Idmon builds demand forecasts and stock plans for small and mid-sized businesses that hold stock. We are data analysts. We use AI the way a good analyst uses any tool: to do the work of a whole team, under our direction, without a team's timescale or a team's bill. We aim to put a working prototype in front of you within 24 hours of receiving your data, at a fraction of what a consultancy charges.
Most of the businesses we work with recognise at least two of these.
Scroll through to see the supply chain of a typical stock-holding business, and where Idmon sits in it.
Suppliers ship to your warehouse. Your warehouse ships to customers. Somewhere in the middle, somebody decides what to order and when.
That decision is usually made on experience and a spreadsheet. Guess low and the racks run empty. Guess high and cash sits on the shelves for months.
Idmon reads what you sold, forecasts what you will sell, and turns that into what to hold and what to order. It feeds purchasing; it does not replace it.
The right items in the right quantities, orders placed in time, and a measured fill rate against your own levels, not a promise.
This is the same process a good analytics team would follow. The difference is who does the heavy lifting. At each stage an Idmon analyst decides what to do and why; a pipeline of agents does the hours of work that decision implies; the analyst judges the result and moves on or goes back. That is how the whole process runs in about a day rather than months: you have a working prototype to critique within 24 hours of us receiving your data. You are involved where it matters: agreeing the question, reacting to what we find, and receiving the result.
Whether this will work depends on your data. A 30-minute call is enough to find out.
Book a consultationA simplified version of the review app we deliver. Pick a product. Every forecast carries its range, its test score in plain English, and the months the model never saw.
The app looks like Idmon. What it shows is yours: your product names, the units you plan in, the questions you asked at intake. And the way it works is not fixed: if you want the horizon longer, the range wider, a different score, or a queue of products that need a human decision, you ask and we change it.
Same question. Very different journey.
Every model is tested against months we held back, and its measured error sits beside every forecast, so you know how far to trust it before you rely on it.
A new product or a new market has no history to test against. We can still forecast it, but the forecast will say plainly that it is untested, and how we arrived at it.
Idmon tells you what is likely coming and what that means for stock. Deciding what to order is still your call, and your judgement matters.
What to forecast, how to treat messy data, which model to trust: those are an analyst's judgements. The agents do the work those judgements call for.
It is used for your engagement and deleted afterwards, unless you ask us to retain it for updates.
If we cannot show you how a number was made, we do not report it.
We do not publish a price list, because the right figure depends on how many products you sell, how many data sources we reconcile, and how far ahead you need to see. We do tell you the shape of it, and the exact quote comes out of the first call, before any work starts.
Need more than a forecast? Margin, stock cover, customer patterns, several data sources at once. We scope that separately, with the same analyst, the same process and the same fixed-fee approach.
Sales by product by day or week, going back as far as you have. Two years is comfortable; one is workable. Anything else you hold helps: prices, promotions, stock-outs, returns, stock on hand. An export from your ERP or a spreadsheet is fine.
We aim to have a working prototype in front of you within 24 hours of receiving your data. You critique it, we refine it, and then the time is yours to put it to use in the business. Messier data or a very large range can take longer, and we tell you at intake rather than discovering it later. We will need you for the first conversation and the final walk-through, and little in between. The analyst and the pipeline do the rest.
Most of it is. Preprocessing exists for exactly this: duplicate SKUs, renamed products, missing weeks, currency changes. We show you what we found and what we did about it before anything is modelled. If the data cannot support a forecast at all, we say so on the first call.
No. It is used for your engagement and deleted afterwards. If you take a refreshed forecast, we keep only what is needed to run the next refresh, and you can ask us to delete it at any time.
An app where you can explore every forecast, product by product, with the range, the test score and the history behind it, and the stock levels and orders it implies. A short written summary of what the analysis found. And the complete audit trail: every number, and the command that produced it. All of it is yours.
Before choosing a model for each product, we hold back the most recent history and test every candidate against it. You see the error each model made on data it had not seen. We pick per product, and we report the error alongside every forecast.
Thirty minutes, no obligation. You leave knowing whether your data can support a forecast at all, and what it would take.
What happens next: