Dynamic pricing: the same maths can serve profit and planet
If you tried to buy a World Cup ticket this year, you've met dynamic pricing at its worst: a price that climbed while you watched. Fair enough — the anger is earned. But the technology itself is not to blame. The math is neutral. It does exactly what you tell it to.
Flash summary
Dynamic pricing is just a price that moves with conditions. Two ML models do the work: a forecaster (what will people want?) and a decider (so what's the right price?). A human still writes the goal — and that goal is everything. Three European companies are pointing the same maths at food waste, clean energy and e-waste, profitably.
What it actually is
Dynamic pricing is simply a price that moves. Instead of one number printed once and left alone, the price adjusts as conditions change: how much stock is left, how many people want it right now, the time of day, the weather, what it costs to make or store the thing. Airlines have done it for decades. Your supermarket does a version of it every time it slaps a yellow sticker on bread an hour before closing.
The only new part is the engine doing the deciding. A human can re-price a shelf of bread. A human cannot re-price two million products every fifteen minutes while reading demand as it shifts. Software can. That's where machine learning walks in.
The ML bit, in simple words
Under almost every modern pricing system, two models quietly do two different jobs.
- The forecaster: "what will people want?" It studies history — last week, last year, this exact Tuesday in past seasons — alongside live signals like weather and competitor prices. Techniques with intimidating names (gradient boosting, deep sequence models) are really just very sharp pattern-spotters. They turn "it feels busy" into a number.
- The decider: "so what's the right price?" Often built with reinforcement learning, it works like a chess engine improving: try a move, see the result, update the instincts. Here the "move" is a price, and over thousands of small experiments it gets good.
The crucial, under-appreciated part: a human still writes the goal. "Maximise revenue at all costs" and "sell everything before it spoils without gouging anyone" produce very different behaviour from the exact same algorithm. The model has no opinion. We do.
Fair and profitable were never opposites. You just have to build for both.
Three companies aiming it at the planet
Too Good To Go — pricing against waste
As a restaurant or supermarket closing time nears, the model marks down food that would otherwise be thrown away, matching surplus to someone who'll happily collect it. In 2025 the platform helped rescue just under 157 million meals, which it estimates avoided around 424,000 tonnes of CO₂e. That is dynamic pricing optimising for "don't waste this."
Dynamic energy tariffs — pricing toward clean power
Providers like Tibber and Frank Energie pass through the real, hour-by-hour price of electricity. When wind and sun are abundant, power is cheap and clean, and the price drops — nudging car charging and dishwashers into the greenest hours. Dutch households on dynamic contracts grew from under 50,000 in 2022 to roughly 423,000 by early 2026. The savings go to households that actually shift their usage, so the tool rewards behaviour; it doesn't replace it.
Back Market — pricing to keep devices alive
On a refurbished-electronics marketplace, pricing models keep used phones and laptops moving at prices that clear stock and widen access. A working device stays in a pocket instead of dying in a drawer — and every reused phone is one that didn't need to be manufactured from scratch. The optimisation target is circulation, not extraction.
The point
Same technology. Three different goals. These companies pointed powerful mathematics at problems worth solving. The revenue still arrives. It just arrives alongside meals saved, gas displaced, and phones kept alive.
FIFA pointed the same maths at "extract the maximum." Too Good To Go pointed it at "waste the minimum." The technology didn't pick a side. A human did — when they set the target. That's the most important line of code in any pricing system, and it isn't in the model at all.
What would this maths find in your operations?
Pricing is one of a dozen places where the same models serve profit and planet at once. An AI Value Scan™ maps and scores them for your operation — before you build anything. The goal you choose is the strategy; we help you find the goals worth choosing.
Beyond GenAI: AI’s big picture →
Sources: Too Good To Go Impact Report 2025 (≈157M meals, ~424k t CO₂e); Dutch dynamic-tariff adoption via IO+ and NL Times; pricing ML mechanics via Grid Dynamics and peer-reviewed surveys. Figures current as of June 2026.