“We consider our company to be data driven.”
Over the course of a long career, I’ve seen certain beliefs take hold in a way that’s completely untethered to reality. Today, the term “data-driven” is universally seen as a desirable state. You will not find a CEO of a publicly traded company who boasts of making decisions by instinct, or who claims to “shoot from the hip”. But how much of this focus on data aspirational and how much is real?
The early days of working with clients are sometimes very valuable to a consultancy. We’re human, and we’re team players, and we can’t help at least sipping whatever Kool-Aid® is on tap among people we identify with. But the view from outside can be telling. When we ask people to agree or disagree with the statement above, we sometimes perceive a sort of syllogism:
All smart companies are data-driven.
Our company is smart.
Therefore, our company is data-driven.
Besides…it’s in our annual report, so it must be true.
As our engagement matures, we hope clients are looking at decision-making at all levels and asking themselves whether decisions are really flowing from data. Sometimes we find that key decisions (like retail site selection as an example) are made with unreasonable constraints. Perhaps the commercial real-estate broker only shows properties in his own inventory, highlighting those that benefit him the most – perhaps over a well-lubricated steak dinner. Data, if it comes in at all, is used for justification of a decision that’s already been made.
It’s true that not all decisions can be handed over to the “Quants”. Sometimes, the available data will not point the way with any certainty. Sometimes there are unquantifiable considerations that must be weighed in the balance. Sometimes, the right answer for the current quarter may be the wrong answer for 3 years hence.
Our advice: Get into the habit of challenging your company’s belief in its own “data-driven-ness”. Encourage the habit of questioning standard operating procedure (hint: if SOP is 5 years old, it’s probably not data-driven). We also advocate that business leaders subject the data-driven answers on biggish things to human review. Data can provide answers, but not prudence.
Not surprisingly, we encourage a third way – a methodical approach that we think will unlock more potential without incurring additional risk. I’ll try to lay out the basics of how we encourage clients to approach matters of access and transparency here:
Shine a light on any over-arching policy that enables managers to avoid thinking about the problem. The truth is that different data involves different levels of risk. You’re paying these people in part for their judgment, so make them use it.
Each data object (schema, database, table, field) needs to be assessed on 3 questions.
Is the data object in its entirety subject to one or more regulatory codes (HIPAA, PCI DSS, etc.)?
Is the handling of the data (either at-rest or in motion) subject to contractual constraints?
Is there an obvious risk or benefit to changing the level of accessibility within the company?
Make sure that if the company does not own a given dataset outright, that any applicable leasing or rental agreement allows the change you are contemplating without additional cost. If additional cost is involved, you should be able to negotiate it to the point of minor impact.
The tools that non-experts might use to explore data may be cost-prohibitive when a group of employees dabble infrequently and seldom produce quantifiable results. Work with software vendors on access plans that make economic sense, and don’t shy away from expecting them to be flexible enough to support your strategic decisions. But common sense should prevail. Do manufacturing employees really need marketing data? On the other hand, might marketing types benefit from manufacturing data?
The third way adds common sense to policy. It replaces the concept of data gatekeeping with the concept of data stewardship. Most of all, it seeks to make available all the data that MAY help a given employee to bring extra value to the company.
Regarding transparency, we strongly suggest that companies avail themselves of modern data catalog software. Providing access to data when the meaning of those data objects is opaque tends to do more harm than good. Valuable time is wasted because users are using wrong, outdated or otherwise sub-optimal data. We’ll be covering this topic at greater length soon.