"We turned over £2.6m on our own site last year. Our accountant says chargebacks are not material because we only lost £6,000. Then our payment provider put a rolling reserve on us in January and nobody could explain why. What should we actually be putting aside?"
Accrue for the disputes you have not been told about yet, not the ones already sitting in your inbox. At a balance sheet date that is about six weeks of chargeback cost: refunded sales, dispute fees, and stock you never get back. On the illustrative £2.64m card book below, that provision is £1,157.
The founder above is not unusual. Chargebacks are the one cost in ecommerce that arrives months after the sale, gets booked against the wrong month, and never shows up in a Shopify or TikTok dashboard as a number you can plan against.
Every figure below is a synthetic example, not a client's books. The platform and scheme rules are checkable against the sources named in the text.
What is the short answer?
Accrue the expected cost of disputes that will be raised against sales you have already made. Three buckets, dated at the balance sheet date:
One: refunded sales. The orders you expect to lose, valued at what the customer paid.
Two: dispute fees. Charged on every dispute, win or lose.
Three: stock. The landed cost of goods that never come back.
The size of the accrual comes from your own lag curve, not from an industry average. On the curve below, about six weeks of chargeback volume is still in the post at any year end, which is 1.5 months of your annual run rate.
What you do not do is wait for the letter. That is not accounting, it is reacting.
What is a chargeback, and why is it not a return?
A chargeback starts with the cardholder, not with you. They call their bank, the bank creates a formal dispute on the card network, and the payment is reversed before anyone has heard your side. Stripe's documentation describes the mechanic plainly: the issuer creates the dispute, the network reverses the payment, and the processor debits your balance for the payment amount plus the dispute fee. The money is held for the whole dispute.
A return is different in every way that matters. The customer asks you, you agree a refund, and you know the number on the day you process it. A chargeback is a deduction you did not authorise, on a sale you have already counted, with an outcome you do not control.
Three dates drive the accounting:
The dispute window. Card networks typically allow a cardholder to raise a dispute within 120 days of the original payment, and the rules allow longer in some situations. That is why a December sale can hit your March bank statement.
The response window. Once a chargeback is raised you usually have 7 to 21 days to respond, depending on the network.
The decision. The issuer usually takes 60 to 75 days to decide once you have submitted evidence.
That timing is the whole reason a provision exists. If disputes resolved in the same month they were raised, you would not need one.
Why does a chargeback cost more than the refund it replaces?
Because the fee is charged on the count, not on the outcome. Stripe's UK pricing, checked in October 2026, lists £20 for each dispute received and £20 for each dispute you counter manually. You get the countered fee back if you win. You do not get it back if you lose, and the received fee is not returned either way. In rare cases network fees are added on top.
Put that against an average order of £110:
| Outcome | What you lose | Cash cost |
|---|---|---|
| Refunded before it becomes a dispute | Legitimate sale unwinds | £110 |
| Dispute won | Nothing, two to three months later | £20 |
| Dispute lost | Sale refunded plus fees | £150 |
The last row is the one founders miss. You refund £110 and you pay £40 in fees, so a lost dispute starts at £150. If the goods never come back you add £38.50 of landed cost, which takes the average lost dispute to about £163 once you allow for the third of those orders where the stock is gone.
There is a second cost that never appears in the P&L. Card schemes monitor dispute rates, and a high rate puts your account on report. Visa's acquirer monitoring programme, as summarised in Stripe's operator documentation, starts at a 0.5% dispute ratio with five reports in a month, and escalates at 1.5% and 1,500 reports outside CEMEA. Mastercard and Visa both review these programmes, and the trend in threshold changes has been downwards.
Once you are on report, the practical consequence is a rolling reserve. The processor can hold back a slice of every payout for months. That is cash, not a cost, and it does not show in your profit. It will still stop you buying stock in September.
What should the chargeback accrual actually cover?
Three liabilities, not one. Mixing them into a single "bad debt provision" line is how brands end up with a number nobody can explain at the year end meeting.
| Bucket | What sits in it | Illustrative year end |
|---|---|---|
| Refund liability | Sales you expect to reverse, at the price paid | £660 |
| Fee accrual | Received and countered fees on disputes not yet raised | £420 |
| Stock provision | Landed cost of goods never recovered | £77 |
| Total | Balance sheet liability | £1,157 |
Two accounting points sit underneath that table.
A chargeback that unwinds a sale is a revenue adjustment, not an expense. The customer has paid, you have recognised the sale, and the dispute reverses the consideration. Under FRS 102 the refund element reduces revenue and sits as a liability until it is settled. Booking the whole thing to administrative expenses overstates revenue and overstates costs, which flatters your gross margin and tells you nothing useful about the sales that actually stuck.
The fee and the stock are costs. Neither is a reversal of the sale. The fee is a payment services cost. The stock write-off is an inventory impairment, at landed cost, not at selling price.
FRS 102 Section 21 is the test for the provision: a present obligation from a past event, a probable outflow, and a reliable estimate. A dispute not yet raised on a sale already made meets all three if your own history says the disputes arrive.
How do you work out your own chargeback rate?
Count, not value. The schemes measure disputes against the number of transactions, so a brand with a high average order value reaches the threshold on far fewer disputes. On £2.64m of card sales, a 0.5% rate is 120 disputes a year at an average order of £110, and about 38 at £350. Same revenue, same rate, one third of the disputes.
Then build the lag curve. Take the last twelve months of disputes, match each one to the date of the order, and bucket by days. A curve like this is typical for apparel:
| Days after the order | Share of chargebacks raised |
|---|---|
| 0 to 30 | 40% |
| 31 to 60 | 30% |
| 61 to 90 | 20% |
| 91 to 120 | 10% |
Now apply it at 31 December. Sales made in December are on average a fortnight old, so around 20% of their chargebacks have been raised and 80% have not. November sales are 45 days old, so 55% have been raised. October is 75 days old, September is 105 days old, and the sums are in the table below.
| Month of sale | Share not yet raised at 31 December |
|---|---|
| December | 80% |
| November | 45% |
| October | 20% |
| September | 5% |
| Equivalent months of chargebacks | 1.5 |
That is where the six weeks comes from. It is not a rule of thumb pulled from the air. It is your own curve, added up across the last four months of sales.
What does the year end provision look like?
Take an illustrative sports brand with 2,000 own site card orders a month at £110, which is £220,000 a month and £2.64m a year. Chargebacks run at 0.5% of orders, so 10 a month and 120 a year. Four in ten are lost. One in three lost orders never comes back, and landed cost is 35% of the selling price.
| Cost line | Count | Per item | Year |
|---|---|---|---|
| Lost sales refunded | 48 | £110.00 | £5,280 |
| Received fees on won disputes | 72 | £20.00 | £1,440 |
| Received and countered fees on lost disputes | 48 | £40.00 | £1,920 |
| Stock never recovered | 16 | £38.50 | £616 |
| Total cash cost | £9,256 |
That is £771 a month in cash, or 0.35% of card sales. Modest, which is exactly why the founder in the opening thinks there is nothing to do.
The provision is not the annual cost. It is the slice of it that belongs to sales already made and disputes not yet raised. At the year end that is 15 disputes, being 1.5 months of the 10 a month run rate:
| Provision at 31 December | Working | Amount |
|---|---|---|
| Refund liability | 6 lost disputes at £110 | £660 |
| Fee accrual | 9 won at £20 plus 6 lost at £40 | £420 |
| Stock provision | 2 orders at £38.50 | £77 |
| Total | 1.5 months of annual cost | £1,157 |
If you have been booking chargebacks only when they arrive, your December accounts are short by about £1,157, your September accounts were flattered by disputes that belonged to August, and nobody can see the pattern because every month looks slightly wrong in a different direction.
There is a growth trap in the same maths. If card sales grow 10% a month, December's sales are nearly three times January's. A provision built on the annual average understates the tail by a wide margin. Build it on the latest quarter's run rate, then reconcile.
Where do sports brands get hit hardest?
Sports apparel and footwear have their own pattern. Five places it shows up:
Drop delivery promises. A pre order that slips by three weeks turns into a wave of goods not received disputes, and those are the ones you lose, because the tracking shows nothing. One drop of 600 orders at £165 is £99,000 of revenue. An 8% dispute rate on that cohort is 48 disputes, 34 of them lost, and about £7,830 of cost. That is ten normal months of chargeback expense from one delivery date.
Sizing and fit. A customer who cannot work out whether to order the small or the medium disputes the transaction instead of returning it. The returns and sizing leak is the same root cause with a different ending, and it is covered in our post on modelling returns and sizing leakage.
Custom and teamwear orders. Personalised kit is exempt from the 14 day cancellation right under the Consumer Contracts Regulations 2013, regulation 28. That does not stop a chargeback for damage, delay, or a kit that does not match the artwork. It only removes the change of mind route, which is the cheap one.
Card testing on launch nights. Limited drops attract bots. A rush of small authorisations is a fraud signal, and the fraud reports that follow count towards the scheme monitoring alongside your genuine disputes.
Subscriptions. Sports nutrition and refill plans generate credit not processed disputes when a cancellation is not honoured or a refund is promised and not paid. Those are entirely within your control, and they should never reach a card network.
Marketplace sales behave differently. On TikTok Shop and Amazon the customer's dispute is handled on the platform, which then deducts from your settlement under its own after sales rules. You still accrue for the deduction, but it is not a card scheme chargeback in your name. If most of your volume sits there, read the TikTok Shop seller page before you build the model.
How should chargebacks land in the P&L and the VAT return?
All the figures above are cash, because that is what leaves the bank. Your P&L sees less of it if you are VAT registered.
A £110 refund is £91.67 of net revenue and £18.33 of output VAT. If you issue a credit note and repay the customer, the output VAT comes off box 1 and the sale comes out of box 6. The £20 fee carries VAT you can reclaim, so the net cost is £16.67. The stock write off is at landed cost, with the import VAT already recovered.
Run the annual numbers again on that basis:
| Cost line | Cash | P&L charge |
|---|---|---|
| Refunded sales | £5,280 | £4,400 |
| Dispute fees | £3,360 | £2,800 |
| Stock write off | £616 | £616 |
| Total | £9,256 | £7,816 |
Where the money is taken back and you never get paid, the route is different. That is a bad debt, and VAT bad debt relief has its own conditions. Under HMRC's Notice 700/18 the debt must have been written off and remained unpaid for six months from the later of the due date and the date of supply, and the claim must be made within four years and six months. Do not assume a chargeback automatically gives you the VAT back in the quarter it lands.
What should you fix before the rate gets worse?
Most chargeback cost is avoidable, and the fixes are cheaper than the provision.
Refund before it becomes a dispute. A complaint you resolve on the day costs you the sale. A dispute costs you the sale, £20 to £40 in fees, and a mark on the count. Stripe's own guidance on early fraud warnings makes the same point from the other side: refunding a charge close to your dispute fee is often the cheaper decision. Set a rule for your team and do not let pride decide it.
Stop promising dates you cannot hold. The goods not received dispute is the one you cannot win. If a drop is going to slip, tell customers before they find out from a tracking page that has not moved. A delay email costs nothing. A wave of disputes costs more than the margin on the drop.
Build the evidence pack before you need it. Tracking and proof of delivery, the product page as it looked on the day, the size guide version, the terms, and the email chain. A dispute response has a 7 to 21 day deadline, and a pack assembled in a panic is a lost dispute.
Fix fit at the source. Size guides, model measurements, real customer photos, and an exchange offer that makes keeping the item less attractive than swapping it. The detail is in our guide to Shopify returns accounting.
Reconcile monthly. Match every dispute to its order, month by month. Without that, the rate is a guess, and a guess is what you will accrue. The method is the same one we use for returns and expiry accruals.
Watch the count ratio, not just the cost. A brand can absorb £9,256 of chargebacks and still be one bad drop away from a reserve. If your monthly disputes are climbing while your order count is flat, you have a product or delivery problem, not an accounting problem.
FAQs
How much should a sports brand accrue for chargebacks?
About six weeks of your annual chargeback cost at the balance sheet date, or 1.5 months on the lag curve used here. It covers refunded sales, dispute fees, and stock you never recover. On an illustrative £2.64m card book with a 0.5% dispute rate, that is £1,157.
Is a chargeback the same as a refund in the accounts?
No. A refund is a known amount you choose to pay, and the sale unwinds in the month you process it. A chargeback is a disputed payment debited with a fee before anyone decides the outcome. The refunded sale reduces revenue, the fee is a cost, and goods that never come back are a stock write off.
Do chargebacks count towards Visa and Mastercard monitoring?
Yes, and they are counted against the number of transactions rather than their value. Visa's acquirer monitoring starts at a 0.5% ratio with five reports in a month. A complaint you resolve before it becomes a dispute does not enter that count, which is one more reason to refund early.
The Close
Chargeback accounting is not difficult. It is just early. You are putting a number in the accounts for something that has not happened yet, using your own history as the evidence.
Get the lag curve right and the provision becomes a line you can defend at the year end, a rate you can watch month by month, and a cost you can price into the product. Skip it and you will keep discovering, three months late, that a good quarter was not as good as it looked.
If your own site card book is growing and you cannot say what a lost dispute costs you, book a call.