Before I dive in, we would love to get your response to the DCA Industry Survey – just click here. It’s open to anyone in digital commerce, takes about ten minutes, and you’ll get a copy of the final study in December. Thanks!
In September, the Digital Commerce Alliance gathered 19 digital commerce leaders at TransUnion’s headquarters for a day-long roundtable about the evolution of consumer data and data exchange. The knowledge we brought together in that room was powerful, representing over 250 years of work in technology, analytics, marketing, retail and more. A full account of our five-hour conversation would fill a book, and I don’t quite have time to write it . . . so here are six of the things I learned.
- “Rules of the Road” are about market power
We could have spent the whole day talking about the rules that govern data usage, and in a sense we did – the rules govern every aspect of the conversation. But on the question of what the rules are, there’s a pretty simple answer for many companies: whatever the biggest client in the supply chain says they are. Here’s why. Rules don’t just come from Europe or California or India; they also come from private contracts, regulations applying only to some companies, or the assurances some companies provide their customers. It can make surprisingly little difference when a new rule goes into effect, because the organization may already be committed to operating at or above that standard. Big clients tend to have high standards, and those tend to cascade down.
- Most data is messy
For reasons mentioned at #1, there are serious limits to how data is used and requirements for how it is protected. But that doesn’t mean the data is kept neat and tidy. A data set tends to be kept in a way that makes it easy to do the one thing it was originally meant to do. Since more information is captured than used, the resulting “data exhaust” is (so to speak) sitting in a pile in the corner—and it’s usually a mess. Companies are starting to understand its value, and customers are starting to demand more. As one executive said, “everyone is in the data business now.” But it will take a lot of work to get value out of all the data a typical company is permissioned to use.
- For many companies, cookies don’t matter – but for every company, it matters what happens next
The Data Exchange Roundtable was born out of a conversation among DCA members about the death of third-party cookies. The impending end to third-party cookies has already led to more creative use of alternative data sources and prompted collaboration between companies trying to create the “next generation” of consumer intelligence. Most companies in the Roundtable don’t directly rely on third-party cookies, though in at least some cases they are upstream in the data supply chain. But the innovations that arise to fill the space left by third-party cookies, if and when they disappear, are likely to impact everyone.
- Everyone’s got a handful of puzzle pieces
The overall market picture is that every company is walking around with a handful of pieces to a 1000-piece puzzle. They’re missing the box, too—and the puzzle is the consumer. In the past, this partial view was good enough. But as some competitors gain enough puzzle pieces to start to see the full picture, pressure will build for everyone to have a coherent approach to knowing more about the customer.
- Clean rooms and federated learning are a good tool for getting the other puzzle pieces in place
Our discussion of clean rooms and federated learning pointed to two potential ways for companies with a few puzzle pieces to work together to serve consumers better. I explained clean rooms in a prior post—but in short, they allow data sets from different sources to be compared without violating privacy obligations. This is generally done through tokenization or other anonymity-preserving techniques. To continue the analogy, clean rooms can be used by parties who each have an incomplete set of puzzle pieces to see more of the picture.
Federated learning is different. Where clean rooms pull data into the same place to be analyzed, under federated learning an AI “brain” is trained in separate sessions on different data sets that never come together in one place. The effect is similar: different data sets are used to generate insight without violating the rules of the road. In federated learning, the software doing the analysis makes a separate house call to each data set.
- AI’s power is in tapping unstructured data
This is a bit speculative, but I think a potential consequence of generative AI will be to allow rapid analysis of unstructured data sets and potentially the automated identification of inaccuracies. A factor underlying much of our discussion was the messiness of data, which arises in part from fast-changing business uses and data sources. Conventional software operates on a garbage-in, garbage-out basis, and cleaning all that up is what humans are for. If we’re able to substitute machine intelligence for human intelligence in that role, it could relieve the need for at least some of our biggest IT investments and allow us to unleash AI on even relatively poorly organized data sets with good results.
As mentioned up top, these are just a few thoughts from a very full conversation. We’re convening DCA members again on November 8th for our Fall Summit at Mastercard in New York city. After that, our final live event of the year is another Roundtable at Collinson-Valuedynamx’s London offices on December 13th.
Participate in the DCA Industry Study!
We would love to get your response to the DCA Industry Survey – just click here. It’s open to anyone in digital commerce, takes about ten minutes, and you’ll get a copy of the final study in December. Thanks!