paint-brush
How to Get Started With Effective Ad Automationby@hackerclo6xmwpi00003b6sx801itdo

How to Get Started With Effective Ad Automation

by Maria ADecember 15th, 2023
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

The article explores practical strategies for automating user acquisition, using Facebook examples from the author's experience at Scentbird. It emphasizes considerations for optimization at different stages of the ad cycle. It concludes by highlighting the value of automation in process establishment and cost optimization, underscoring the importance of company and department maturity in the implementation process.

Company Mentioned

Mention Thumbnail
featured image - How to Get Started With Effective Ad Automation
Maria A HackerNoon profile picture

In this article, we’ll cover some particular approaches to automation processes using Facebook examples from my previous work at Scentbird.


Scentbird utilizes the subscription model and business growth directly connected to active user base growth. Business margins are needed to optimize customer acquisition costs. For any startup, it is critical to demonstrate steady growth with all the KPIs met, primarily including ROI.


In Scentbird, I began as a media buyer and was responsible for user acquisition on Facebook. I ended up managing 100% of performance traffic with up to a $2M monthly ad budget.

The ideas on why and when it’s the best time to automate are covered in the article “Why and When to Automate Media Buying”.


Automation itself sounds like an obvious and simple solution in the IT epoch; however, there are several pitfalls that we should keep in mind to avoid spending excessive resources like time and money.


Here, we’ll discuss several practical things:

  • On what channels can we automate ads?
  • Where to start with automation
  • Which parts of the ad cycle may be automated?
  • Several particular approaches to automation.


On what channels can we automate ads?

This question usually has a pretty simple answer. You can at least partially automate any channel with an API, like Meta Ads, Google Ads Campaigns, DSPs, Apple Search Ads, etc. Usually, if the channel has an API, you’ll have to perform code automatization. However, looking for services that provide no-code automation for these channels may be a good idea.


These services may go with relatively high prices, like RevealBot or Smartly.io, or with a bit lower price (or even free), like Luna for Apple Search Ads, because it is an Apple partner with specific terms with Apple.


A general rule of thumb is that the larger your budget, the more advantageous it becomes to hire a developer—whether on a part-time basis or for a specific project—to build and support the automation system for your company.


Where to start?

As we discussed in the previous article, it makes sense to consider automation when at least:

  • There are one or several scalable digital marketing channels with sufficient LTV/CAC ratio;
  • Routing tasks to manage each channel takes at least 2 hours per day.


The keyword here is “routine”; thus, you are sure this channel is here with you and not going anywhere soon. Also, you or your team perform relatively the same tasks frequently.


Step 0 in the automation process separates the routine tasks (daily/weekly maintenance) from the creative ones (A/B tests, new placements tests, etc.) AND begin to perform them in the same way each time. The best way here is to write an algorithm (a process) for performing these tasks and stick to it for a while to ensure that it’s stable and you haven’t missed anything.


Which parts of the ad cycle may be automated?

A typical ad lifecycle consists of the following parts:

  • Launch;

  • Active running;

  • Turn off.


Let’s discuss each part.


Launch

This is the trickiest part. Here, you should pick the creative, the target audience, and the budget.


Usually, it’s more complicated to automate because:


  • Most no-code solutions haven’t integrated automation yet.
  • Even if they did, you can’t be sure they have all the needed features. Thus, the solution might not be flexible enough.
  • The logic may be pretty complicated.
  • If you automate this part yourself, this is the place where you’ll need the most maintenance from the code perspective in the future.


The audience grid approach will be applicable from the target audience's perspective. It has different names in different companies, but the logic is simple and almost the same.


In particular, you need to group small target audiences into several large ones based on each small audience's performance. You can determine the performance after running several audiences simultaneously and check the statistical significance of the difference.


For example, for Facebook lookalike audiences, the lower the percentage, the better performance is. The performance may vary significantly between different publishers for DSPs like Vungle, ironSource, or Google DV360. This data is highly business-specific, so although you know how the audience is distributed in other companies, it’s better to recheck each case.


Having the audience grid, you can add it to code or no-code automation. It’s a good idea to review the grid structure occasionally. Still, the main idea is to fix it and use it similarly until significant algorithm, product, or inventory changes happen.


The next important thing is the budget allocation. For different channels, budget policy will vary a lot. Let’s discuss the Facebook example. You may notice different patterns after running media buying in the channel for a while. For example, there was a time when a strategy was launching an ad set on Facebook with a small budget and then scaling it when it appeared to “work well” yielded better statistically significant performance than launching an ad set with a “working budget” right away.


Now, CBO simplifies this task, but in any case, you should have a typical process for default budget allocation, scaling, and downscaling, depending on the performance. Having this process in place, you can determine the budget you are launching with.


And the last (but not the least) part is the creative.


For creatives, you typically should have two cycles:

  • The “main” cycle, where you put creatives that are proven to perform “best” - a part of your evergreen campaigns;
  • The “test” cycle, where you test your new creatives - a part that may vary depending on the seasonality, performance, number of new creatives, ROI targets, etc.


Depending on the channel, algorithm, and product-market fit strength, the creative tests may be either profitable or unprofitable. Furthermore, depending on a particular channel capacity and the number of new creatives for testing, you can either have the opportunity to test all the creatives on this particular channel at the required pace while maintaining the health channel performance (ROI) or not. So you can consider different approaches for test creatives allocation.


Let’s consider the general thoughts about the creatives’ tests that may be useful:

  • You may fix the specific audiences, where you keep a small number of best creatives for benchmarking, add the new creatives for tests, and then remove them after the test.


  • If you perform media buying worldwide, you can define the regions with similar performance, divide the creatives between these regions, test only in one region, and consider that these creatives will perform similarly in another region.


  • The same idea stands for the channel in case you can group the channels by performance.


  • Also, remember that not all the channels provide the instruments for creative A/B tests like Facebook or Google DV 360; thus, you may want to test your creatives on these channels and use “good” creatives on the other channels without tests or with partial tests.


  • If the channel automatically reallocates the budget for the best creatives (like Facebook or Twitter Ads do). In that case, you may add new creatives to the running ad sets along with the best creatives and see whether the algorithm reallocates any significant share of the budget to the new creatives (at least 5%).


Also, there should be a process to remove the burned-out creatives from the rotation.


Combining all these three processes, you can create a decision-making process in pseudocode and understand what functions you need to automate your process and which parts can or can not be automated depending on the no-code tool you are using or the code you are using to write.


The automated launch process will help you maintain the daily ad budget and avoid underspending. It’s especially important when monthly spending on each channel is part of your team's KPIs.


Lifecycle

Lifecycle is about managing your ads during the time they’re on. On most channels, you can change the bid and budget. In rare cases, you can control things like frequency or other parameters.


Usually, this is the simple part. The bid helps you to control CAC; this is the limiting condition on the lower side (if the bid is too low, you may set a higher budget, but the algorithm won’t spend it). You have the budget to control overall spending; this is the limiting condition on the high end (if the bid is reasonable, the algorithm will spend the amount approximately equal to your budget).


Thus, you can create rules to adjust these settings depending on performance. Most no-code platforms will successfully perform this kind of management. Here, you can perform many experiments with strategy and particular details, but still, it’s a simple part from the automatization perspective.


Turn off

This process has a medium complexity compared to launch and maintenance from the automatization perspective. And this is the spot where you can achieve quick and significant results. Most no-code automation tools also let you configure such kinds of rules easily and quickly. Furthermore, for example, Facebook already has a built-in tool to set up rules like that.


Here, you should formulate what is “good” performance and what is bad. In edge cases, it’s obvious, but sometimes we can’t tell yet.


And this is a challenging part: what is the earliest time (and under which circumstances) when we can, with sufficient probability, say that the performance of this particular ad set is not good enough and decide to turn it off?


The more precise you can differentiate the “good” from the “not good enough” ad set, the more the CAC optimization will occur. You’ll often need several iterations to develop a decent solution, but it’s worth it.


Conclusion

Here, we discussed some practical approaches and details of automating user acquisition via channels like Meta Ads, Google Ads Campaigns, etc.


Hope these examples will clarify some aspects of crafting the automation process that helps the company grow without additional HQ. This case describes the key points that will allow you to navigate better.


At the end of the day, automation can bring the company a lot of value in establishing processes and optimizing CAC and FTE. However, it’s essential to consider the company's maturity in general and the marketing department's maturity in particular before you even come up with this idea because you need to build it before you can automate it.