Anyone who has been a data analyst or data scientist for even a short while has been there before. You get a request from a stakeholder to pull some data, you run some queries to generate the data you believe is being requested, you send it to the stakeholder, you go back to working on your ever growing backlog of tasks, and you never hear anything about it again.
While it feels good to quickly fulfill a request and get it off your plate, as a data leader, I’ve found that it can lead cross-functional partners to develop the wrong impressions about how data teams deliver value and impact, and even worse, often leads to suboptimal business decisions.
Use this project framework to understand the major phases of an analytics project, common pitfalls and tips to handle them, and turn inbound requests from stakeholders into memorable and leveraged business impact.
Project Phases
At the highest level, we can break down any typical analytics project into five phases:
The Business Question
The Analytical Approach
Data Processing and Cleaning
Analysis
Delivery and Communication of Results
Of course every project is different and the complexity and time of each phase will vary greatly. The main goal is to have clarity on the why of what you’ve been asked to do and develop your project plan with that why in mind.
The Business Question
The underlying business question and context for the request is the cornerstone of any analytics project. As a data practitioner, it is your job to help your stakeholders reveal the business question to you. Establish a shared understanding of the business question before proceeding any further to maximize the likelihood that your work will be used appropriately.
Here are a few common things to navigate:
The request is misaligned with the question. Since a stakeholder is more likely to request specific data versus posing a question upfront, once the business question is known it may reveal a mismatch between the requested data and the underlying question. This is normal and exactly the reason this phase is always the first step. You haven’t wasted your time pulling the wrong data and putting it into the wrong hands.
There is a better business question. As the data practitioner, you have significantly more knowledge about the strengths, limitations, and caveats of your company’s data than your stakeholders do. Don’t take your stakeholder’s business question at face value. If it seems appropriate, reframe the business question or redirect your stakeholder to a better question. Data practitioners sometimes hesitate to do this out of fear or an assumption that their stakeholder is always right. Try this out and you’ll probably be pleasantly surprised at how receptive (and grateful) your stakeholders will be.
The question isn’t appropriate to answer now. This can lead to difficult conversations but the reality is that not every business question a stakeholder a brings to you should always be prioritized. Maybe the question is better answered with user interviews. Maybe only a handful of customers are affected. The key is to make sure your stakeholder understands why.
It’s worth noting that when your stakeholder doesn’t immediately volunteer the context behind their request, it’s often because they don’t know this is information you need. Use this as an opportunity to educate and explain why it’s so important.
The Analytical Approach
Armed with a shared understanding of an explicitly stated business question, you can begin to sketch out your analytics work.
While many of the details here will be dependent on how your company and team operates, the goal is to define an approach that allows you deliver with these considerations in mind:
The urgency of the request. The level of urgency is usually the overarching factor that will influence what approach you should choose.
The accuracy or detail that’s needed. There is usually a tradeoff between urgency and how thorough you can be.
Understand the difference between a real deadline and a fake deadline. Real deadlines, like an upcoming board meeting, are hard constraints. On the other hand, fake deadlines can be a good opportunity to strategically increase scope to demonstrate more value. “If I had an extra day, I would also be able to do…”
If you’re not able to find a viable analytical approach that meets your stakeholder’s urgency and accuracy parameters, speak up. It’s better to have a discussion about cutting scope now rather than a discussion about missed deadlines later.
There are many ways to skin a cat. In a resource-constrained business setting, the best way to skin a cat is by using the simplest option that answers the underlying business question on time.
Data Processing and Cleaning
If you’re a new data practitioner, you’ll soon find out that 70 percent of your time on projects will be spent on data cleaning. If you’re a seasoned data practitioner, you probably think 70 percent is an underestimate.
The importance of data cleaning can’t be overstated. Anything you do after this point is highly dependent on the quality of your data and any decisions you make about how to clean it. For example, if you want to compute average revenue, your result may change if you decide to coalesce null values to $0.
Consider these tips:
Allocate more time. For tables that are new to you or that you don’t use regularly, anticipate that something unexpected will come up and celebrate if nothing does. Even for tables you’re intimately familiar with, applying them to new business problems may reveal new issues.
Know the impact of your decisions. Often there are not obvious “right” data cleaning decisions but you should make those choices with intention. More importantly, consider how your results may change when your decision changes. Use this as a way to pressure test your decisions.
Spend appropriate time making sure you’re building your analysis on a solid foundation to avoid having to awkwardly backtrack on your findings later.
Analysis
You’re finally at the phase where you expected to spend all of your professional time. Congrats! Know that all the work you did before this point will make the job of analysis that much easier.
Remember to:
Know the business question. This is important to keep in mind at every phase, but is particularly important here. Make sure that as you execute on your analysis, whether it’s one simple query or a complex analysis, you remember the reason why you’re doing it in the first place.
Make deliberate choices. Similar to how there’s not one way to clean data, there’s also not one way to properly do an analysis. Make sure you can speak to the choices you made.
(When relevant) Try out multiple versions. If you will be doing a presentation of your analysis in the next phase, this is phase where you should be trying out different ways of presenting your findings with the goal of maximizing the likelihood that your audience will understand, believe, and act on your findings. Don’t be afraid to make two headlines or a bar chart and a line chart, and think about which would make more sense if you were in the audience.
Delivery and Communication of Results
Whether it’s sharing a query result in Slack or a fancy presentation in the boardroom, the final delivery of your work is one of the most important responsibilities as a data practitioner.
Delivery and communication seem like a simple task but it can be surprisingly difficult. Make sure your stakeholder:
Understands your data or findings
Believes that the established business question has been answered
Shares their follow up questions
The goal is to make sure your stakeholder understands the data you’ve equipped them with and will use it in your intended way to make the right decisions.
Consider these pitfalls:
Including too much information. Recognize that your stakeholder or audience may not understand or even really care about all the details of your work. At the end of the day, they (and really you too) only care what they should do or what you recommend. Leave all the nitty-gritty details for a team meeting and focus on the key findings.
Not following up. It’s important to confirm with your stakeholder that your work has addressed their needs. If not, this is the time to work with your stakeholder to understand the root cause of any misalignment.
At the end of the day, they (and really you too) only care what they should do or what you recommend. Your credibility as a data practitioner is the most valuable currency you ensure that stakeholders act on your work.