Insight
Who might stop buying, who might not pay, how much you will sell next month. Models trained on your data, with the margin of error stated up front.
Who this is for
Reports say what happened. At some point the question changes.
The budget is set in December and the variance shows up in March. In between there is nothing to steer by.
You learn a customer is gone when the order does not arrive. The signals had been in the data for months.
You have a total overdue balance. You do not have a list of who is riskier than whom.
What you’re losing now
Not in the abstract. In decisions taken a month later than they could have been.
A customer kept with one call at the right moment is cheaper than one found from scratch.
Without an estimate of collections, planning payments is an exercise in optimism.
An average catches neither seasonality nor shifts in demand. The result is capital tied up or an empty shelf.
What we deliver
Not a model — answers to questions you already ask. Each with its own margin of error.
On open invoices, a score that says who is riskier than whom. Not a bad-payer label — an order in which to call.
Built on signals from your own data: how long since they ordered, how the frequency changed, how many returns they made.
An estimate for next month, as a range rather than a single number. A planning reference, not a budget target.
Every model comes with its accuracy measured on data it did not see during training. If a model is not good enough, we say so and do not ship it.
How we work
Not every question has an answer in the data you hold. That is settled in the first week, not at the end.
Duration: 1 conversation
We pick one or two questions, not ten. One prediction that gets used beats five that get built.
We check whether there is enough data and enough history for a model to learn from. If there is not, we say so here.
The model trains on part of the history and is tested on the rest, which it has not seen. That is where the reported accuracy comes from.
The list of at-risk customers or the forecast lands on the screen you already open, not in a separate report.
The timeline depends on how much history you have and how clean it is. We agree it after the check in step two, not before.
What changes
Departed customers surface in the quarterly report, and the sales forecast is an average of recent months.
A live list of at-risk customers, and a forecast with a confidence range next to it.
What stops you
From the same stage
Prediction needs clean history. Profitability and the sales analysis are what produce it.
Know exactly where you make profit, where you lose, and how your cash-flow will look in 3–6 months. With “what-if” scenarios.
Understand which clients bring value, where you’re losing opportunities, and where your sales team should focus.
The metrics that matter, on one screen, updated automatically. Not fifty charts — the indicators you actually run the business on.
We analyse every table in your systems and surface what falls outside the pattern: repetitions, missing counterparts, unusual values, concentrations. No hand-written rules.
Next step
Tell us which decision you want to make better. We’ll tell you whether we can help, and how, concretely.