From gut feel to gold standard: AI-powered pricing that ships as a product.
Pricing exhibition space is one of the highest-leverage decisions a brand owner makes across their portfolio — and one of the least rigorous. For most, it meant looking at last year's number, adding a percentage, and hoping for the best.
There was no systematic way to weigh shifting macroeconomic conditions, varying demand across brands, or the one question nobody could cleanly answer: what is the revenue-maximising price for this brand, right now?
The gap between intuition and optimum was costly in both directions. Price too low and revenue is left on the table; too high and space goes unsold. With no model to interrogate, commercial teams couldn't quantify the error — let alone correct it.
Discretionary discounts filled the vacuum, quietly eroding revenue the market was willing to pay.
Pricing a portfolio means accounting for factors that vary enormously across geographies, industries and cycles. History alone isn't enough — it must be fused with the right external signals, engineered thoughtfully, across eight families of internal and macroeconomic data.
A model that recommends prices is only useful if teams act on it. Black-box outputs get overridden on instinct. It had to show its working — so a director understands not just the recommendation, but why, and can put it in front of a client.
A one-off pricing report is a consulting engagement. A capability that generates it on demand, for any brand, wired to live data, versioned and reliable enough to sell — that is a different order of problem.
A model that recommends, not forecasts
Built from scratch on 160,000 transactions across 82 brands, the engine inverts the usual problem: it recommends the price that maximises revenue, rather than predicting revenue at a given price. A companion model, seven times faster, enabled rapid iteration without waiting on full runs.
Built for commercial teams
Outputs are price–demand curves, revenue surfaces and feature-importance breakdowns that make the reasoning transparent. The model captures price sensitivity with notable accuracy — a ±10% price movement maps reliably to ±10% revenue change across the portfolio.
Productised, not delivered once
The pricing logic was consolidated into a single versioned codebase wired to live market data, generating a board-ready report — charts, tables, scenarios, market context — in under three minutes. Days of skilled manual assembly became a capability customers could buy.
End-to-end data science and product engineering — from problem framing and feature engineering through to a versioned, in-product commercial capability. We took the engagement from a vague analytical question to a paying product.
Brands using the engine achieve 6–8% annual price increases — outperforming the B2B average by 2.7 points, with 7.2% revenue growth versus 4.5% for the sector. A 4% yield improvement alone delivers a 10× return on the engagement. What changed wasn't just the numbers — it was the confidence to act on them, with a model behind the decision rather than a hunch.