Is Your Digital Advertising Data Descriptive or Predictive?

by George Konetes | Aug 7, 2020 | Digital

In this age of digital advertising, data is readily available like never before. But it is easy to be lulled into a false sense of confidence in our data just because we have it. Not all data has equal value, nor should it be used to make business decisions. Sometimes our data is descriptive of what happened, and sometimes it is inferential of what may happen. Knowing which is which will enable you to make better, more accurate, and more profitable decisions.

What Happened

Descriptive data tells us exactly what happened. Your fundraising ad was shown to 1000 people, 15 of them clicked on the ad and of them, 5 people donated. The information is accurate, it is helpful, but it is not predictive. In other words, if we show that same ad to 1000 more people, we have no statistical confidence that we will get the same or similar results. The numbers are just too small.

What Will Happen

Inferential statistics, on the other hand, draw from larger data sets and bigger sample sizes. This data has the power to infer that similar results will happen next time with a certain confidence level.

An Example

Facebook provides an A/B test tool that enables you to create an advertising campaign and duplicate that campaign, changing one variable, to test which produces a better result. Say you want to test what impact two photos have on cost per new donor acquired. You can set up an experiment and test that photo variable. The media spend needed to obtain a certain level of statistical significance is then calculated.

This sounds like a great tool, and it can be, but it reveals a problem that many people do not realize. In many cases it takes around $3,000 worth of media spend to obtain enough data to get a result with an 85% confidence level. Which means the data is robust enough to infer there is an 85% chance that if the test was repeated, you would get the same outcome. If you want a higher confidence level, you usually need to spend more. Which means any test that is much smaller is likely to not be statistically significant.

The Takeaway

Many “tests” that are conducted are too small and limited to provide inferential learnings. And sometimes decisions are being made with more certainty than the data gives cause for. The bottom line is to know what your data can really tell you and make the best decisions possible with the information you have.

Need help collecting or interpreting your marketing data? We would be happy to talk.

If you are interested in Public Relations strategy or consultation to help protect your brand, Infinity Concepts is here to help. CLICK HERE or call us today at 724-733-1200.