For a printing house in Poland, adBrain built a price- and discount-optimisation system that recommends discounts across both offline and online channels. The model balances the chance of winning an order against the margin given up, so pricing decisions are driven by data rather than gut feel. It is a clear case of adBrain’s decide capability: a live decision, made under uncertainty, with guardrails.
The problem
Printing is a high-mix, quote-driven business: every job is a little different, and the discount that closes one order would be a giveaway on another. Discounts were being set by intuition and habit, inconsistently across staff and channels. Too high, and margin evaporated; too low, and orders went to a competitor. Nobody could see the trade-off clearly enough to optimise it.
What we built
- Demand- and margin-aware discount recommendation. The model estimates how likely an order is to convert at a given discount and recommends the level that best balances win probability against retained margin.
- One model, two channels. The same intelligence powers online discounting (shown to the customer during the order) and offline recommendation (shown to staff at the counter), so pricing is consistent everywhere.
- Guardrails and overrides. Recommendations respect minimum-margin floors and business rules; staff retain the final say.
How it works
| Stage | What happens |
|---|---|
| Quote context | Job attributes, volume, customer, and channel are read. |
| Estimate | The model estimates conversion likelihood across discount levels. |
| Recommend | It returns the discount that maximises expected margin. |
| Apply | The discount is surfaced online to the customer or offline to staff, within margin guardrails. |
Why it matters
Discounting is one of the few levers that moves both conversion and margin at the same time, which is exactly why it is so easy to get wrong. Putting a model behind it turns a gut decision into a measured one, and applying the same model online and offline removes the inconsistency that quietly erodes profit. Where the data supports it, the effect of the model is measured against the prior baseline rather than assumed.
The expertise & the challenge
This project is part of adBrain’s machine learning decisioning and predictive & causal modelling practice, written up as Machine Learning & Decisioning. It answers the challenge we discount by gut feel and leave margin on the table. Price and discount optimisation extends naturally to dynamic pricing, promotion planning, and margin analytics across retail and services.