GMV is what buyers
spent. It is almost
never what you earned.

One number, counted
before anything gets
subtracted from it.

Gross merchandise value is the total value of goods sold through a platform, marketplace or channel over a period, measured at the price the buyer paid, before anything is taken out. That is the whole definition. The complexity comes entirely from what people do with it afterwards.

The word doing the work is gross. GMV is counted before refunds, before cancellations, before platform fees, before payment processing, before shipping, before discounts in most reporting conventions, and long before cost of goods. A marketplace reporting $1B in GMV has not earned $1B. It has processed $1B of other people's transactions and kept a fraction.

That fraction is the interesting number, and it has its own name.

Read that table twice, because it is the entire argument. One transaction produces a $100 headline, roughly $9 of platform revenue and about $80 to the person who actually made the thing. All three numbers are real. Only one of them is GMV, and it is the one least connected to anybody's bank balance.

It is worth being precise about who should use the word at all. GMV earns its place when the reporting entity does not own the inventory: a marketplace, an app store, a payments platform, a creator commerce channel. In those cases revenue alone would badly understate the scale of what is being operated, and GMV is the honest way to describe it. A single brand selling its own products has no such gap, which is why a single brand quoting GMV is usually making a presentation decision rather than an accounting one.

Calculating GMV is
easy. Agreeing what
counts is the work.

The formula is average order value multiplied by number of orders, over a defined period. Nobody argues about the arithmetic. Every argument is about which orders belong in the count, and those decisions change the answer by double digits.

The five decisions that change the number

1. Refunds and cancellations. Gross GMV counts them, net GMV does not. In apparel, where return rates run high, the two figures describe noticeably different businesses.
2. Shipping and tax. Most conventions exclude both, because neither is merchandise. Including them inflates the figure and makes period comparison unreliable once rates change.
3. Discounts. Count at the price the buyer actually paid. Counting at list price is a straightforward overstatement, and it is more common than it should be.
4. Failed and fraudulent orders. Exclude anything that never settled. An order that failed authorisation is not merchandise value, it is a log entry.
5. Timing. Order date or ship date, chosen once and kept. Switching between them across periods produces growth that is purely a definitional artefact.

Write those five choices down and put them next to the number, permanently. Across the businesses I have worked with, the single most common cause of a GMV figure nobody trusts is not error, it is that two teams computed it under different rules and neither documented which.

Average order value is doing half the work in that formula, which is why it deserves its own attention rather than being treated as an input. The AOV and conversion rate comparison covers which of the two is actually worth pushing on at different stages.

Commission GMV is the
slice you are actually
paid on.

Commission GMV is the portion of gross merchandise value that a commission is calculated against. It exists because almost no programme pays on the headline number, and the exclusions are where affiliates and partners lose money they assumed they had earned.

The exclusions vary by programme, which is exactly why the term needs its own name. Common ones: refunded and cancelled orders, orders from customers who were already the platform's, shipping and tax portions of the order value, orders placed with a competing coupon code, and orders outside an attribution window that is often shorter than anyone reads.

The four questions that define any commission GMV

1. Is it gross or net of refunds? Net is normal and fair. Gross with a clawback later is the same thing with worse cash flow.
2. Does it include shipping and tax? Usually not, and on low-value high-shipping categories that difference is large.
3. What is the attribution window, and is it last-click? A seven-day last-click window and a ninety-day first-touch window describe completely different businesses.
4. Are existing customers excluded? Most programmes exclude them, which is reasonable and routinely missed when people model their earnings.

There is a practical consequence for anyone building a business on top of a commission programme. Model on commission GMV, never on headline GMV, and apply a haircut for the exclusions before you decide the channel works. Across the partner and affiliate programmes I have looked at, the difference between the two figures has commonly run twenty to forty percent once refunds, existing customers and out-of-window orders come out. A channel that looks marginal on headline volume is usually unviable on qualified volume, and finding that out after you have hired against it is expensive.

Take rate tells you
what a platform is
genuinely worth.

Take rate is platform revenue divided by GMV. If a marketplace processes $1B and books $30M, its take rate is 3%. It is the single most useful number for comparing platform businesses, because GMV alone can be bought with subsidy and revenue alone hides the scale underneath it.

Take rate also moves, and the direction is diagnostic. A rising take rate means a platform is either adding services merchants want or extracting more from a captive base, and the difference between those two shows up in churn rather than in the ratio. A falling take rate means the opposite: growth is coming from volume the platform monetises poorly.

The clearest live example sits in the public filings. Across the four periods covered in the BigCommerce and Shopify financial study, BigCommerce's take rate fell from 1.22% of GMV to 0.96% while Shopify's rose to 3.10%. Both companies were growing GMV. Only one was converting it into a business at an improving rate, and the take rate said so well before the narrative did.

One caution on reading take rate across companies. It is only comparable when the businesses monetise the same way. A platform that earns purely on commission and a platform that earns on subscription plus payments plus lending will show different ratios for reasons that have nothing to do with health. Compare a company against its own history first, and against peers only when the revenue mix genuinely matches.

Three industries, three
slightly different uses
of one word.

The definition is stable. What changes is which number people put in front of it, and that is where most of the confusion in this topic originates.

The last row deserves a note. A single brand quoting GMV rather than revenue is almost always doing it because GMV is the larger number. There is no marketplace, no third-party sellers and no take rate. It is gross sales with a better haircut, and it is worth asking why the plainer word was avoided.

TikTok Shop is the context where this bites hardest right now, because creator commissions, platform fees and return rates all sit between the GMV headline and the money. The affiliate economics breakdown works through what actually survives that stack, and the answer is a good deal less than the dashboard implies.

The pattern across all four contexts is the same once you see it. The party reporting GMV is the party furthest from the money, and the number gets smaller at every step toward whoever actually shipped the product. That is not a criticism of the metric, it is a description of what a platform is. The mistake is only ever in reading a number from the top of that chain as though it belonged at the bottom.

Four ways this number
is used to make things
look better than they are.

GMV is not a dishonest metric. It is a useful one that is easy to deploy dishonestly, and the patterns repeat often enough to list.

1. Quoting GMV where revenue belongs. The most common and the least subtle. If a company reports GMV prominently and revenue in a footnote, the take rate is the reason.
2. Reporting gross GMV with a high return category. In apparel, where returns are structurally high, gross GMV and net GMV are materially different numbers and only one describes the business.
3. Counting GMV the platform subsidised. Volume bought with discounts is real volume and unprofitable revenue. It inflates the headline and depresses the take rate at the same time.
4. Annualising a peak. Taking a Black Friday week and multiplying. Obvious when stated plainly, surprisingly common in a deck.

A fifth pattern deserves separate mention because it is harder to spot: changing the definition mid-series without saying so. A company that quietly moves from net to gross GMV, or from ship date to order date, produces a growth curve that looks like performance and is arithmetic. This is why the definition belongs next to the number permanently rather than in a footnote that gets dropped in the next deck template.

None of that makes GMV useless. For a marketplace it is genuinely the right top-line measure, because the platform does not own the inventory and revenue alone would understate the size of what it operates. The failure is not using GMV. It is using GMV without the take rate beside it.

If you take one thing from this page, make it the pairing. GMV without take rate is a volume claim with no earnings attached, and a volume claim is not a business. Ask for both, every time, and the number stops being able to mislead you. For the brand-side version of the same discipline, the contribution margin breakdown covers what survives all the way to the bottom.

In a diligence process,
GMV is the first number
anyone takes apart.

If you are heading toward a raise or a sale, assume that GMV will be rebuilt from source data by someone who does this professionally and is not impressed by it. That reconstruction is routine, and the gap between the reported figure and the rebuilt one sets the tone for everything after.

What a buyer or investor actually does is straightforward. They pull order-level data, apply their own definitions, strip refunds, remove intercompany or test orders, and rebuild the series. Then they compare it with what the deck said. A small variance is normal and expected. A large one is not a rounding issue, it is a credibility event, and it repriced deals I have sat in on.

Building that reconciliation early is the cheapest diligence preparation available, and it is covered alongside the rest of the pack in the data room breakdown. If a sale is the actual destination rather than a distant possibility, the sell-side playbook covers what else gets rebuilt from source.

Questions people ask
about GMV and how it
relates to revenue.