Question
A new product, a new ICP or a young company has no closed deals for a model to train on. What works instead, and why learned models are weakest exactly here.
The short answer
With no closed deals to learn from, score deals on weights you set yourself or an industry benchmark, then let your own history take over as it builds. A new product or a young company has nothing to train on, which is when a score you can read and change beats one that has to be learned.
How Kaypo answers this on your pipeline
30 minutes with the founder
Thirty minutes on your pipeline or the example one. Pick a time and it is booked.
Every model that learns from your history needs enough of it. Launch a new product into a new ICP and the history that exists describes a different business selling a different thing to different people.
HubSpot documents this about its own predictive score: it needs a critical mass of closed deals before predictions stabilize. The same constraint applies to any learned model, however large the training set behind it.
A new product with no closed deals. A new segment where the buying process is different. A young company with thirty closed deals total. A pivot that made last year history misleading rather than useful.
These are not edge cases. They describe most companies under fifty people and every company launching something.
Define the criteria yourself, from what you believe a good deal looks like in this motion. Which roles have to be involved, which process steps have to start, how fast a real buyer replies.
You have something the model does not: you know what you are selling and to whom, before any of it has happened.
Then check it as evidence arrives. Ten closed deals is not a training set and it is enough to tell you whether your criteria are pointing the right way.
A score you can define is available on day one. A score that has to be learned arrives after the quarter you needed it.
Kaypo starts from weights you set, so it is useful before you have any history at all. As closed deals accumulate, the back test runs them against your own outcomes and tells you which of your assumptions held. You adjust, and it gets sharper every month.
The result goes to each rep in Slack every morning with the reason beside every number, from the first week rather than the second quarter.
Recorded in the sandbox, the same example pipeline you can open yourself. Pick any part.
Deals. Every open deal, scored on what the buyer did.
Deals can be scored without historical data, if the model is one you define rather than one that has to be learned. Set the criteria from what you know about the motion, then test them against outcomes as closed deals accumulate.
A learned model typically needs a critical mass before predictions stabilize. A model with weights you set is useful immediately and improves as history arrives.
One where you define the buyer criteria, required roles and milestones yourself. Historical baselines cannot help with a motion that has no history.
The product page for this: How Kaypo works.
Reviewed November 8, 2026 against the product as it behaves today.
30 minutes with the founder
Bring this question and your own deals. Pick a time and it is booked.
Next step
The real product on invented B2B software companies. Nothing to connect and no signup.
Open Deals in the sandboxThe whole product on an example pipeline: scores, next steps, plays and agents. No signup and no call.
$1,500 a month, or $15,000 a year, for one admin and up to 15 users. Free for 14 days, no card.
Or get the free report on your own closed dealsOr talk it through with a person, 30 minutes