Your sale goes live Friday at nine. By twenty to ten there are nine emails saying the same thing in nine different tones: I bought this on Tuesday, it is 25 percent off today, can you refund me the difference.
Somebody now does this nine times. Find the order. Work out what the buyer actually paid, which is not the list price because they used a welcome code. Work out what the item costs today, which is not the list price either because the sale is an automatic discount. Subtract. Decide whether a Tuesday order is inside a window nobody has written down. Open the refund screen, override the amount, untick restock, add a note, reply. Four minutes each when nothing is strange, and something is usually strange.
A merchant put this to the Shopify Community in April 2025. They were calculating the price difference after each promo and issuing partial refunds by hand, and they asked whether anyone had found a more efficient or automated way to manage these post-purchase refund requests. The single reply agreed the task sounds simple, pointed out that discount codes and dynamic offers make it harder, asked two clarifying questions, and that was that. The thread is still open. That is the state of the art.
The short answer
Shopify gives you the partial refund and nothing else: no price adjustment setting, no price protection window, no policy field, no price history, no Flow trigger for a price change and no Flow action that issues a refund. So the work is yours to build.
Snapshot variant prices daily and before every bulk edit, classify each drop as a promotion, a mispricing or a permanent reprice before anything refunds, compute the per-unit difference after allocated discounts on both sides, apply a published policy of window, minimum, cap and exclusions, settle in store credit by default and cash above a threshold, and reach eligible customers before they have to ask. The judgment, telling a sale apart from a typo and reading a request that arrives as a sentence rather than a form, is why this belongs with an agent rather than a condition builder.
Shopify has no price adjustment anything
There is no price adjustment feature, no price protection window, no policy field. What exists is the ordinary partial refund, and the mechanism is fine: open the order, click Refund, enter a quantity, then overwrite the Refund amount field in the Summary with the figure you actually mean, deselect Restock items because nothing is coming back, and submit. You can keep issuing further partial refunds until you reach the total available to refund, and you can send the money to the original payment method, to store credit, or split it across both.
So the refund works. Everything around it is missing, in three specific ways.
Flow cannot see a price change. The product triggers are product created, deleted, status updated, variant added, variant deleted, variant back in stock, variant out of stock, and variant inventory quantity changed. None fire on price. A long-running community thread says it plainly: there is no Shopify event for price updated. On 6 October 2026 a Shopify staffer added to that thread that the Flow team is testing a product updated trigger which can fire on selected product and variant fields including price, limited to testers, with no general availability announced. Until it ships, the native approach is a scheduled workflow that wakes up daily and goes looking.
Flow cannot issue a refund. The order actions are tags, note, cancel order, capture payment, mark as paid, archive, add a line item, send invoice, metafields and get order data. Refund exists only as a trigger, refund created, never as an action. There is no create gift card action and no store credit action either. The write is reachable only through Send Admin API request.
Shopify keeps no price history. The order line item records what the buyer paid, so that half is permanent. The product record holds only today's price. The moment the bulk edit that starts your sale finishes running, the before state is gone unless you captured it yourself.
Put those together and the native answer to what did this customer overpay is a person with two tabs open and a calculator.
The number is harder than it looks
Four things make the arithmetic wrong more often than anyone expects.
List price is not what they paid. If the original order carried a code, Shopify allocated that discount across the line items. The honest comparison is per-unit price paid after allocated discounts against the per-unit price that same cart would pay today.
List price is not what it costs today either. Automatic discounts, volume breaks, market-specific pricing and B2B catalog prices do not live on the variant. A 25 percent automatic discount shows a $100 list price and charges $75.
Some orders have already been adjusted. The second promo of the month catches the same customer twice, and the second calculation has to run against the already-reduced effective price, not the original.
Shipping thresholds move. Take $20 off a $95 order that earned free shipping over $85 and you have quietly given away the shipping too.
Get this wrong in your favor and you look cheap in public. Get it wrong the other way across a 200-order sweep and the adjustment costs more than the promotion did.
The costs nobody counts
The visible cost is the refund. Three more sit under it.
You keep the fee on money you gave back. With Shopify Payments, refunds do not return your credit card fees, and the same holds on a third-party gateway in a country where Shopify Payments is available. Only merchants on a third-party gateway in a country without Shopify Payments get transaction fees back proportionally, and only on refunds issued through the admin. Every adjustment is the difference plus processing you have already paid on it.
It lands in your numbers as a refund, not as promo spend. A price adjustment is a discount granted late, and Shopify files it under refunds. Your promotion looks cheaper than it was, your refund rate looks worse than it is, and nobody can tell you what the last sale actually cost.
The requests you never see cost the most. The customer who does not email returns the whole order and rebuys at the sale price, which costs you a label, a restock, a possibly scuffed unit and the processing twice. Or they open a dispute. Or they say nothing at all and simply wait for the next sale, forever.
Time matters too. Refunds are generally possible for up to 120 days after the order, with anything older depending on the customer's bank. A request sitting in a shared inbox for five weeks is not a free request.
The playbook worth building
As a Dugong playbook, in plain language:
# trigger
Every price change, plus every
inbound price adjustment request
# steps
1. Snapshot prices daily and before
any bulk price edit
2. Classify each drop: promotion,
mispricing, or permanent reprice
3. Pull orders inside the window
holding that variant
4. Compute the real per-unit delta,
after allocated discounts on
both sides
5. Apply the policy: window, floor,
cap, exclusions, already adjusted
6. Settle: store credit by default,
cash above the threshold, code
where neither fits
7. Speak first, before the customer
has to ask
8. Book it against the promotion,
not against refunds
Step two is the step that prevents a disaster. Not every price drop is a sale. A typo correction from $249 to $24.90 and back, a currency rounding fix, a clearance markdown on a discontinued line, a supplier-driven reprice: all of them look identical to a rule that only reads a number. Refund against the typo and you refund most of the product.
Step six decides what this costs you. Store credit keeps the money in the business, and most customers take it when it is offered in the first reply with a straight explanation rather than after two rounds of argument. Cash to the original method is the right answer above whatever figure makes credit feel like a dodge, which for most baskets sits somewhere between twenty and fifty dollars.
Step seven is the entire return on this build. The same twelve dollars buys a completely different reaction depending on who speaks first. Sent unprompted on Friday morning it reads as a store that pays attention. Sent on Monday after two emails and a screenshot, it reads as a store that had to be pushed.
The policy is the configuration
Most stores do not have a price adjustment problem. They have an undocumented one, which is why every request turns into a small negotiation and two people give different answers in the same week.
Write down five things and publish them. The window in days from the order date. The smallest difference worth processing. The largest you will adjust without a manager. The exclusions, meaning final sale, clearance, bundles, gift cards, wholesale and orders that already used a code. And what the customer receives. Published, that page stops a good share of the requests from arriving and pre-answers the rest. Written down, those same five numbers are the whole configuration of the automation. Everything after them is reading email and doing arithmetic.
Where AI earns its place
Two judgment calls here, and neither of them is subtraction.
The first is on the way in. These requests never arrive as a clean form. They arrive as the blue hoodie I got last week is cheaper now, with a screenshot, sometimes about a variant the customer did not buy, sometimes about a competitor's price, sometimes from somebody who has asked three times this quarter. A model matches the sentence to the order and the line item, reads the screenshot, runs the real calculation on both sides, checks it against your policy and drafts the reply with the actual number in it. What reaches a human is a decision, not a search.
The second is on the way out, and it is the one worth paying for: judging whether a price movement is a promotion at all. That call wants context. Is a campaign running. Is this product discontinued. Is the drop plausible as a percentage for this category. Did it arrive alongside a bulk edit of 400 other variants. A threshold cannot see any of that, which is exactly why a threshold gets this confidently wrong.
Two things stay human. The policy itself, because it is a margin decision rather than an operational one. And any adjustment above your cap, which should land in a queue with the number already worked out and a one-line recommendation attached.
Set it up this week
Export your current variant prices to a CSV today. That file is your first snapshot and it costs four minutes.
Then pull every partial refund from the last sixty days with no return attached and add them up. That number is what price adjustments already cost you, and in most stores nobody has ever looked at it.
Then write the five numbers down and publish them. The automation is straightforward once a policy exists. It is impossible while one does not.
Running a sale with a price adjustment story, the BFCM
morning that ate a day of support, the sweep that cost more
than the promo, the policy you have never written down? The
inbox is open: field-notes@dugong.live. We are
collecting case studies for the next issue.