How the sales program is decided
Forecasting vs. optimizing: what is the difference? → What the numbers mean: shadow price, elasticity, break-even… → The complete cycle: from the plan to the daily visit →The model is a calculator that reviews all possible combinations of what to sell, where, when, and how much to buy, and keeps the one that yields the highest margin without breaking any business rules.
Think of it as the business GPS.
You tell it where you want to go (the highest margin) and which streets are closed (the rules). It doesn't guess: it calculates all routes and gives you the best one. If you change a data point, it recalculates in seconds.
The data
- What has been sold of each product, in each zone and month
- Price, cost, and cost of delivering it to each zone
- Inventory today and incoming purchases
- How long it takes for an order to arrive from the supplier
- Pending orders to be filled and the company's target
everything comes from the platform, no one types it in
The rules
- Selling more than the market buys (historical data + an effort margin)
- Selling what is not in the warehouse
- Exceeding the purchasing budget or warehouse space
- Leaving already sold orders unfilled
- Falling below the target (or, if allowed, paying a penalty for not reaching it)
the "closed streets"
The objective
Earn the highest marginprice − cost − logistics − warehousing, summed across all products, zones, and months
the GPS "destination"
The plan
- How much to sell of each product, in each zone and month
- How much to buy and when to order it
- How much inventory to close with each month
- Which rule is holding back profitability and how much it's worth to loosen it
and it is distributed into targets per salesperson
The engine tests, discards, and keeps the best one
For example, with 40 products, 8 zones, and 12 months, there are nearly 4,000 sales decisions, plus purchasing and inventory decisions. A person cannot review all their combinations. The mathematical engine (linear programming) can: it immediately discards those that break a rule and, among those that remain, finds the one with the highest margin.
It is not a prediction or an opinion: it is the best possible answer with the data and rules we gave it. If the data changes, the plan changes.
Two products and only 100 units in the warehouse
| How to distribute the 100 units | Cream A | Serum B | Margin |
|---|---|---|---|
| Half and half | 50 | 50 | $6,500 |
| What sold the most last year | 40 | 60 | $6,200 |
| First what yields the most | 70 | 30 | $7,100 |
Shadow price. The model also tells us how much we would gain with one more unit in the warehouse: $50, because another Serum B would be sold. This way we know if it is worth buying more inventory, paying for more space, or increasing the budget, before spending.
In real life, there are thousands of units, products, zones, and months at the same time. The idea is exactly the same.
Forecasting vs. optimizing
How does it differ from planning with time series (forecasting)? View comparison → Next reading
The complete cycle of the plan
Plan, schedule among salespeople, take it to their visits, execute the day, evaluate against the plan, and adjust. View the cycle → Next reading
What the model's numbers mean
Optimal value, activity, slack, shadow price (dual), marginal value, break-even, and elasticity, with an interactive example. View the values →
Every figure in the plan can be explained with a data point and a rule.
What if the cost rises or demand falls? Change the data point and compare.
The plan is distributed among salespeople and, combined, their targets meet the plan.