favorite of mine. If you wondered why I bothered writing out the examples for precision and recall above, we’ll get to that shortly. The F1 score is always equal to a number between 0 and 1 and it represents the harmonic mean of precision and recall. F1 Score, F1 = 2(P*R/P+R) When the formula is applied to our Oprah problem, we get 0. F1 being a fairly advanced evaluation metric, we can say that our model is completely useless for this particular set of data and decide to handle our imbalanced dataset problem by building a different model. Real-life Examples of Imbalanced Datasets If you’re a professional data analyst and you encounter a real life example of imbalanced data, it’s usually going to be worth looking into. A great example is encountered in . If you work for a bank and are trying to predict credit card fraud, you’ll encounter imbalanced data sets. The competition between fraudsters and machine learning models has been going on for a long time and each side keeps finding new ways to “outsmart” the other. finance In , predictive models are being used to better diagnose various diseases and have been saving lives for the past few decades. As you can imagine, the most infrequent diseases create imbalanced datasets and classifiers learn from those to help doctors with diagnoses that would rarely come to mind. health Finally, we have so many examples of imbalanced data sets in that I sometimes believe some solutions were created solely to help marketers sell more services. Customer turnover, refund requests and even legal action taken against companies can all create imbalanced data sets and getting ahead of those is invaluable to top companies. marketing Final Thoughts By the time I finalized writing this article, I deleted my original conclusion (yes, I always write that part first) and decided to write a new one. I did so because I was sure that this will be a short and sweet piece that’s easy to both read and write. However, I found myself constantly digressing trying to explain different machine learning principles, thinking of new examples, and the like. I even learned a few new things while fact-checking myself when making edits. I consider myself a machine learning veteran, yet every time I write an article like this, I remind myself what originally got me excited about it. It’s a bottomless pit (in a good way) of theories, problems, and solutions - all created for the betterment of mankind.

Imbalanced Datasets & How to Properly Deal with Them

What is email authentication? Simply speaking, you confirm your identity as a sender when you authenticate your emails. Email authentication is a collection of techniques that provide verifiable information about the received email. It verifies the domain ownership of the message transfer agent which participated in the transfer of an email message. Without authentication, spoofers and spammers have a far easier possibility to impersonate you and cause harm to your reputation. Your reputation? Yes, email authentication also benefits your email reputation. You improve email deliverability by authenticating your emails. With that, you reassure the email servers of your identity and get your emails delivered to primary inboxes without an issue. If you wish to improve , we recommend reading that article for comprehensive guidance. email deliverability How does email authentication work? While the entire email authentication process occurs behind the scenes, the are set by the senders. Those authentication policies define the regulations by which an email can be authenticated. The sender configures the email server and technical infrastructure to conform to those policies. authentication policies Once an email server receives an email it conducts the email authentication process by investigating the details of the received email and comparing them to the defined rules of the domain owner. Afterwards, the email is either delivered to the primary inbox, sent to the spam section or sometimes even flagged or rejected. Without any authentication, the delivered email messages are highly likely to be considered and get rejected. That directly affects the email reputation, which grants us more insight into the question mentioned in the section above - how does email authentication affect email reputation? spam The set policies for email authentication are separated into standards, and there are three prominent email authentication standards. And once you have set all three email authentication protocols - you will be able to send emails freely and confidently. What are email authentication standards? In email authentication, there are three standards used prominently, also called email authentication protocols or methods. Those standards are These standards are supplementary to the basic email transfer protocol, SMTP, as it does not include any authentication mechanisms. SPF, DKIM, DMARC. SPF and DKIM are used most commonly and they serve different purposes. SPF validates if the email was sent from an authorized . DKIM verifies if the content was not changed by any interfering server during the transfer. IP address And even though SPF and DKIM are the most common, DMARC is the most sophisticated one. DMARC conducts both SPF and DKIM verification processes and performs an additional procedure. We shall speak of that procedure further down the article. We suggest reading this article for a more extensive guide on . For now, let us discuss some details on the available email authentication protocols, also called methods and standards. email authentication Sender Policy Framework (SPF) SPF is a simple TXT file included in the of the domain. When creating an SPF record, domain owners specify the IP addresses authorized to send emails on behalf of that domain. DNS records The server receiving an email runs a DNS verification of the return-path address. After locating the SPF record within the DNS it scans the list of authorized addresses. The result of this process is positive if the address is present within the SPF record in the DNS of the server. In an opposite situation the email triggers a soft fail alert. The email is then marked with an authentication failure. A server still accepts the email but has a reason to discard it. SPF is a widely adopted email verification method and is considered essential for any SPF increases the chances of valid email delivery. email sender SPF record example If you were to search the domain and locate the DNS records, an SPF record should look so: v=spf1 include:_spf.google.com include:mailtrap.io include:_spf.reply.io ~all Let us analyze this SPF record and learn how to build one ourselves: is the beginning and is referring to the version of the SPF. SPF1 is the most widely adopted version and is also referred to as the SPF Classic v=spf1 is for specifying the domains used to send emails from. include: concludes the record by specifying all other unauthorized domains. ~all Some records include IP addresses instead of server names, like so: v=spf1 ip0:000.00.00.0 ip6:000.00.00.00 include:_spf.google.com include:mailtrap.io include:_spf.reply.io ~all How to properly authenticate with SPF? If you are using a Mailtrap Email API/SMTP service, then you already have a prepared SPF server you can simply copy and paste in your DNS records. Otherwise, when building an SPF record there are many details to consider. SPF comes with qualifiers, and the details of those qualifiers define the purpose of the SPF record: is used to trigger a soft fail, accepting an email but marking it with SPF fail. ~ is used to reject emails. – is used to trigger a pass, allowing email messages. + is a neutral qualifier, indicating an absence of policy. ? The plus mark qualifier is also automatically assumed if there are no qualifiers indicated by the end of the DNS record. And all of the qualifiers above are simply suggestions and can be entirely ignored by the server if needed. Domain Keys Identified Mail (DKIM) DKIM is also considered an important email authentication method to possess. It is a digital signature attached to the transferred email and is validated by the receiving server upon arrival. Upon each DKIM validation, a public key is generated and added to the DNS records of the domain. The DKIM signature is encrypted with an individual and a private key. The signature itself contains details of the email, such as the email header and body. The entire DKIM signature is created upon the creation — do forgive the tautology — and delivery of the email itself. Upon arrival, the server compares the public key from the DNS records to the private key. If everything matches - the test is passed. But if the email message went under any modification the DKIM test fails and the email message is discarded. DKIM record example DKIM record does look more complicated than an SPF record: DKIM-Signature: v=1; a=rsa-sha256; d=example.net; s=newyork; c=relaxed/simple; q=dns/txt; t=1117574938; x=1118006938; h=from:to:subject:date:keywords:keywords; bh=MTIzNDU2Nzg5MDEyMzQ1Njc4OTAxMjM0NTY3ODkwMTI=; b=dzdVyOfAKCdLXdJOc9G2q8LoXSlEniSbav+yuU4zGeeruD00lszZ VoG4ZHRNiYzR As previously, let us dissect the entire record to understand it better to build ourselves: is a selector and labels the associated DNS. s= hashed email message body bh= the digital signature in a hashed format b= These are the most important three. It is also important to understand, that a domain itself can have multiple selectors. Multiple sectors serve different purposes, such as an email for newsletters, promotions, sales, etc. How to properly authenticate with DKIM? DKIM is highly advised to be configurated with SPF. DKIM focuses on a singular aspect of phishing prevention and provides no mechanics for indicating to a server how to treat failed attempts. Authenticating emails with both of these email authentication methods, DKIM and SPF, provides a far more secure email exchange and information on where the email message originated from, and where it has been modified. For proper email authentication with DKIM, it is important to update your private and public keys once or twice a year. verification Domain-based Message Authentication Reporting and Conformance (DMARC) DMARC is not considered an authentication method, as it builds upon other email authentication methods, such as SPF and DKIM. DMARC adds domain alignment, which is an additional layer of security. The domain alignment procedure verifies SPF and DKIM. This procedure can either be “relaxed” or “strict”. Choosing the correct variant is particularly important. If you select “strict”, you need to make sure to include all transactional and marketing domains to avoid rejecting important emails. To clarify, DMARC does not require SPF or DKIM, but having all three is considered a best practice to properly authenticate your emails. The trick would be to select a “relaxed” method, but use to control which emails you interact with and which you do not. With DMARC you are able to configure additional policies for failed verifications. That helps with the generation of email performance reports. email suppression lists DMARC record example DMARC record example is as follows: v=DMARC1; p=none; rua=mailto:postmaster@mailtrap.io; ruf=mailto:postmaster@mailtrap.io; fo=1; adkim=r; aspf=r The tags in the record are more extensive than we have seen in other two email authentication methods: is the version of DMARC. v is the policy of DMARC. p email addresses receiving reports about emails failing DMARC check. rua email addresses receiving individual and instantaneous reports about failed checks. ruf optional domain alignment for DKIM. adkim domain alignment for SPF. aspf How to properly authenticate with DMARC? Even though I mentioned configuring SPF and DKIM with DMARC is the best practice, you might find it surprising that it is not the most popular one. And that is the precise reason spoofing is often a problem. If you have already configured SPF and DKIM, you must add DMARC to have the strongest email authentication method for your domain. However, it is worth taking time when configuring DMARC to ensure proper authentication. If the configuration was flawed the DMARC tests fail. Conclusion And now you know the tips and tricks on how to properly authenticate your email. Many might consider email authentication a complicated procedure, yet proper guidance can help significantly. We understood how email authentication improves email deliverability. And with improved email deliverability you have better control over . It is important to take time in understanding email authentication and patiently configure it within your business. It improves your email reputation and asserts healthier business communication. soft and hard bounces

How to Properly Authenticate Your Emails

I’ve led many projects with deadlines over my career, but one stands out from how much I learned about planning. My team needed to redesign all five applications within three months. The problem was the sheer amount of work. We needed to re-architect some core UI components and update a ton of code. We only had three people on the team. This is a lot of work for a small team. We were going to miss the deadline. But we got it done. Ahead of schedule. How did we do it? We couldn’t change the amount of time, but there were two other options. We could add people or reduce the project scope. Both are costly. Adding people disrupts other teams and changing the scope may disappoint stakeholders. So how do figure out which lever to pull to hit the project deadline? We created a plan. In this post, I’ll explain what you need to do when planning for projects with deadlines. With a great plan, you’ll be able to advocate for expanding the team and cutting scope. So where do you start? Disclosure: This article includes affiliate links to any books I reference. I may receive compensation when you click on links to these products. Planning Projects with Deadlines The first step to managing a is to answer two questions: projects with deadlines Can we add people to the project to hit the deadline? If we add people, can we even meet the deadline? To get answers, we need to figure out the project’s effort. So grab a spreadsheet and write down the to-do list of the entire project. Don’t worry about user stories or project ceremonies. Just create a to-do list of you’re aware of. If a project area depends on someone else’s work, go find that person and have them write their to-do list. everything Spreadsheets work GREAT for scoping out a project with a deadline: They’re simple to write out a list of things. You can add metadata to each to-do item You can analyze and manipulate the data. Write out everything that you know at this point. It’s ok if you discover unknown scope later. This phase of a project with a deadline is about gathering data and advocating for more resources. Essential Data for your Project Plan Now it’s time to add metadata to each task in your project’s spreadsheet. I recommend the following: – Which systems or product areas will someone work in for this task? For example, the sign-up feature, data filters, a settings page, etc. This information is CRITICAL for later. Project Area – Make a of how many days it will take for ONE engineer to complet’e a task. – code reviews, QA work, deployment costs, etc. Stakeholders need an idea of when the project will ship. Effort Estimate rough estimate Include non-coding work – (Optional) Add special context for a task that others may need. For example, a hard dependency from your infrastructure team to provision a server. Notes With this metadata, you can create some data-driven summaries of the project’s work: Use the SUM() function to calculate the total number of days needed for the project. You can even calculate the effort of specific project areas. Color-code each row by the project area to get a quick understanding of where people will work. NOW we can start answering those expensive questions from earlier. Want to learn how to become the successful leader your team needs? to learn how. Sign up for my newsletter Will Adding People Help Hit Deadlines? Many software engineers have internalized the famous saying from : The Mythical Man-Month Adding manpower to a late software project makes it later. Brooks, Frederick P. The Mythical Man-Month . Addison-Wesley. While true for some projects, there are times when adding people will help a lot. The trick is finding how engineers can work on projects with deadlines. some concurrently “Working concurrently” means finding work where an engineer can work independently from teammates. These independent project areas are exactly . Engineers can start writing code and not worry about coordination costs. where you want to add headcount Remember when I told you to identify which project area a task contributes to earlier? Those project areas highlight where people can work concurrently. Say your team has two engineers and you identified four project areas. You can add two more engineers to the project and increase velocity. You now have one engineer independently working on each project area. In this situation, your new project estimate is the area with the most effort. What may have taken four months to complete may now be one. Your stakeholders would love to hear this news. Can We Cut Scope from Projects with Deadlines? So we’ve added headcount to the project, but let’s assume you STILL won’t hit the deadline. What do you do? Reduce the project’s scope. At this point, you need to meet with stakeholders and walk them through your plan. Your spreadsheet is critical here. It anchors the discussion and shows that you thought through every aspect of the project. Explain each requirement’s effort, your estimates, and how you outlined the work. You can expect two responses from your bosses. First, they’ll see they asked for too much and redefine the scope. You’ll learn what’s exactly needed for the deadline. There may be follow-up work after the date, but there’s less scope for the deadline. The second outcome will identify the wrong assumptions you made. Maybe you thought you needed to with five-9s of availability. Turns out, that’s not important. on your project plan highlights spots where extra effort isn’t needed. over-engineer a system Getting feedback Cutting this scope will be an immense help. The Last Option (Seriously…avoid this) Adding headcount and cutting scope will address most problems for projects with deadlines. But what if the team STILL won’t hit the deadline? . Is the deadline arbitrary or externally mandated? What happens if the team misses the date? Ask why the deadline exists Understanding HOW the deadline came to be may unveil other options. If there are no other options, then working late (or ) is your only choice. Asking a team to work overtime is the LAST thing you should consider. The team will lose a lot of trust if you don’t approach this topic carefully. crunching Sometimes, it’s a reasonable ask. The is one instance, but asking a team to work overtime needs to be a rare event. Asking once every few years is all you’ll get before people start quitting. Log4Shell incident Conclusion When your team has to work on a project with a deadline, the first step is to create a plan. Make a spreadsheet and outline the to-do list of every task needed to complete the project. Your job is to answer the following questions: Can we hit this deadline with the current team? Where can we add headcount to the project? What scope can be removed? But this exercise will do more than answer these questions. By outlining the project’s work, you can lead it. You have the map of what the team needs to work on. You can coordinate how everyone will work. With a plan in place, it’s up to you to ensure the team gets to the finish line. Good luck. Want to learn how to become the successful leader your team needs? to learn how. Sign up for my newsletter Also Published here

How to Properly Estimate Tasks So You Never Miss Another Deadline

Self-service business intelligence, or BI, has been on the to-do list of many organizations for quite a while. Marketed as a tool that allows users from non-technical backgrounds to get insights at the pace of business, self-service BI, however, is leaving many organizations disappointed when it comes to implementing it practically. Failure stories abound, with companies never getting what self-service BI has originally promised. That is freedom from IT for line-of-business users to create powerful and accurate reports to drive business growth. In this blog, you will find out what self-service BI exactly is, why organizations fail at it, and what steps your company should take to . implement a successful self-service BI solution What self-service BI is Self-service BI definition Self-service BI is often defined as a form of BI that uses simple-to-use BI tools to allow non-tech-savvy business users (sales, finance, marketing, or HR) to directly access data and explore it on their own. Self-service BI differs from traditional BI that is owned by the IT or BI department as a centralized function. In the traditional approach, it is these teams that are in charge of everything. They prepare the required data, store and secure it, build data models, create queries, and build visualizations for end-users after collecting their requirements. The idea of self-service BI is closely related to data democratization that is focused on letting everyone in an organization access and consume data. The ultimate purpose is to generate more insights at the organization level and drive better business decisions. Key benefits of self-service BI — Shifting control to end-users means skipping time-consuming stages of the traditional BI process. In self-service BI, end-users don’t have to wait for days or even weeks until their report finally goes live after getting through elicitation and approvals. They also don’t have to deal with the tedious change request management process when realizing that more visuals are necessary. This is because they can chop, tweak, and add data on the fly to uncover important trends, patterns, or anomalies. Faster time to insight — By empowering business users with thorough domain knowledge to perform their own data analysis on an ad-hoc basis, self-service BI produces better-quality insights while freeing the IT or BI teams from handling routine tasks related to data. Instead, these teams can focus on harder problems like setting up data pipelines to get cleansed and transformed data to the right destination at the right time and maintaining important data governance processes. Improved operational efficiency — Apart from optimizing IT and BI capabilities for time and cost savings, many self-service BI adopters take a step further. They arm subject matter experts with knowledge and tools for performing . In other words, they raise citizen data scientists who know how to generate ML-driven predictions critical to business. With data science talent coming at a hefty price tag, this kind of investment is probably one of the best a data-driven company can make. Cost reduction advanced data analytics Core features of self-service BI tools To enable the powerful benefits of self-service BI mentioned above, self-service BI tools should have the following essential features: that enable self-service BI tool integration with databases, CRM, ERP, marketing analytics, finance software, and other on-prem and cloud systems to serve analytics needs in the most efficient way. Data connectors that range from book-quality canned reports with customizable settings to ad-hoc drill-downs while allowing users to schedule distribution or divide the results into subsets for different audiences. Vast reporting capabilities that allows users to select data fields and visuals and drag and drop them into report canvas for exploration and storytelling. Intuitive drag-and-drop or click-based interface that simplify the process of creating dashboards based on user preferences and needs. Data visualization templates Many organizations take their self-service BI to the next level by enriching it with capabilities in and . Augmented analytics platforms enable users to discover more data, evaluate uncharacterized datasets, and create what-if scenarios. This way, business can react to its evolving needs as quickly as possible, achieving the utmost nimbleness. data science machine learning Why organizations fail at self-service BI 1. Unrealistic expectations An organization that just starts throwing data at novice users is facing a serious risk of poor-quality reports. It will be very lucky if these users with different qualifications wind up with non-misinterpreted data without first learning the basics of reporting. For instance, a happy user creating their first report on total sales in a historic period might end up with average numbers instead of a SUM, knowing nothing about default aggregations for various measures. Or on the contrary, they can submit inflated numbers. There is also risk of data inconsistency that might affect weighted averages when they need to be displayed with different levels of granularity. Further on, a non-power user might rest satisfied with just a casual analysis that has supported their initial beliefs. The confirmation or cherry-picking bias trap is not something an untrained user is necessarily aware of, especially when under pressure to explain a certain pattern. 2. Reporting chaos Self-service BI doesn’t mean zero IT involvement. Letting users toy around with data with no governance from IT usually leads to reporting anarchy. With no governance, there could be redundant reports from different users working in silos and delivering the same analysis or reports from different users analyzing the same metrics but using different filters and hence delivering conflicting results. Reports from different departments can rely on different naming conventions for quantity, value, or time or use the same terms but not necessarily the same definition. Multiple versions of the same database, errors in databases that are never fixed, the creation of objects used only once … The list is endless. Governance is not something that a data-driven organization can boycott in the world of self-service. No matter how badly a company wants to free users for conducting their own analysis, IT still needs to be involved to maintain high data quality and consistency. 3. Lack of adoption Truth is, not everyone likes to work hard. Most business users just want a simple dashboard that will give them the numbers. Valuable insights, however, often lie levels deeper that go beyond plain business performance analysis. Another psychological factor that may hold back an efficient self-service BI is resistance to change. It is not uncommon for many organizations in the early stages of their self-service BI journey to see frustrated business users coming back to BI or IT to request a report as they did in the good old times. Older approaches are safer. Unfriendly self-service BI environment setups also might be a problem. What may seem for IT or BI teams to be an easy-to-use tool for collecting and refining results can have an overwhelming and demotivating amount of features for a casual user without technical skills. Pivotal tables and spreadsheets might be dull, but users when getting stuck. are quick to revert to them 10 tips from ITRex on how to implement self-service BI successfully Below is a list of essential takeaways from ITRex experience in building efficient self-service BI tools for both smaller business and large companies, including for the world’s leading retailer with 3 million business users (read more about‌ this project ): here . You first need to define what you want to achieve with self-service BI, be it as simple as reducing delayed reports or providing data access organization-wide. Self-service can mean anything to different people, so you should be clear about your project. It’s also important to understand early the scale of implementation, the types of users, their technical proficiency, and your expectations of deliverables. Set your self-service BI strategy . You should wrap your head around what your stakeholders look for in data and what their data-related success metrics are. Interview them to collect their functionality, usability, user experience, and other inputs. Then continuously ask them for feedback as you iterate. Apart from making sure you build a relevant self-service BI tool, you will also give your stakeholders a sense of ownership and improve their engagement. Keep all stakeholders on board throughout the project . This is also essential. Your IT has all the information on your data environment, existing data sources, data governance controls in place, and data access management. They will help you choose or build a self-service BI solution that is easy to maintain, monitor, and manage in terms of user access and integration of new data sources. Involve the IT department . Self-service BI governance encompasses the following: Set up robust governance Data governance policies and procedures to ensure your data is consistent, complete, integral, accurate, and up-to-date. Here you will need to and adopt leading practices in master and metadata management as part of it develop a broader data management strategy Governance of business metrics to define them uniformly across your self-service BI environment and rule out any deviations Governance of reports to set a procedure for their quality validation Data security to define who gets access to what data in your self-service BI and establish data lineage . There’s no one-size-fits-all strategy. Your users have different needs and skills your tool should precisely cater for. You will probably need to balance between flexibility and sophistication to allow your users to ask new questions while staying self-reliant. A custom self-service BI solution will make it easier to achieve. Select the right tool . A single source of truth is implemented as part of solution architecture to enable decision-making based on the same data. For this, companies build a or another kind of central repository that provides a 360-degree view of all their data from multiple sources and makes data access, analysis, enrichment, and protection much simpler and more efficient. It’s worth the investment. Establish a single source of truth data warehouse . Three types of training programs for end-users are a must: 1. data analysis and visualization, 2) the basics of joining data and building data models, and 3) continuous peer-to-peer training. Educate users . It will help a lot if you either establish a center of excellence or have an expert community on Slack or Teams so that your end-users know where to go to fill in gaps in knowledge. Build a community . They will help drive engagement by increasing access to data for users with no analytical background and providing oversight as needed for better-quality reporting. Consider embedding BI specialists in business units . Choose a limited environment for starting your self-service BI project and build from there using an agile approach. This way, you will fix problems early before scaling up. Start small Watch this two-minute video of a project from the ITRex portfolio to learn how self-service BI augmented with AI can drive efficiency gains for a large enterprise if done right. https://youtu.be/KVARUwIRFZE if you consider embarking on a self-service BI journey. With their battle-tested approach, we will help you avoid common pitfalls while bringing your project to success. Drop ITRex team a line

Stories, News & Mentions about PROPERLY

Properly’s Stories on HackerNoon