How To Prioritize Data Science Projects

Prioritizing projects is an important part of any managers daily tasks.

However, as a data science team manager or lead, it can be very difficult to assess which projects should actually be done and which projects should be put on the back burner. Between translating business needs into technical requirements and managing expectations, managers can get overwhelmed by all the new requests coming their way.

Being a part of a large organization typically means a data science team might be juggling the requests of almost every department. Especially if there is only one data science team that is attempting to wade through all the projects sent to them from the entire company.

The issue is that every manager thinks their problem is the most important. From their perspective they are somewhat right. Only somewhat…

The company is a large organism and what might seem like a good move for one manager, might not be the most optimal use of resources for everyone else. So how does the company ensure that the right projects are being done at the right time?

The solutions haven’t changed too much from other technical work an management. However, it seems like the problems persist where managers squabble over who should get what resources or which projects should be cut.

In order to help, our team has listed out a few tips we have seen to be helpful in the past. Some are easier to implement than others. In the end, it does depend on the size of a company and how the data teams are structured.

Here are a few of our tips…

Create A Committee Of High Level Directors

Creating a committee has pros and cons and we would like to discuss them because although it can be easy to say. Creating an effective committee that actually drives innovation rather than stifle it doesn’t always happen.

Pro

Having a balanced committee means you allow multiple perspectives to come into play. This improves the typical bias when a single department approaches the data science team with their demands. When a company has a solid committee that attempts to look at the entire landscape of the company and their customers needs they can more clearly assess the their current standings from a higher level. Instead of getting stuck in the weeds.

Con

Committees cause drag as sometimes teams get stuck waiting while committees are discussing the next best move. In addition, new requests might get stuck in committee for a long time before they are ever addressed. This can stifle innovation.

A Middleish Ground

There are a couple of methods to mitigate the pros and cons of having a committee. One of the ways that can be effective is to allow the data science team to have some personal autonomy on smaller projects. This will improve the team’s ability to move quickly and take away the risk of having small projects get bogged down in the process. In this case, a strong data science lead or manager makes an immense impact. An experienced data science manager can assess if the project really is small or big and know how many resources will be required.

Projects Should Be Driven By Company Goals Not Technology

I recently heard a director state that they had a new directive this year. Their directive was to use tensorflow. That is like saying you have a new directive this year, and it is to use a hammer but your actual task is to paint the Mona Lisa (I am sure there actually is a video of someone on YouTube painting the Mona Lisa with a Hammer).

There is a plethora of new technology that can tantalize our fickle minds to be drawn to use them. Even with all these cool new tools, it does not mean that they all need to be used. Sometimes a simple solution will provide much better ROI and competitive advantage compared to using the newest tech on the block. So when picking projects, especially with all the hype around data, make sure it is driven by your company’s goals and not technology.

Use Data Not Emotions

Many of us are at fault for getting a crystallized idea of what we think the right next step might be without first questioning the data. In order for anyone to make an informed decision, people need accurate facts. Even with facts, there is the plausibility that some managers might already have a preconceived idea and instead of trusting the data they are provided, might instead go with their idea anyways.

Design A Company That Encourages Discussion From Individual Contributors.

Individual contributors sometimes get pigeonholed as only doing their work and not looking at the big picture. Sometimes this is by choice, other times this is because of the process they have to work in. This does not mean individual contributors don’t have a good perspective on what could be a very effective project. Especially for internal initiatives. If multiple analysts and individual contributors are struggling with a problem , then it might be good problem to solve. However, this is an issue directors might not know about. Thus, it isn’t a bad idea to reach out and see if there are any basic projects that could be done to help analysts perform better work.

Truth be told, every company does operate slightly differently and the best way to manage projects does vary. If what your teams seem to not be working, then it might be time to re-calibrate the processes. It will probably be a hybrid of many of the tips above and some of your companies personal standards.

Call-To-Action

If your team needs help prioritizing its data science or developing an director level data science plan, then reach out today! We would love to help move your team in the right direction.

Data science initiatives can costs millions of dollars and we want to ensure that all your resources are being utilized to the maximum effect.

How To Grow As A Data Scientist

Boosting Bagging And Building Better Algorithms

How To Survive Corporate Politics As A Data Scientist

8 Top Python Libraries For Machine Learning

What Is A Decision Tree