Sales prediction at Dell

In plain terms

A big company already sells to a lot of customers. Some of them would buy a second product as well, and nobody knows which — so the sales team calls everyone, or it guesses. A propensity model is a way of ranking that list: you show it who bought what in the past, and it scores the customers who most resemble the ones who already said yes.

That's what this was, and it's worth reading with the caveats attached. It was an internship, the model was one piece of a much larger sales machine, and the pipeline figure below is an estimate built from average deal sizes rather than money anybody collected.

The project

During a summer internship in 2018, I built a propensity model (Azure ML) on Dell's historical customer database to identify cross-sell targets for a high-margin software product. It surfaced about 20,000 qualified accounts representing an estimated $129M in pipeline, based on segment-level average deal sizes. The model was hosted on Azure so sales and marketing teams could get predictions and feed back new training data.

It was an internship project, so I'll keep the claims modest — but it's where I first saw that the interesting work is usually where data, revenue, and tooling meet, and I've been working in that intersection since.