Storytelling and other essentials

Years of supporting Digital marketers have demonstrated the Importance of story telling. More so today when we can use the innovative tools of causal inference, which are rife with paradoxes and contradictory results. There are stories that need to be told, occasionally there are stories that you should keep to yourself but most of all when dealing with data you need to be honest. Few moments were less trying when you have have to tell the client that not only is the trend they presented to their boss without checking has reversed but that it had never happened. Data analysis can be very complicated and if you learn people cannot resist the temptations to inflate figures then you need

The big reversal

Some time back an analyst of an enterprise client asked me to look at their account and figure out if a recently launched service was a success or a failure. I had expected this call a while back. I had written the measurement plan after consulting with the system analyst then since the dev team had allocated three hours for implementing the document I was asked to help and implement it using a tag Managment solution and used the allocated three hours to verify all the edge cases. At this point the marketing department got an invoice and decided their budget was running low, as they had ended up paying for the dev teams oversight, and told me to postpone the reporting indefinitely. The ops did a blue green launch and then a full launch and then lots of data started coming in.

After a few months of data collection the analyst looked at the data and saw lots of abandoned sessions. The measurement plan was of little help, instead of a 50 page presentation I had provided a long page of strangely named virtual pages referencing a two UML charts they had sent me. The plan had been one of the simplest I had turned out and the trick had been choosing just the right names. The fact that the service had many exit points and only three entry points was not helpful and that success could reset or terminate the sessions early. The terms marketing sieve or anti funnel come to mind. It was real difficult to see if end users were reaching a goal or giving up in transit. Reporting would have managed the aggregation of the data using a number of clever regular expressions and some more to allow one to drill down into the different modules. The analyst could not see almost any successful sessions. The problem was a second more complex service was now being finalized and Marketing needed to present the previous project to the CEO as part of a quarterly meeting.

So I used GA To find the results manually, I saved massive url and their screenshots to permit reproducing the analysis at a later day as any analysis we did not automate would have to be repeated and the clients would then compare to old data and it would invariably lead to contradictions. Alse the CRO team liked to provide highly optimistic reports of their success. This would lead to moments of great consternation when the trend suddenly reversed. The queries were run any differently the outcome would And convince managment when the analyst and later their department heads calls to finalize the presentation. The data ended up showing that both overall and for the three main use cases the service was a win for end users and had reduced a significant load at the call center at peak hours. This was one of the best results in the quarterly presentation and though we told the story we never did a reporting solution. The takeaway is that few analytics products reach production.

Data collected and cleaned was analyzed and reported and then comes the time to take action. But when we considered segments and sub segments the data recommended a opposing action. For almost every segmentation the recommended action reverses.The situation was that it was impossible and to understand if a recently launched product was a success or a failure and furthermore if this was due to the promotion or the product design, and what should the client do next, drop it, keep it and if so which channels to use and which market segments to proceed dfocus acquisition efforts. The elephant cannot come from the data alone.

Like in many situations the client

Another situation is the A/B testing. Where clients interested in becoming more data driven are often tempted to cut a test short of a conclusive outcome and risk making a decision that will negatively impact their business instead of driving it forward.

Storytelling also operates on a cognitive level helping transform the abstractions of data analysis into more engaging material.