The most important factor in determining if a given data science project will succeed or fail in a business environment is not the quality of the results. In an ideal world, that would be the case, but unfortunately it isn’t true in the real world that we live in. I know I have some explaining to do with that comment, so read on!
Solid Results Don’t Even Get You Halfway Home
First of all, I’ll make clear that absolutely, positively producing accurate results is crucially important. Every professional creating data science processes must ensure that results are valid and accurate every time. At the end of the day, however, from the viewpoint of the people who sponsor data science and analytics projects, the results themselves are at most 50 percent of the criteria that will determine if they view the project a success.
At least 50 percent of the success of a project will be based on how well the results are put together in a presentation and how that presentation is delivered. Is the presenter able to position the results effectively? Can the presenter interpret things in a way that makes sense to the audience so they will be comfortable taking action? When building a data science process, you can’t just focus on generating compelling results, however tempting that may be. You must also leave time to focus on the right interpretation, positioning, and selling of the results to the businesspeople who asked for the analysis.
A business team won’t care about your weeks of effort and all the gory details you waded through to get to the results. They care about the results. You must get the results across to the business sponsors effectively, or the results may as well not exist. Producing great results is necessary, but not sufficient to having the project viewed as a success.
Raw Results Can Seem Dull
In the marketing industry, models can be used to identify which customers should be contacted with an offer. After the offers are made, the lift from those efforts can be measured exactly. If it worked, more of the same can be done. If it didn’t work, the team can stop and try something else. Great mathematically and financially, but not very exciting in and of itself.
One of the biggest budgets on a lot of companies’ books is mass advertising in the form of TV, radio, newspapers, and so on. Such media do have an impact. However, it is nearly impossible to get a highly accurate measure of that impact, the methods to do so are notoriously tricky, and, in my opinion based on my experience using them, dicey at best. Yet advertising is still pervasive, even when there are other, more measurable options available that budgets could shift to. Why is that?
A Lesson From The Advertising Industry
What do ad agencies do when they pitch their plans? They show up with a multimedia presentation. They’ll have music. They’ll have video. They’ll have new catch phrases. They’ll talk about how much customers will love the brand. They’ll get everyone tearing up with a touching scene of a family interacting with a product in a way that makes their life better. The agency will get the audience so pumped up about the story they’ve heard that they’re ready to sign up. Even the fact that the advertising can’t be measured cleanly on the back end that doesn’t matter much, because the audience has bought into the vision and the excitement of what that ad agency is suggesting they do. In other words, agencies do a great job telling their story.
The intention is not to pick on ad agencies (please don’t send the nasty letters). Rather, the intent is to compliment them! While advertising is not nearly as measurable as some other activities, it has maintained a huge share of expenditures. That is in part because of the advertising industry’s ability to make what they do compelling to their sponsors through a great story. Advertising agencies fully understand and leverage the power of presentation and communication, and data science and analytics teams can learn a lot from those ad agencies. Imagine how successful a data science project can be if highly measurable actions based on solid analytics are paired with the excitement level that advertising activities instill in the business community.
Results Are Necessary, But Not Sufficient
It takes hard work and practice to develop the ability to distill a lengthy and complex set of results down to digestible sound bites. At times, you will feel you are watering things down too much. You will feel like you are focusing too much on fluffy slides instead of meaty algorithms. While it is necessary to have the details to defend the findings available, the details shouldn’t be brought out until necessary. The business team’s eyes will glaze over, they’ll tune out, and they won’t act on the results if discussion gets too technical. You must deliver results in a way that keeps the sponsors engaged and interested. To do that, you need to put in the effort to create a compelling presentation and you must accept that to do that you will have to stop chasing a little extra lift in favor of making sure you present successfully. Producing great results is necessary, but not sufficient to having your data science project viewed as a success.
Note: This blog is based on content from chapter 8 of my book Taming The Big Data Tidal Wave, Wiley 2012.
Originally published by the International Institute for Analytics
