By Ernesto Fernandez, Harvard Business School, and Marianne Pelletier, Staupell Analytics Group
The insights that data analytics and machine learning enable into our work are powerful, and prospect development shops are right to embrace them. As more consulting firms pile into the sector advertising transformative solutions, organizations are taking these analysts up on their offer to turn their data into game-changing insights.
However, as our leaders stand prominently at the bow of the ship pointing our way into these waters for the first time, it is typically those of us in business intelligence (BI) and development operations (DevOps) who are actually steering the ship.
As BI/DevOps professionals, we are our organizations’ natural representatives to these specialists. This is true even though we are unlikely to fully comprehend all of the competing stakeholders and priorities that will determine the project’s (and our own) perceived success. You are the “data person” and this is a “data project,” so this thing is yours, ready or not.
The project manager must ensure that their analyst’s time, their team’s vague questions and the contents of a less-than-perfect CRM converge into an actionable plan. If you find yourself in this position, we offer you this roadmap to getting a handle on your project as the person who is (or at least will be held) responsible for everything from your leaderships’ ambitions to your analyst’s deliverables.
At its core, a data analytics project is a research study, and therefore needs some application of the scientific method to be effective. Accordingly, we recommend this set of steps:
1. State the Problem(s)
Resist your organization’s urge to “send them the data and see what they say” before deciding exactly what you want from your analyst. Ask, “What has brought us to the point that we are now ready to pay for this research?” Articulate organizational needs in concrete terms and express your management, operational or strategic issues and failures honestly, in writing.
2. Define Goals
Push beyond the sullen confession of grievances and frustrations to a more hopeful vision of a better future. Ask, “What would ‘better’ look like?” Establish specific objectives for overcoming the needs, issues and/or failures expressed in your problem statement(s) in writing.
3. Formulate Research Questions
Ask, “What answers do we need in order to get to ‘better’?” Determine what insight and understanding you need in order to realize your goals. Identify what former Secretary of Defense Donald Rumsfeld called “the known unknowns.” Take the most time with this part and involve all stakeholders, remembering that the entire project must and will be directed by the exact questions you pose to your analyst, in writing.
4. Assess Capacity
Ask, “What people, money, tools and data do we actually have to work with?”
People. Take inventory of your resources of time and talent. Who can work on this and how long can you afford to have them doing it?
Money. Set your analyst’s budget, including any overages and surcharges you may want to be able to afford if you choose to “upgrade” the project later on.
Tools. Determine if your existing information infrastructure will support the desired outcomes.
Data. This one is a bear, and actually deserves its own article, but suffice it for now to say that you must catalog all available data, value by value, and prepare a comprehensive codebook explaining everything you know about every field you have available. In writing.
5. Revise Your Research Questions
Consider your original research questions in light of your newly-assessed capacity and ask, “What questions can we actually answer?” Together with your analyst, revise your research questions and come to an agreement about what is possible based on the available resources of people, money, tools and data. In writing.
6. Set Expectations in a Scope of Work
Once you come to an agreement with your analyst, ask, “How long will this take and what will it actually look like?”
Enumerate your deliverables; set deadlines for those deliverables; schedule presentations for those deadlines; and come to an agreement on your analyst’s long-term responsibilities to support your organization’s interpretation and implementation of their results, as well as contingencies in case things do not go as planned. Be painfully specific and get granular as you hash out your scope of work. In writing.
7. Answer Research Questions
Once you have laid all this groundwork, step back and let your analyst do their job. Take a sincere interest in the progress without rushing or smothering. Remember, you are paying your analyst for success. So ask, “How can we help our analyst be successful?” Touch base often enough to answer questions, weigh in on interpretations, correct assumptions and spot-check results as they arise. In writing, when possible.
8. Test Results
In addition to spot-checking results for problems as your analyst goes along, build in your own time to exhaustively review and test everything that will be shared with your team. Ask yourself, “Are we really looking at and talking about the same data?” Do not give minor errors, regardless of whose errors, a chance to snowball.
Without exception, the analyst must be expected to answer all of your questions to your satisfaction (but not necessarily to your comfort) before it reaches anyone else’s desk. Try to reproduce the same numbers and results. Challenge yourself to find endogenous and confounding variables. Anticipate special or fringe cases and see where they show up in the results.
Most of all, remember that you are the gatekeeper responsible for ensuring that your leadership only sees results that you have vetted and are worth their time. Always strive for collaborativeness and collegiality with your analyst, but if they are in any way resistant to you fulfilling this important responsibility, call them out for it in plain language. In writing.
Once you have vetted your analyst’s answers to your research questions, ask yourself, “Did we find a better question?”
Be mindful that iteration takes much more time and effort as interested participants join in, so make sure you have first exhausted your own ability to help refine the question before bringing others into the process. Save everyone time by predicting the predictable reactions and asking the predictable questions.
Your goal is to get to a set of results and recommendations that your organization will know what to do with. In writing.
10. Convert Findings into Strategy
Once your analyst has answered your and your organizations’ many questions to your team’s satisfaction (remember that having results proven is different from having results changed because your stakeholders do not like them), it is time to convert these analyses into strategic recommendations. Rather than letting yourself, your analyst and your colleagues become mired in abstracted concepts and theories, start with this question: “What are we going to do with these results in our day-to-day?”
Present these recommendations in simple, compelling visualizations and clear language that is native to your organization, especially to the decision makers.
To assure the best use of the analysis, cater to your current culture. If your team or management like a certain reporting tool, use that. Make the results familiar by repetition and by inserting them everywhere and anywhere that they make sense. In addition, be available to answer questions or even retrain — one-on-one and away from the judgmental gazes of others — to encourage adoption.
In the end, your upper management must use the new information from the analysis themselves if anyone else is to be expected to learn and use it. Make sure that management has all the training and tools needed to get comfortable with the results and commit to an action plan.
The potential of a project manager to maximize outcomes for their organization are difficult to overstate. You will be called on by your organization to serve variably as a subject matter expert, an interpreter, an investigator, an informant, an air traffic controller and at times a therapist. You will also be an invaluable asset to your analyst as their own personal Virgil, lighting their descent through the layers of your organization to help them uncover the quality results for which your organization is paying good money for. Adhering to this methodology will help you structure your work and set a manageable but efficient pace for everyone involved. Most of all, it will help you ensure you are doing your own best work while always serving the needs of your only real client: the organization itself.