Don't Let Your Data Collect Digital Dust

A very important topic was introduced during a HN Insider conversation between Bryson Koehler, CEO of Revinate, and Jeff Jonas, CEO and Founder of Senzing, Inc. – that being entity duplication and the need to resolve identities in order to have enough confidence for targeted segmentation.

Significantly, the two former IBM alums discuss how thinking you have just 20% more customers than you actually have can skew a data set enough to reduce confidence to the point where it forces decision makers to maintain a generalist approach to marketing. It’s a matter of risk aversion wherein if you have a presumption about someone and you’re wrong, this can cause more headaches than if you were to keep your messaging at a broader level of personalization.

In this sense, data quality is paramount towards market intelligence which is also decision intelligence. For instance, identity resolution can help to build a more accurate knowledge graph about how hotel guests interact with promotions in an omnichannel environment so that you not only know what offers will most likely resonant but also when the ideal time to market to these customers is – for example, three months, three weeks or three days out from arrival.

A useful tool that has emerged to aid in guest profile deduplication is ‘entity-centric learning’ which uses machine learning (ML) to assemble what’s known about a customer through a unison of profiles, which often means taking into account cultural nuances as well as how addresses vary across different territories. This form of ML also has use cases in fraud prevention whereby fraudsters purposefully act to separate information through different channels in order to obscure their identity, with entity-centric learning able to piece together overlapping partial profiles to give a probability of risk.

Ultimately, what Koehler stresses is that data is an organism that requires constant gardening so that it can be assembled correctly to deliver valuable business outcomes. In late 2024, we are still very much within a commercial inflection point for various types of artificial intelligence (AI), with ML providing many different valuable applications if it can make observations off of high quality data, while other more nascent forms like retrieval-augmented generation (RAG) proving to be lucrative in making sense of very large data sets. Next year will likely be the year of data, so be sure to invest in cleaning and centralizing it, lest it start collecting digital dust.