Mr. Pagani, the whole world is talking about the value of data and companies cannot escape from several providers who want to analyze their data. You think data analysis doesn't reveal the information that's really inside? Right?
I think „richness“ of data comes from analyzing it from multiple possible angles. However, I’d argue that nowadays the biggest problem for companies is not the analysis stage. The problem starts earlier, as most companies suffer from a poor data strategy. Traditionally, people in the research industry have been guardians of our clients data. We did the gathering, the processing and the delivery. And for the most part, we did a reasonably good job. However, currently, there are more and more relevant sources of data. The companies generating the data have no idea about the type of data governance protocols that we are used to in our industry. And that’s why I think many analytics projects go terribly wrong. The behavioural data is analyzed and strategies are designed, but the data many times lacks the explanatory power it needs to answer the most basic business questions.
You have done a study on the data structures of products in companies and have found a certain amount of chaos. Please tell us what the result of your research was.
To get started, allow me to quote a CEO we mention in our Esomar paper, “Data science is the easy part. Getting the right data, and getting the data ready for analysis, is much more difficult. The majority of our time is spent getting the data
”. Most companies are capturing only a fraction of the potential value they could be extracting from data and analytics.
This was one of the key findings of our research and we have called it the ‘conjecture of decreasing incremental returns of unnormalized data assets’. As we’ve explained through our paper, when data scientist teams in our clients want to understand and analyze the complete picture of a consumer, with their motivations and journeys and competition. Like parts of a huge jigsaw every time a data analytics team acquires and tries to integrate a new dataset to analyze, the lack-of-norms ‘issue’ kicks in.
A global head of market research for one of the leading beverage companies in the world has told us for our paper that they’ve been working with their worldwide Retail Audit vendor to standardize the product lists – at the SKU level – across all countries in the world: “the project took three years, and we are still far from having a global taxonomy that we can fully trust. And let’s not even try to align Retail Audit products with those that appear in my brand tracking studies. That is one challenge we have given up on”. We should not give up on this. It’s not an easy problem to solve, but we owe it to our clients to do better. Integrating the most common datasets into a coherent data-structure is a challenge most of our clients that are serious about Data Analytics are struggling with right now. And it’s a historical problem of our own making. But that is the problem of companies. They may have a CTO who should take care of that. Why does market research have to deal with this?
Many companies are struggling with their digital transformation journeys and I believe that has a lot to do with the fact that more often than not, initiatives are led by technology heads, instead of analysts. With this framework, I believe our research industry has an important part to play in helping out with those challenges. It’s relatively recent that brands started using their data assets to move away from mass marketing to mass personalization. However, the process is proving a lot slower than anticipated. Mass-personalization requires companies to bring together and analyze a huge volume of data. Datasets that are brought together to understand the different aspects of the ‘customer journey’ tend have no common structure, and that lack of data standards is undermining the value we could provide to our clients. And what is your proposal for a solution of the problem?
The main issue we identified during our study can be seen as the lack of a common language among vendor companies. Technically speaking, the issue lies with the poor -and many times inexistent- corpora of domain application knowledge. While there could be a basic structure that is common and shared across all companies, and especially within a company, no organization/body has ever taken the time to create it. The problem is endemic and serious, and it is crippling the huge potential that properly harmonized ‘big data’ sets would have for our clients’ business. In our paper, we argued over the advantages of creating an industry-wide Ontology Based Data Access (OBDA) for marketing products. The data engineering effort that it would take to create and align the existing unnormalized datasets would be minimal compared to the incremental value for clients that are individually doing the same thing, one-by-one. We propose that our marketing research industry tries to tackle the lack of normalization issues in a formal and systematic way. We have put a proposal to the Esomar Standards Committee for them to get a commission together to study the creation of this data standards.