Data For Subscriptions

The episode surfaces a few genuinely useful ideas - data mediation as a business-model-driven pull process rather than a data-push exercise, and usage data serving dual/triple purposes beyond pricing - but these are buried in prolonged, meandering exchanges full of hedging, anecdote recycling, and mutual validation. The insight-to-filler ratio is low for a 48-minute runtime.

“it should be the pricing data model determines what data you need to collect and the velocity of the periodicity of the data and the level you need data”

“data mediation can also have dual, if not triple use to understand pattern usage, to be able to forecast, well, what is the likelihood in the next period or the next quarter”

The NoSQL analogy applied to subscription nomenclature is a mildly interesting reframe, and the argument that AI's resource-intensive cost structure is now economically forcing usage pricing into the mainstream is a reasonable fresh angle, but the vast majority of the discussion recycles standard analyst talking points ('start from the business model,' 'don't let technology limit you') that circulate widely in the space.

“NoSQL doesn't mean not SQL. It means not only SQL. That's what the no means. And I think really what we're talking about is not so much subscription is not only one time”

“could I use Genai to basically determine what data I require to populate to make this business model work”

Stephen Harrell has relevant practitioner depth - Oracle, Silicon Valley startups, and product leadership at a subscription billing company including introducing AI - but he has spent four years as an analyst removed from hands-on execution, and the episode leans more on general research observations than hard operational experience at scale.

“I've been an analyst for just over four years now. I spent majority of my career um, on your side of the fence”

“the company previous to joining Ventana which is, was a, ah, company involved in subscription management um, and billing software. So I led product there and I led the introduction of AI to that product”

The episode name-drops BMW heated seats, Philips medical imaging subscriptions, Oracle perpetual licensing, Snowflake, and a South Africa insurance anecdote (with a concrete '6 months' data point), but the buyer's guide research the whole episode is nominally anchored to is never actually cited with findings, numbers, or vendor comparisons - leaving the evidentiary base thin.

“I worked on a project for a property and casualty insurance company in South Africa... they said to it, we need to operationalize this and it said six months”

“I think Philips is offering um, some of its medical imaging on a, on a subscription basis so a smaller practice can avail themselves of the latest technology”

The host adds genuine context from personal experience (Swedish energy companies, AI credits model) and structures the conversation reasonably, but questions are consistently broad and prefaced with lengthy self-answers; there is no pushback, no challenged claim, and the host defaults repeatedly to validation rather than probing follow-ups.

“I second your thought there”

“if you would take um, you know, two minutes and give us, and everybody was listening a bit of advice of what should one think about when one wants to set up data mediation in a good way?”