Cut the price from $24.99 to $23.45 after the best rival offer undercut you to $23.40 from $24.10, 12 minutes ago. Matching the lowest offer would have earned $0.62 less per day. Featured Offer odds go from 41% to 68%.
Amazon repricing · US · CA · UK · DE
Every price change, explained.
Marginary reprices your Amazon listings and writes down why — in a sentence that still makes sense a week later. Nothing is priced until you have given an item a cost and a floor, and one button stops all of it.
Held at $18.99 instead of moving to $17.80 because a safety rule stopped the move. Your minimum price of $18.99 stopped it going lower.
Raised the price from $31.00 to $32.40 because expected daily profit peaks there. These odds still lean on the general model, with only 4 recorded observations on this listing.
Every check writes one of these, including the ones where nothing moved. Sentences above are illustrative — the structure is the product's, the figures are samples.
What you actually get
A repricer that has to show its working.
Seven things it does, and one thing it will never do: move a price without leaving you a reason you can read.
The answer is a sentence, not a probability
"72% Featured Offer" tells you what a model believes. It doesn't tell you why your price moved.
Marginary writes one plain sentence per decision — what it did, what the competition did to cause it, what that beat, what stopped it going further — assembled from what the cycle already worked out and stored. A decision from last Tuesday reads exactly as it read on Tuesday.
Rule #4 fired. Price updated.
Cut the price from $24.99 to $23.45 after the best rival offer undercut you to $23.40, 12 minutes ago. Matching the lowest offer would have earned $0.62 less per day.
It tells you when it's guessing
A brand-new listing has no history, so its odds are the general model wearing a percentage sign. Marginary says so — in the sentence — until that listing has enough of its own measured observations to speak for itself.
As this listing records its own behaviour, the bar fills and the general model steps back. The sentence changes with it.
Every rule keeps a scoreboard
Seven days per rule: how often it fired, what it did to your Featured Offer share, and what it was worth per day — labelled as modelled, because it is.
A rule that fires constantly and improves nothing is a rule you should delete, and you shouldn't have to read a log to find it.
412 fires +9.2 pts
88 fires +2.1 pts
1,304 fires −0.3 pts
Fires most, changes least. Worth deleting.
What it refused to do is written down too
Holds are recorded with the same care as changes. Your floor, your per-change cap, your daily and hourly limits, the price-war brake — whichever one bit gets named.
A repricer that silently does nothing looks identical to a repricer that's broken, right up until you check your sales.
Held at $18.99 instead of moving to $17.80 because your minimum price stopped it going lower.
Held at $42.50 because prices on this listing have fallen four times in an hour. Marginary stopped following them down.
Held at $11.25 because this item has already changed price 3 times today, which is the limit you set.
That floor is worked out, not typed in
You think in "never below 15% return", not "never below $18.43". Marginary takes what you paid plus inbound shipping and other costs, runs the full Amazon stack — tiered referral rates, FBA fees by size tier and weight band, storage, the tax Amazon adds to its own fees in Canada — and inverts it.
An item with no cost basis is never priced at all. That's checked before every single decision, not once at import.
Change a cost and the floor moves with it, instead of going stale.
The price it picks is an answer, not a reflex
Undercutting maximises your chance of a sale, not your money. Marginary estimates how much of the day you'd hold the Featured Offer at each price in your range, how many units that moves, and what each one earns after fees — then picks the peak.
When you'll run out before your restock lands, it prices up rather than selling next week's units cheap today. All of which it will explain, in the sentence.
Illustrative curve. The real one is fitted per listing from that listing's own recorded history.
See the decision before it happens
Ask what a price would do before you set it. Any candidate price runs through the same fee model, odds and demand curve the live engine uses, so the preview and the engine can never disagree.
Replay a strategy against a listing's own recorded market history to see what it would have decided. It's honest about its limit: it can tell you what these rules would have chosen in those conditions, not how competitors would have reacted.
Change a cost, a stock level or a fulfilment method and the whole ladder recalculates. Nothing is written.
Objection, handled
The things it will not do, no matter what you tell it.
Two of these cannot be switched off. The rest are yours to set, and they apply to everything — your own rules, our decisions, and manual changes alike.
Your price never leaves your range
Not by a rule, not by the optimiser, not by accident. You set a lowest and a highest price per item, and nothing crosses either.
An item with no cost basis is never priced
A listing pulled from Amazon knows what it sells for and nothing about what you paid. Without a real floor it sits blocked, with a line telling you what's missing.
Biggest jump in one change
As a percentage, a dollar amount, or both. The smaller one wins.
Changes per item per day
So your listing doesn't look unstable to buyers.
Changes across your whole shop per hour
One cap on everything, so nothing runs away.
Smallest change worth making
Pointless one-cent moves get skipped.
Price endings
Round to .99, .95 or .00. Always down, never up.
Price-war brake
When prices on a listing keep falling fast, Marginary stops following them down.
Wait while Amazon is selling
Amazon usually wins on price. Sitting still beats losing money.
Emergency stop
One button freezes every price where it is. Marginary keeps recording what it would have done, so you lose nothing by using it.
An item with no cost basis is never priced.
Checked before every single decision, not once at import. Nothing you have not finished setting up can be moved by anything — not a rule you wrote, not the optimiser, not a manual change.
Getting started
Four steps, and you can stop and come back.
Connect Amazon
You grant permission to read your prices and, later, to change them.
Get your items
Pull your catalogue straight from Amazon, upload a CSV, or paste one in. Imported items arrive paused. Run the pull twice and it updates rather than duplicating.
Tell us your costs
What you paid, shipping in, anything else. This is what makes the floor mean something.
Pick how to price
Start with Profit maximiser. It optimises inside your range, backs off when Amazon holds the Featured Offer, holds in a price war, and raises prices when you're running low on stock. After that it runs on its own, as often as you choose — every five minutes through to once a day.
Pricing
One thing to decide: how many items you're pricing.
A plan is one number — how many items Marginary prices for you at once. Paused and archived items cost nothing.
Every plan is the whole engine — you are buying capacity, not capability. Starter's first fourteen days are free. Full pricing details
Before you ask
Questions sellers actually ask.
Will it undercut my way to the bottom?
What happens if Amazon starts selling my item?
Which marketplaces does it work on?
What if I import items before I know my costs?
How do I know something's gone wrong?
Are the sales numbers real?
Read a fortnight of its decisions, free.
Marginary runs every check against your real catalogue and writes down every decision and the reason for it. Nothing moves until an item has a cost and a floor, and one button freezes the lot.
14 days free on Starter · emergency stop from day one