A New Data Strategy Can't Make Up for Lousy Delivery

We all have that friend who talks about the new diet or new workout routine they’re doing. Maybe you are that friend. The one that never seems to lose weight, but talks about how this new diet might change things.

It’s not that there was anything wrong with these diets. Some might be based on dubious science or have unrealistic goals, but usually the reason they don’t work is that the people on them don’t actually follow them. They give themselves a cheat day that turns into a cheat month. Eventually, they have just stopped completely. At no point during this did the person actually decide to quit their diet. Their efforts just faded away.

That’s how I feel about new data strategies. When things are going poorly, decision makers naturally try to fix things by developing and implementing a new strategy. Much like a new diet, it really only works if the organization has the ability to follow through with it.

Recently, I had published another article that argued data quality degrades over time because organizations don’t build and enforce the right processes. A new data strategy is often offered as solution to those issues. Sometimes that’s true, especially if your last strategy was unrealistic and poorly thought out.

That’s usually not the case though. I’m skeptical myself that every strategy out there was that poorly thought out. I think it’s just that the people making them usually underestimated the amount of leadership it takes to deliver on them. A so-so strategy that’s well executed will do better than a phenomenal strategy with half-hearted efforts. The world is full of good ideas and idea men. The real talent delivers on it.

It’s understandable why people want to come up with new ideas. That’s where all the glory is and so people are incentivized to come up with new strategies. To many people, improving something someone else built doesn’t make you look like a leader. Building something new does.

For example, which of the following would you rather tell your boss and your team?

• “We are going to build a machine learning program that will identify what channels deliver the most leads”
• “We reduced the defect rate from 30% to 5%”

Many people will go straight to the first one. That’s not a bad goal to have, but it doesn’t do much for you if you can’t fix the defect rate. Stakeholders have a hierarchy of needs and data quality is the first one. It doesn’t matter what fancy things you deliver. Once they find out your data is wrong, they’ll stop trusting you.

There’s also the issue with transaction costs. It takes a lot of change management to implement a new strategy. If you used a campaign naming structure to indicate demographic data of your target campaigns, but now you want to use campaign naming structure to indicate campaign types, you have to re-train your staff and change workplace habits. For that reason, you shouldn’t ever make a strategy change lightly.

One Last Note

You might read this and be tempted to say that I don’t think you should ever change data strategies. That’s not true. The world changes around us and if we don’t change with it, we fall behind. I do think decision makers change strategies more than they need to. Netflix didn’t go from a mail delivery service to a on-demand service in three months. It took years. It just seems like lightning speed in retrospect, but it was a concerted effort and they didn’t ignore underlying issues in their services while doing it.