Are you wondering how to scale the impact of a data science team? Or maybe you want to hear more about how to select projects that have a broad impact across your company. Well either way, you are in the right place! In this article we tell you everything you need to know about scaling the impact of a data science team. In particular, we focus on solutions that can be applied without increasing the headcount or resources allocated to a team.
We start out by discussing the reasons that you should care about scaling the impact of a data science team. After that, we take some time to discuss high-level strategies that can be employed to increase the impact of a data science team. We follow this up with examples of specific tactics that can be employed for each strategy. Finally, we synthesize everything we have discussed into a single principle that can be used to guide decisions.
Why should you care about scaling impact?
Why should you care about scaling the impact of data science projects? The reason will vary based on the role you play in the company. In this section, we will discuss reasons that individual contributors and managers alike should care about scaling the impact of the projects they work on.
As an individual contributor, creating scalable solutions that have an outsized impact will enable you to progress faster in your career. Promotions are often tied to the scale of the impact that you have on your organization, so building out highly impactful projects will help you get promoted faster. Building out highly impactful projects will also increase the amount of trust the business has in your work, which will generally grant you more freedom to choose what you work on and authority to influence important decisions.
There are many reasons for data science managers to care about scaling the impact of their team. If your team is viewed as highly impactful, then it will be easier to get new headcount and resources for your team and expand the scope of your team. This will simultaneously contribute to your own career progression and make it easier to find growth opportunities for existing team members. If your team is seen as a critical team that produces strong output, they will also be in a better place in case of events like layoffs or restructurings.
Strategies for scaling the impact of data science projects
What are some strategies that can be employed to increase the impact of a data science organization? Here are a few strategies for increasing the impact of data science projects and data science teams.
- Build solutions that solve multiple business problems. One of the best ways that you can increase the impact of a data science team is to focus on building generalizable solutions that can be used to solve multiple different business problems. This increases the number of problems your team is able to solve without increasing the amount of work the team needs to complete.
- Maintain focus on your most important work. The next way to increase the impact of a data science team is to ensure that data scientists are able to maintain focus on long term solutions. This way the team will focus more time on scalable solutions that solve multiple problems and less time on ad hoc solutions that are only used once.
- Build reusable data components. Another way to increase the impact of a data science project is to create reusable data components that can be utilized by other data scientists and data professionals. If you consistently build components that can be reused by other data scientists, this will reduce the average time it takes to complete a data science project.
Tactics for scaling the impact of data science projects
What are some specific tactics you can use to scale the impact of data science projects? Here are some examples of specific tactics that ladder up to the strategies we mentioned in the previous section.
Build solutions that solve multiple business problems
- Seek out solutions that solve multiple problems. The first way to increase the impact of your data science projects is to seek out projects that will solve multiple problems across the business. In order to do this, you need to have a broad understanding of problems and priorities that exist across different areas of the business. In order to develop this understanding, you might need to meet with stakeholders across your business unit and adjacent business units to understand their priorities. Ask them what some of the main challenges they face are. Ask them what information they wish they knew about their customers or the areas that they oversee. After you meet with stakeholders from a few different areas, you can synthesize all of the information you have gathered and determine whether there are overlapping pain points or pieces of information that would be useful to have across multiple different areas.
- Reframe projects to solve multiple problems. Another tactic you can take is to reframe existing problems that you have been tasked with solving to create a solution that can be used to solve many different problems. For example, imagine you worked for a website in the travel industry and you were asked to build a solution that would identify customers who would enjoy staying at a particular hotel. A more general way to reframe this problem would be to understand the characteristics of that hotel and broadly identify users who prefer destinations with those characteristics. For example, if the main selling point of the hotel was that it is right on the beach, it might be useful to identify which users favor beach vacations. If you could build out a solution that identified which users liked to go on beach vacations, this solution could be used broadly across the website. You could use it to rank hotels that are on the beach higher in search rankings, preferentially surface ads that have beach scenes on them, and suggest flights to countries with large coastlines.
Main focus on your most important work
- Say no to work that is not impactful. The first tactic for maintaining focus on important work is simply saying no to any requests that will not be sufficiently impactful. Note that the applicability of this tactic may depend on the strength of your relationships with your stakeholders. If you are working with a new team that you have not established trust with, it might make sense to perform a less impactful task to build trust. If you are working with a team that you have a strong relationship with, you should feel more empowered to say no.
- Introduce a low friction barrier to stakeholder requests. If you find that your team is getting distracted by frequent requests from stakeholders, you may be able to reduce the number of incoming requests by introducing a low friction barrier into the request process. This simply means increasing the level of effort that it takes to submit a request by a hair. For example, if your stakeholders usually send their requests over in one-off slack messages, you could update your process to require them to fill out a formal request form and answer a few questions about their request. The idea here is that if a request truly is important and the data that is being pulled is going to influence business decisions, then a small amount of friction is not likely to prevent stakeholders from submitting a request. A low fiction barrier may, however, prevent a stakeholder from submitting a request for a low impact request that is not going to drive business decisions.
- Build tooling that allows your stakeholders to self-serve. If you find yourself getting the same types of requests from your stakeholders week after week, it may be worth building self-serve tooling that enables your stakeholders to answer their own questions. The type of tooling that you build may look different depending on what type of questions you are getting from your stakeholders and who your main stakeholders are. If you work close to data acquisition and get frequent questions about how to name new user events that are being introduced into your website tracking data, you could create a generalized framework that provides guidance on how to name events. If you get frequent requests related to the performance of different sources you use to acquire web traffic, it may make sense to create a self-serve dashboard that your stakeholders can use to answer questions about the performance of different traffic acquisition channels.
- Purchase self-serve tooling from a vendor. If it does not make sense to build self-serve tooling yourself, you can also purchase a third party solution to enable your stakeholders to self-serve. There are many vendors that build tools that aim to enable non-technical stakeholders to access data and answer their own questions about the data.
Build reusable data components
- Create frameworks that standardize the way tasks are performed. It is often useful to create frameworks that standardize the way that a certain task is performed. If there is a standard recommendation for the way a common task should be performed, it will reduce the amount of ambiguity future team members who want to perform the same task have to wade through and allow them to make decisions faster. It will also increase the uniformity of the codebase, which will make it easier for new team members to onboard onto projects they are not familiar with.
- Package up reusable code. It is also useful to package up code into reusable components that can be recycled by team members who want to perform similar tasks in the future. This will reduce the overall amount of code that needs to be written and maintained by the team.
A unified principal for increasing impact
If you have been paying close attention, you will realize that all of the strategies and tactics that we have discussed thus far can be summarized with one underlying principle. That principle is that you should spend more time building reusable solutions that solve multiple problems and less time on one-off solutions that do not offer sustained value. Whether the end user of your product is a data professional or a business stakeholder, you will have a larger impact on the organization if you create a solution that multiple people want to use. That means that you should do your best to reduce the number of ad hoc analyses that you perform so that you can have more time to focus on long term solutions that solve multiple peoples problems.
Check out our article on data science best practices for all of our best recommendations on how to increase the efficacy of data science teams.