Category manager
Runs one group of products: prices, promotions, suppliers
Gets each planned promotion with its forecast range and the discount that pays most, and decides which run and at what discount
The effect of every promotion forecast automatically, before it starts
We build this system to order for retail chains. AI reads past promotions and receipts, weighs where extra sales come from (new buyers, similar products or customers buying ahead) and forecasts a planned promotion’s profit at each discount, as a range. Ordinary software works out the week from that forecast and checks the stock. A promotion that loses money is held, and the category manager decides
Who works with it
Marketing manager
Plans the promotion calendar and its budget
Sees what each promotion in the calendar is expected to bring before it starts, and what the last one actually brought
Buyer
Orders products from suppliers
Gets the forecast units for each promotion in time to order the stock for it
What comes in
What the system reads for each promotion, and where the forecast goes
Three ways out
The path of one promotion
A planned promotiona product and a discount
AIThe forecastrules and AI
- Into the promo planpays at this discount
- Category managerunsure, or stock shorta person decides
- Held for a personloses moneya person decides
Example promotions
Pick a promotion and a discount, and see what it would bring
Made-up products of a grocery and home chain. The forecast is a range: the band on the curve, never one promised number
Ground coffee, 1 lb, a week-long promotion
What the promotion would bring AIextra profit in the promotion week
no gain10% pays most0%10%20%30%
Discount
The week at 20% off, by the rules Rule
- Units sold in the week AI
- 720 (630–810) · usually 300
- Margin in the promotion week
- $2,102
- − Margin in a usual week
- −$1,590
- − Taken from similar products AI
- −$252
- − Bought ahead, missing later AI
- −$223
- = Extra profit
- $38 (−$123 … $199)
Rule In stock: 2,000. Forecast: up to 810
What AI weighed AI
- MoreSales rose by about half in the last two promotions on it
- SameMost buyers come back for it every week anyway
- MoreA competitor’s promotion on coffee ends this week
AI’s reading: a small discount brings new buyers; a deep one mostly gives the margin to regulars
Where the promotion goes AITo the category manager
The middle of the forecast pays, but the low end loses money, so the category manager decides whether the risk is worth it
a person decidesSoftware, AI and a person
Who does which part of the work
Software and AI forecast every planned promotion. A person decides which run
Ordinary software
Monotonous work with clear rules
- Collects sales, prices, costs and stock by product and store, and the history of past promotions
- Works out the promotion week from AI’s forecast: the margin, the usual week, what similar products and later weeks lose
- Checks the forecast against the stock in stores
- Holds a promotion that loses money on the forecast
- Writes the forecast and the route into the promo plan
AI: understand, weigh and choose
The small decisions on every promotion
- Reads past promotions: how much more sold, and where the extra buyers came from
- Forecasts the units a promotion would sell at each discount, as a range
- Weighs how much would come from similar products and from customers buying ahead
- Finds the discount that pays most on its forecast
- Says when it has too little history, and leaves the promotion to a person
A person
Every promotion and every exception
- The category manager decides which promotions run and at what discount
- A person decides whether to run a promotion that loses money
- The category manager decides on products AI cannot forecast
- The buyer orders stock for each promotion
- Marketing plans the calendar and the budget, as before
How a project starts
Four stages. After each one, you decide whether to go on
First on past promotions, then on planned ones
- Usually firstDiscovery and specification
- In
- How promotions are planned and judged today, and the sales history you keep
- You get
- A written specification: which categories come first, what data the system needs, what it will cost and in what order. The document is yours whether or not we build it
- ThenTest on past promotions
- In
- Last year’s promotions, with sales before, during and after each
- You get
- What the system would have forecast for each, next to what each actually brought
- ThenPilot on one category
- In
- The planned promotions in one category you choose
- You get
- A forecast range and a route for each. The category manager decides as usual and compares the forecast with the result
- When you decideRollout
- In
- The categories you choose
- You get
- Every planned promotion there gets a forecast before it starts. Your team changes the rules
Six questions we ask first
- How many promotions do you run a year? We plan the system and the pilot around it
- What share of sales happens during promotions? The test on past promotions starts with the biggest
- How deep are your usual discounts? The forecast covers that range first
- How do you judge a promotion today: by sales, by margin, or not at all? The test compares the system’s forecast with it
- Do you keep sales and stock history by product, store and promotion? The forecast is built from it
- Which cash register, ERP and analytics systems do you use? The pilot connects to them
Tell us how promotions are planned today
A 30-minute call. We will say what your promotion history allows and where a project would start