OMD @ The Consumer Engagement Technology Conference 2015

Last month, I was invited to speak at a masterclass put on by the Consumer Engagement Technology Forum. I presented the topic, “The Role of Data and Analysis in Driving Consumer Engagement”, to an audience of 60 representatives from both local and international companies here in the region. Here, I summarize the five key points from my presentation.

  1. Having a clear measurement framework of consumer engagement is critical. At any typical organization, many data sources are available, including data on: sales, transactions, leads, footfall, website analytics data, apps, email/CRM databases, social data, media and marketing, PR, brand health, as well as data from other research such as surveys.
  2. The organization, de-silo and reintegration of data can be achieved by linking data sets using User-IDs and time-stamps, or by simple data fusions on overlapping variables in different data sets. The objective is to create a single, unified view of customers over time in a manner that is consistent.
  3. The methods of analysis do not have to be complex, but have to be objective driven. Analysts sometimes forget that it’s “people who are producing” the data. The best practice is to constantly ask, “Why? Why did the consumers do this?”, and dig deeper into the psychology and motivations of consumers. Put the analysis into context of consumer behavior and don’t only deploy mathematical and statistical techniques blindly.
  4. Thoroughly follow the “Occam’s razor” principle for every analysis – don’t try to boil the entire ocean and make calculations more complex than they need to be. Entities should not be multiplied beyond necessity and the simplest explanation (all other factors being equal) is usually the right one.
  5. Validating analysis results is of utmost importance. Out-of-sample forecasting – where models are used to calculate the expected levels of sales or engagement – is the most powerful way to ensure that the analysis is useful and produces correct predictions. The average prediction error should be below 10%, ideally in the 5-7% range. Once this test is passed, the model can be used for business decisions. Other ways of validation include surveys or experimentation.

In summary, the main point I want to get through is that data analytics should be used to reduce uncertainty and minimize the risk of strategic and tactical business decisions. If the analysis does not produce results that deliver on this objective, or fail to pass the validation phase, further investigation needs to be conducted on the factors underlying the consumer engagement process being analyzed.

Share.

About Author

Igor Skokan

Igor is the former head of Annalect, our insights and data analytics arm. He ensured that all data was carefully analyzed and examined in order to unearth smart, sharp and timely insights that could be applied across the board. In a previous lifetime, Igor was a Slovak nuclear scientist.

Leave A Reply