September 22, 2026 · Disputed Team
"How do you actually define 'win rate'?"
It's a question we hear constantly from operations, finance, and leadership folks in chargebacks. You've probably asked some version of it yourself. You may also have asked it inside your own company, about your own numbers, and gotten two different answers from two different people.
The problem is there are multiple ways to calculate win rate, at multiple points in a dispute's life, and each one answers a different question. Report the wrong one to the wrong audience and you'll either flatter yourself or scare yourself for no reason.
Here are the four we think matter the most, and why.
Formula: Won / (Won + Lost). Grouped by the week you submitted, only counted once every case in that cohort is fully resolved (~9 weeks out).
Represent win rate asks: of the cases you actually fought and got a final answer on, how many did you win. The most basic scorecard: fight, then find out.
This is the cleanest measure of a dispute team's skill. It has no noise from what came in, what got skipped, or what an issuer decided to do later. If it drops, that's a direct signal to go look at how the cases themselves are being built.
The 9-week window isn't arbitrary. A representment goes to an issuing bank, gets reviewed by a back office, and works its way through a chain of approvals before you find out whether you won or lost. Look at RepWR before that chain has finished running and you're not looking at a finished number, you're looking at whichever cases happened to resolve fastest.
Formula: Won (week N−1) / Chargebacks received (week N−1, representment-eligible: excludes accepted). Viewed in week N.
Same idea as RepWR, but checked early. For last week's disputes: wins so far versus cases received, before most of them have even been resolved.
This is your earliest warning system. A drop in RWR flags something worth checking in last week's losses right now: a representment change, an issuer or scheme behavior change, incorrect representments going out, or a genuine quality slip.
It needs a baseline to mean anything, though. RWR always reads low, because most of that cohort hasn't resolved yet. A "dip" only makes sense compared to what a normal week-1 RWR looks like for your program, not compared to RepWR and not compared to zero. Know your own baseline before you react to the number.
When your team wins a dispute, the cardholder gets a second option: they can reopen it, sending it into a pre-arbitration process.
Fighting in pre-arbitration usually costs money on top of the disputed amount, so most merchants don't have the standing arrangement in place to do it, or don't take the fight. When nobody fights it, the case doesn't stay a win. It gets counted as a loss.
That distinction, a win that held versus a win that got reopened and dropped, is the whole reason the next metric exists.
Formula: Won / (Won + Lost + Pre-Arbitration). Same submission-week cohort, matured ~9 weeks.
Some wins get challenged again later and flip to a loss. Matured win rate is representment win rate adjusted to count those flipped cases too, so a win that didn't actually hold up doesn't get counted as one.
RepWR alone can flatter you if wins are reversing later. If MWR sits noticeably below RepWR, your wins aren't sticking, and it's worth investigating what happens after submission, not just what got submitted.
This one needs a baseline as well, specifically the normal gap between RepWR and MWR. Every program has some erosion between the two, which is expected. But what you want to watch out for is a quarter where that gap suddenly widens. You only know it's unusual by comparing it to what the gap has historically looked like for you.
Formula: Won / (Won + Lost + Pre-Arbitration + Accepted + Not Fought). Grouped by the week the dispute first came in, matured 13 weeks.
This one zooms out to every case that came in that week, including the ones you chose not to fight. A case you gave up on and a case you lost are both money gone, so both count here.
RepWR only judges what you chose to fight, which means a great RepWR can be hiding a lot of cases that got quietly accepted or never worked. MBR closes that gap. It's the real financial picture, including your triage decisions, not just your fighting skill.
Notice that the four are not all grouped the same way. RepWR and MWR are grouped by the week you submitted the representment: the work your team did. MBR and RWR are grouped by the week the dispute first arrived: the volume your business generated. Same underlying cases, two different questions, so don't mix the two groupings inside one report without saying which one you're using.
| Metric | Formula | Cohort | Matures at |
|---|---|---|---|
| RepWR | 310 won / (310 won + 150 lost) | submission week | ~9 weeks |
| RWR | 140 won / 620 received (early check) | received week, early | viewed weekly |
| MWR | 310 won / (310 + 150 lost + 40 pre-arb) | submission week | ~9 weeks |
| MBR | 310 won / (310 + 150 + 40 + 180 accepted + 120 not fought) | received week | 13 weeks |
RepWR reads 67%. MWR, once the pre-arbitrations that got reopened and dropped are counted, reads 62%. MBR, once every case that came in that week is counted, including the ones nobody fought, reads 39%.
Same underlying cases. Three very different pictures of the same quarter, and all three are correct answers to different questions.
Every formula above is calculating wins divided by dispute cases.
What they don't take into account are the dollar amounts they represent. You can run the same four formulas by dollar amount instead, wins recovered divided by dollars disputed, and you should, because a team that wins 70% of its cases by count but loses the handful of high-dollar ones is a very different story than the table above tells.
Keep the two separate, though. The rate answers how often you win. The dollar version answers how much winning is worth. Blending them into one number, like folding fees or dollar weighting into a "rate," just produces a figure nobody can trace back to either question.
Three questions to settle about your own reporting, in this order:
A win rate without those three answers attached is a number nobody can compare to anything, including your own number from last quarter.
And when a vendor quotes you a win rate, the same three questions work unchanged.