Mathematical modeling is a powerful tool, especially when the model is backed up by Big Data. Which is why data science has become all the rage in the business world. Data scientists are the toast of the town.
We had a high-profile example of the power of mathematical models during the last election: Pundits' predictions were all over the electoral map. But the models developed by the "quants," based on dozens of state and national polls, came out on the money.
Yes, models are powerful. But a note of caution is in order. A model is only as good as the data that goes into it. And it isn't enough just to have a lot of data, even Big Data. The data has to be correct, and it has to be the right data – data that offers relevant insights.
Even the best models, let us remember, are simplified representations of the complex real world. As analytics expert Andrew Anderson notes at the Adobe Digital Marketing Blog: "All models are wrong. Some are useful."
Models are often so impressive-looking however, that they can be taken for received truth. The temptation to regard them as authoritative may be especially strong for those, such as marketers, whose own expertise is not mathematical. If you are accustomed to living in a world of uncertain and subjective responses – relying on what consumers say – curves and statistics seem all the more solid.
But the models are only as good as the data that does into them. And when the data involves humans it is probably still subjective.
The election models got tested, finally and decisively, on Election Day. In the business world there is rarely such a clean, once-for-all opportunity to observe results.
Testing must be continuous, and it must be testing against the real world. Testing a model against itself can find implementation errors, but it won't tell you if the data you have is relevant data. Only real-world tests will tell you that.
