A police chief says, “Gather evidence so we can find out who’s guilty,” not “Gather evidence to show this person is guilty.” HR analysts should be so lucky. Too often we’re asked to provide evidence that someone’s initiative has increased productivity, or that an investment in this department will cut turnover, or the ever popular “I want data showing my team is underpaid.”
Managers have a stake in the outcome of an analysis, and they wouldn’t be human if they didn’t try to nudge the results in their favor. Organizational theorists point out that there are inevitably two concurrent dramas around any issue. In one drama everyone is trying to maximize the well-being of the organization, in the other drama people are looking out for themselves. We talk about the former drama and pretend the latter one isn’t happening, but both are in play.
Analysts get into trouble when they ignore the reality that managers will try to cherry-pick data to support the answer they want to hear. If analysts go along with cherry-picking data, then they will undermine their credibility. If they pay no heed to what the manager wants to hear, then they’ll face managers nitpicking the data and second-guessing their methods. We need to face the fact that analytics involves politics. Managing the politics is part of an analyst’s job.
How to Manage the Politics
Analysts should always start a project by having a conversation about the stakeholders. Discuss what they want and how badly they want it. This alerts you to the political dynamics you will need to manage.
If the manager starts out explicitly asking for data to support a certain outcome, you need to push back. You can reframe the request by saying, “That’s an important idea. It will be interesting to see if the data supports your hypothesis. Let me start work on it and I’ll get back to you.” The key step is framing the manager’s request as a hypothesis to be tested, rather than acquiescing to the original framing of “give me data that supports the outcome I want.”
When you do an analysis, if the data looks like it’s not going to support a stakeholder’s preferred outcome, you should gradually warn them that they may be disappointed with the results. This will allow them to brace themselves for bad news and get them thinking about alternatives. Most people will come around if you take the time to ease them through the process. They do care about what’s best for the organization, although they also want it to be good for them. The key is not to dump unexpected, unwelcome results on their plate. Even if the results are available immediately, you will want to give them a “preliminary” view of what the data is indicating, then hold on to the final results until they’ve had some time to process their disappointment.
How to Pre-empt the Problem
While there will always be politics, it would be nice if managers recognized that cherry-picking data is a bad idea and that fact-based decisionmaking will, in the long run, be in their best interests. Hence, the analytics function, as a whole, needs to push for a fact-based culture. An organization will struggle to get value from analytics if the culture turns a blind eye when managers twist facts so that they get the results they want. The head of analytics shouldn’t just be saying, “I need an investment in analytics tools,” “I need an investment in new skills,” or “I need an investment to improve data quality.” They also need to be saying, “I need an investment in a fact-based culture so that we can benefit from analytics.”
The culture manifests itself in practices such as managers always bringing data to the table and when the data is sparse, coming with suggestions about how they could gather fresh evidence. The culture also shows up as an attitude of being happy if the data shows your initial view was wrong. “Happy” may sound a bit strong but let me explain. If you thought the faster route to a location was Road A and then the data shows Road B is faster, you could be upset because the analyst proved you “wrong” or you could be happy because the analyst found a way to get to your destination faster. You want a culture where people lean towards feeling happy if data shows their intuition was wrong (or at least pretending they are). Finance has been able to implement a fact-based culture, analytics should be able to do so too.
These cultural changes—bringing data, seeking out new data, and accepting the conclusions whatever they may be—set the stage for the deepest change: people willing to accept a result that isn’t good for them personally. If an organization has a fact-based culture, and generally treats people well, then the drama of “doing what’s best for the organization” will prevail over the drama of “doing what’s best for me” more often than not. That’s the culture and analytics team will want to work in—which makes building such a culture worth it.