Dynamic pricing: how to measure the real effect on your revenue

Flotly Team - published 10 April 2026 · 6 min read

A method to test the effect of variable pricing on your revenue without confusing a hypothesis with a measured result. No promised figure: a measurement protocol.

The promise is everywhere: “Increase your revenue by 30% with AI!” A figure like that means nothing without three things: a reference period, a fleet scope, and the list of other changes made during the test. Without them, it is not a result, it is a sales pitch.

This article doesn’t promise you any percentage. It gives you the protocol to measure your own, and to recognise, when you read a case study, whether the figure quoted was measured or made up.

Two technologies sold under the same word

A rules engine applies the rates you have written yourself: weekend surcharge, discounts beyond seven days, specific rates by vehicle category or by customer profile. It is deterministic, auditable, and it is what most rental software does, Flotly included.

A predictive engine, on the other hand, estimates future demand from historical data and suggests a price nobody wrote in advance. It is a different technology, a different cost, and a different level of proof to demand from the supplier. Before paying for “AI pricing”, ask which of the two you are buying: the question alone often clears up the conversation.

The four indicators to track

The average rate is not enough: it rises mechanically as soon as you raise prices, even if you rent less. Track these four indicators together.

Revenue per available day: revenue ÷ (number of vehicles × days in the period). This is the main indicator, the only one that moves with both price and occupancy.
Occupancy rate: days rented ÷ days available. It tells you whether a rise in revenue comes from price or from volume.
Average value per rental agreement: to tell “selling at a higher price” apart from “renting for longer”.
Margin after variable costs: cleaning, fuel, wear, commissions. Revenue that rises while the margin falls is not a gain.

The protocol, in five steps

Freeze a reference period of at least four weeks, with unchanged rates, and record the four indicators.
Change only one thing: the rate grid. No new advertising campaign, no added vehicle, no website redesign at the same time.
Let it run for at least four weeks over a seasonality comparable to the reference period or, failing that, compare with the same period the previous year.
Keep a log of everything that changed without your doing: a competitor closing, a local event, a vehicle off the road, unusual weather. This log is what will make your conclusion defensible.
Compare the four indicators, not just revenue. And accept that a zero difference is a result, not a failed test.

What skews a test

Changing several things at once: you will never know which one produced the difference.
Comparing August with November: the season outweighs everything else.
Testing on three vehicles: the difference you see will be noise, not a signal.
Stopping the test as soon as the curve goes up: that is exactly how the percentages you read everywhere are born.

The trap: overreacting to demand

Bad dynamic pricing triples prices for a local event and kills conversion. Good dynamic pricing aims for maximum revenue, not the maximum rate. Those are two different things.

So set limits before you start: a floor above your break-even point, and a ceiling you would be comfortable showing a loyal customer. A rule you couldn’t explain over the phone is a rule to remove.

What nobody tells you

The first weeks of variable pricing are unsettling. Your regular customers see prices moving, and you need to be able to explain why. Two simple safeguards: show a price range rather than an instant price, and keep a stable loyalty policy, outside the adjustment mechanism.

A pricing engine doesn’t replace business judgement. It applies a rule faster and more consistently than you do. That is useful, and that is all it is.

Our position, to be clear

Flotly doesn’t have a customer base large enough to publish an average gain, and we won’t do so as long as that is the case. Our pricing engine applies the rules you configure (duration, weekend, vehicle category, customer profile) and doesn’t predict demand for you. If you want a figure, it will come from your own test, using the protocol above.