Measuring Conversion Before It Happens

Measuring Conversion Before It Happens

The age-old adage “Don’t count your chickens before they are hatched” certainly had not taken into account the advances in mathematics and predictive analytics. Savvy marketers today not only need to be able to count their chickens before hatching (forecasting), but even more savvy marketers can influence the number of chickens that are hatched by clever intervention. Welcome to the world of predictive analytics.

Understanding leading and lagging indicators

The concept of leading and lagging indicators can be simplified. Lagging indicators are like your traditional Business Intelligence reports. It is a scorecard. The sales are made and conversion recorded. There is nothing more you can do about it except to see what is wrong and perhaps change your strategy or process in the future. In medical term, it is like the equivalent of an autopsy. You can discover what cause the death but there is nothing more you can do to the dead subject.

Leading indicators can be data that can help you predict the outcome. It is the equivalent as a doctor’s diagnosis. With a good diagnosis, you can prescribe medication or even do preventive measure to avoid certain illnesses.

Since leading indicators can be so powerful, where do we discover the leading data? By definition, leading indicators are data that correlates to the lagging data (sales and conversions are what that are of interest to us marketers). We can measure anything that we think that can ultimately impact sales and we can find out the correlation between what we measure (eg. Number of Facebook likes, number of recommends, positive user generated contents, positive rating etc.) and correlate them with the sales. Correlation can be easily calculated with any statistical tools such as R or even with an excel. The higher the correlation coefficient, the greater the predictive ability of the value (the value falls between -1 to +1). If you discover a negative correlation, do not throw out the variable. In fact, it is telling you NOT to ignore that variable as it means that it can be used to predict a drop in sales! If something has little value in prediction, the correlation coefficient is near to 0 (eg. 0.1 or -0.1).

Why lagging indicators are not sufficient in optimizing

Looking out for good predictors of sales can be difficult and perfecting a model that can predict sales accurately is even harder. Many organizations give up on the hard work and decide to optimize purely base on previous results. On the web, widespread adoption of analytics (partly fueled by free analytics software such as Google Analytics) has caused a surge in demand for analysts that many people who are not well trained in analytics and statistics decided to ride on the surging wave. However, as analysis is difficult, most analysts choose the easy way out and instead of analyzing, they do reporting. Some analysts do not even go beyond the metrics provided by standard dashboards (bounce rate, visits, page views etc.). However, optimizing only on these standard metrics that are lagging data is difficult and often a trial and error. Moreover, due to the lag the data exhibit, real time effort such as marketing activities or other optimization work may make the data confusing. Lastly, optimizing base on these reports only cause small incremental uplift (however, it is always better than no optimization) and organizations seldom see the paradigm shift they hope for.

Using leading indicators to predict and optimize sales

Therefore, we recommend that a proper measurement framework be built before any serious analytical work begins. Sentiments and other leading indicators of sales and conversions may be not easy to measure directly and we need to think about proxies and their predicting powers by calculating correlations. After working them out, we also need to do a predictive model to accommodate the various variables to have a viable model.

Once we figure out all the leading indicators that can predict the rise or drop in conversion, we can then make conscious business decision to optimize the values of those indicators. Not only we would be able to count the chickens before they hatch, we are able to get more chickens than ever.