When you run a Price Optimization strategy more than once, you may see the suggested price for a product move slightly between runs, even though nothing obvious changed. You may also notice that the suggested price is a little different from the optimal price the system displays. Neither is a bug. This article explains what's happening so you know when to act and when to leave things alone.
Why the price moves a little each run
Most of the time, the system suggests the price it currently believes is best for your objective. But every so often it tries a nearby price on purpose, so it can keep watching how your customers respond and keep its demand model sharp. If prices only ever sat at a single point, the system would slowly lose sight of how demand behaves just above and below that point, and stop adapting to real-world changes.
The key thing to know is that this movement is kept within tight limits. The system only tests prices close to the one it already trusts, so it never jumps to a surprising or extreme number. Small run-to-run wobble is the system learning, not losing control.
It's also worth knowing that what the system has learned about your products stays stable between runs. The demand patterns behind the numbers don't get rewritten each time. The only part that carries deliberate variation is the final pick at the end. So a small difference between two runs doesn't mean the system changed its mind about the product. If anything, it's trying to learn more.
Optimal price versus suggested price
The results can show two related numbers, and it helps to know what each one means:
The optimal price is the system's best estimate of the ideal price for your objective. Think of it as the target.
The suggested price is what this particular run actually chose. Usually it matches the optimal price, but sometimes it sits a little to one side because the run is testing a nearby price.
Over time these two numbers tend to move closer together. As the system gathers more evidence about how customers respond, its target becomes steadier and the suggested price settles onto it.
Newer sales matter more than old ones
When the system learns from your sales history, it does not treat every sale as equally important. Recent sales carry much more weight, and older sales count for steadily less the further back they go. A sale from last week tells the system far more about how customers behave today than a sale from a year ago.
This is why a suggested price can shift as fresh sales come in: the system is following your current customer behaviour rather than being anchored to old patterns. It also means that after a real change in demand, for example a new season or a shift in what customers are willing to pay, the suggestions catch up on their own as new sales arrive.
What this means for you
Small differences between runs, and between the optimal and suggested price, are normal. You don't need to do anything about them.
The movement is always bounded, and your safeguards still apply on top, so a run can never produce a price outside the limits you set. See How to Add Safeguards to Your Pricing Strategy.
If you want prices to settle down faster, the best thing you can give the system is steady, up-to-date sales data. Next best thing is to wait and let it learn.
For the wider picture of how a suggestion is built from start to finish, see How Price Optimization Works.