Services
Demand forecasting, inventory optimization and production scheduling.
Everything hangs off the data. A forecast is only as good as the sales history feeding it. A buy plan is only as good as the forecast. A production schedule is only as good as the buy plan. Get the data right and the other three get easier.
Start anywhere below.
Demand Forecasting
Most forecasts tell you what they think will happen. Ours also tells you how wrong it has been lately.
Underneath, it does what any good forecast does: it looks at how much you normally sell, whether that is climbing or falling, and how your year is shaped. Then it does the part most tools skip. It checks its own recent misses, works out which way it has been leaning, and corrects the next month before the month is over. A normal forecast finds out it was wrong a quarter later. This one finds out now.
901 Horizon is our demand and inventory planning platform. One place to forecast demand top-down and bottom-up, track accuracy and bias, plan new products and the sales they take from existing ones, set inventory targets and buy plans, and carry demand through to components. Watch the demo below for a quick tour.
Try 901 Horizon on your own company
Go to the demo tool itself and put in your company’s details. Horizon pulls in data on your business and builds a version of the platform tailored to your company: the demand forecast, inventory position, buy plan and component needs, all on your own products.
Purchasing & Inventory Optimization
Hold less stock and still have what people order. Both, not one at the cost of the other.
Almost every business we meet has too much of the wrong thing and not enough of the right thing at the same time. That is not a buying problem, it is that nobody ever set the rules: how much cover each item needs, and when to order more. Below, decide how often you want to have things in stock and how much cash you can put into it, and watch which orders survive.
901 Horizon does all of this as part of the platform, from inventory targets through to the buy plan. Below is a separate tool to play around with so you can see how it works: set the service level and the budget, and watch the buy plan change.
How often a customer gets exactly what they ordered. 100% is ruinously expensive.
What suppliers actually deliver against what the system assumes.
The constraint nobody models. When it binds, something has to give.
| SKU | Fcst / mo | Safety | ROP | Buy | Spend | Status |
|---|
What the last point of service costs
Average inventory at every half-point of service level. Flat to about 95%, then a wall.
Where the money goes
Green is fully funded and teal is trimmed by the cash ceiling.
Manufacturing & Production Scheduling
The order you build things in is worth hours a day.
Every time the line switches from one job to the next, somebody cleans something, swaps something, and waits. Run similar jobs together and most of that disappears. The catch is that ten jobs can be run in three million different orders, so planners fall back on habit, and habit is expensive. Drag the jobs below into whatever order you like, then let the computer try to beat you.
Sequence · drag to reorder
Keyboard: focus a job, hold Alt with up or down.
The production block
Synthetic data, not a client schedule. The optimizer runs nearest-neighbour from every start, then 2-opt. It is the same approach as the scheduler we built in Python for a door manufacturer, which runs against their live order book daily and took on-time delivery from 54% to 97%.
Data Management & Analytics
This is the center. Everything above reads from it.
We pull the numbers out of every system you run: the ERP, the till, the spreadsheets somebody keeps on their desktop. Then we make them agree with each other and keep them in one place, set up under your account rather than ours. If you fire us on a Friday, everything still runs on Monday.
Hover anything for detail. Click any bar, slice or bubble and every other panel filters to it.
Total sales over time
Monthly, with a twelve-month moving average
Mix by product family
Flips to channel once a family is selected
Top selling products
Within the current selection
Growing
Year over year
| SKU | 12 mo | YoY | Change |
|---|
Declining
Year over year
| SKU | 12 mo | YoY | Change |
|---|
Stock position
Cover against demand, sized by inventory value
Where the inventory sits
Cycle stock versus safety stock, at cost
Projected cover
What to do about it
Thinnest cover first
| SKU | Fcst / mo | Cover | ROP | Buy | Spend | Position |
|---|
The bottom line: the media works, but it does not pay
Stores took turns running double the usual media and running none. Over three periods, stores with media doubled beat stores with it off by +1.6 points of year-over-year sales growth, after removing seasonality and the built-in differences between the two groups. The lift is real (p = 0.02) and pointed the same way every period. But each extra media dollar brought back only $0.68 in sales. At 5% profit flow-through, the extra spend lost about $2,898 per store, or $318,780 across the chain.
Sales lift
Doubled stores vs. off stores, YoY growth
The dot is the best estimate; the band is the 95% range (+0.3 to +2.9). The band clears zero, so “no effect” is ruled out.
Return on media
Extra sales ÷ extra media spend
95% range 0.13× to 1.23×. To pay back in profit at 5% flow-through it would need 20×. This is incremental return, far stricter than what ad platforms report.
Profit impact per store
Over the three test periods
Profit on the extra sales minus what the extra media cost. Only the profit counts against the spend, not the full sale.
Same answer every period
Doubled group minus off group, YoY growth
Swapping which stores get the media, and seeing the gap in the same direction each time, is what weather, seasonality or a lucky split cannot fake.
The question every marketing budget should answer: would these sales have happened anyway? Platform dashboards count every sale near an ad. A holdout test counts only the sales the ad added. Sometimes the answer is to spend more. Here it is to spend differently.
D3 running in your browser on generated data. In an engagement this is Power BI or Tableau against your warehouse, but the rule is the same: a few views tied to recurring decisions, and every visual filters every other one.
Seen enough?
Tell us which of these sounds like your business, or where the data runs out before it answers the question. Scoping a full project plan costs you nothing.