Big Data meet Smart Thinking

It must have been funny in the early days of TV: a bunch of producers would broadcast programmes and they could all claim they had the finest programme and the highest audience. The science of measurement took a long time to catch up with the truth and put numbers on the viewers. Even now, traditional media are tough to measure, and to understand effectiveness is an even bigger stretch.

By contrast, the advent of online media must have seemed like a breath of fresh air to numbers-oriented chief marketing officers, and the non-marketing technology people who often managed online budgets would have been thrilled by the sheer accountability of it all. A good banner gets clicked on more than a bad one, so optimisation was empowered in a way that was undreamt of in the traditional media world.

But, in the headlong rush to embrace measurement and accountability, Big Data can sometimes lose its way. The dark world of Conversion Attribution is a case in point. It works like this: online shoppers generally travel through a number of web locations and eventually complete an online purchase. Companies have grown familiar with the idea of paying the people in the chain that made the sale happen. Simplistically this makes sense. I come into a subject completely cold, I do a search and up pops the Ford Focus. I click on the link and order a car online. Clearly, the search provider wants and deserves a piece of the sale.

At this point two very bad things have tended to happen. The first is that proximity to the sale was credited with evidence of causality. In other words, the last thing that I did was, by definition, the thing that caused the sale. This is known as the ‘last click’ phenomenon. The site responsible for the last click before sending the consumer off to the sale is the one that gets the credit.

The second bad thing is to ignore the concept of causality or, more simply, ignore whether the specific online event has any effect on the probability of the sale being made. If I stood outside Tesco and shouted to everyone going in that they were about to go into Tesco, there is no doubt that I would be on the ‘path to purchase’ and consumers would be aware of my instructions on the location of Tesco; but clearly I made no real difference to the chances of a sale being made. Many of the approaches to Conversion Attribution have taken a frighteningly simplistic view of causality and, if an event directly preceded a sale, companies were happy to pay for that event.

Pointlogic are starting to work in this space now and we are finding there is a lot of traction in bringing Smart Thinking to sit and work alongside Big Data. The nature of the thinking is really to do nothing more than apply the sort of tests of cause and effect that marketers have grown to accept in the ironically more data-poor traditional media world.

When we try to understand effectiveness within online media, it is not enough to highlight that an online event was the last thing that happened. We want to understand the actual contribution of different types of activity and of external events such as TV ads, the weather and all the usual causality suspects.

Our current focus is on effectiveness measurement, which tends to be more retrospective, looking back at what caused what. This is valuable for bringing a sense of the true value of marketing in an online and offline sense. We are also working with digital Big Data specialists to develop ways to link this type of mature analytics to real-time transactional events within online. This, in effect, would mean modifying what you pay for things such as search terms by better understanding whether the person is destined to buy anyway, and maybe even knowing that he is destined to buy and the credit should go to that radio ad you ran yesterday.

We are therefore applying our mathematical skills to make Big Data a little smarter and we are looking forward to doing this with more partners and clients.

To find out more contact Edgar de Gelder ( This e-mail address is being protected from spambots. You need JavaScript enabled to view it ).