A Chargeback Win Rate You Can Actually Compare

A comparable chargeback win rate needs a published denominator and an outcome policy. A percentage based only on contested, completed cases describes one selected group. It does not describe every filed case, every dollar disputed, or the work of analysts who inherited different kinds of cases.

Build a small outcome ledger before calculating anything. The ledger should retain accepted, partially won, pending, and unknown cases rather than squeezing them into a convenient win-or-loss field. Readers can then see both the result and the selection decisions that produced it.

Define outcomes without hiding ambiguity

Use the provider's recorded final state as the starting point. A fully won case, a partially won case, and a lost case should remain distinguishable. Record merchant acceptance as a merchant decision even if the provider's eventual state is mapped to a loss for one report. Pending cases have no final result yet.

Shopify documents partial outcomes as well as full wins and losses. Shopify's process reference. Do not infer a full win from a balance movement alone; match the event to the final case record.

Maintain a separate contested flag with a clear definition. For example, your organization might call a case contested when a substantive merchant response was submitted. Record the submission event supporting the flag. A note saying “planned to contest” does not establish that a response occurred.

Keep unknown outcomes in their own bucket until reconciled through the relevant case administration workflow. If an extraction omits older cases, the missing population must be disclosed before publishing performance.

Publish two views with different purposes

The all-filed view begins with a purchase or filing cohort and accounts for every included case. It is useful for showing how the total case population developed. The contested-completed view describes outcomes among cases selected for contesting that have reached a known result.

Neither view should silently absorb pending cases. In the all-filed view, show fully won so far as a share of all filed cases alongside the full outcome distribution. Label it as an observed snapshot if the cohort is not mature. In the contested-completed view, exclude pending cases from the result denominator but disclose their number.

Do not call a recovery-value measure a case win rate. Winning one USD 20 case and losing one USD 2,000 case produces a different business picture from the reverse, although the case count result can be identical. A separate amount-based metric may be useful, but its treatment of partial recoveries and fees needs its own specification.

Read a hypothetical ledger

A hypothetical merchant reviews 100 filed cases from one cohort:

Thirty cases are fully won so far among all 100 filed cases. The all-filed snapshot is therefore 30% fully won, with substantial other states still visible. That is not a final cohort win rate while twenty cases remain unresolved.

Among the sixty contested cases with completed outcomes, the strict full-win share is 30/60, or 50%. If the merchant separately reports any positive case outcome, that share is 40/60, or about 66.7%, because it includes partial wins. Calling both numbers “win rate” would make comparisons impossible.

The twenty accepted cases matter. Excluding them may be appropriate for the contested-completed view, but it conceals the selection policy unless acceptance volume and reasons are also shown. A team that contests only unusually strong cases can post a higher contested percentage while resolving less of the original case population in its favor.

Make comparisons survive a change in case mix

Before comparing two months, examine reason categories, disputed amounts, inherited cases, acceptance policy, and outcome maturity. A new rule that routes weak cases away from contesting can raise the reported percentage without improving the underlying customer experience or analyst decision quality.

Use a stable report header: “Strict full wins among contested cases with known final outcomes; partial wins separate; cohort defined by filing date; observed through [date].” Include the numerator, denominator, pending count, accepted count, and definition version.

When policy changes, show results under both the old and new definitions for a transition period if the data supports reconstruction. If it does not, mark a break in the series. Connecting incompatible percentages with an uninterrupted trend line gives readers an appearance of continuity that the records cannot support.

The operational decision should follow the distribution, not the largest percentage. A rise in pending cases might require a status reconciliation. More merchant acceptances may call for a review of commercial failures. A shift toward partial outcomes may warrant a separate value analysis. Publishing the denominator gives those questions a fair starting point.

Keep operational dispute review separate from your independently calculated performance metrics.

Review your disputes

Related reading in this collection:

• How to Calculate Your Shopify Chargeback Rate: Count and Value Explained
• Backtest Shopify Risk Ratings Against Later Disputes
• Find Dispute Concentration Before Blaming Your Entire Store