Refunds report
In the app: Reports → Refunds
Needs the 'View Reports' permission.
Shows how much money goes back to customers, who sends it and how long it takes: the refund rate, the split between the AI, your team and refunds made outside Squid, the reasons, and how refunds build up over the months after an order.
Where the figures come from
The report is built from the orders and refunds Squid reads from your connected Shopify stores. Every refund on those orders is counted, also refunds that were made in Shopify itself and never went through Squid. Squid's own refund records identify who issued a refund and why. Refunded totals always come from Shopify; the separate figure for changes to the AI's proposed amount compares Squid's own proposed and edited amounts.
- Shopify test orders are left out.
- Amounts are converted to your reporting currency at the exchange rate of the day of the order or refund itself.
- The line under the title shows how far the data is synced ("synced through …") and where your order history starts. Figures for days before that start are not complete.
What you can see instead of figures:
- "Refund reporting is not enabled for this organisation": order reading is switched off. "Open integration settings" opens Settings → Integrations.
- "No Shopify store is connected": connect a store first. A store with no saved Shopify access is not listed. The list does not check whether saved access still works.
- A notice that a store's orders are still loading, with "Check again".
- "No orders were placed in this period": the headline refund rate shows a dash because there are no orders to divide by. Return requests on older orders may still have a return-to-refund conversion rate. "Widen to the last 12 months" widens the period.
A teammate whose access is limited to certain stores only sees those stores.
Period, filters and view
- Period: "7d", "30d" and "90d", or a "Start date" and "End date". The default is 30 days. "Widen to the last 12 months" is also offered when there are no orders and that wider period is not already selected. Days start and end in the organization's "Report timezone" from Settings → Organization.
- "Stores": one or more stores to include or exclude.
- "Refunded by": keep only refunds issued by "AI", "Human agent" or "Outside the helpdesk". Rates still divide by all orders, so with "AI" selected the rate reads as the share of order value the AI refunded.
- "Exclude cancellation refunds": leaves cancelled orders out, both from the orders and from the refunds. This does not check whether an order shipped; it checks whether it was cancelled.
- "Value" / "Count": measure refunds by money or by number. It changes the tiles and charts that can be read both ways.
- "Overview" / "Cohorts": the report itself, or the cohort table described at the end.
- Refresh and "Export CSV".
Which refunds a period holds. The period selects orders by the day they were placed. The figures then count the refunds on those orders, as long as the refund was made within 60 days of the order. A refund made this week on an order from two months ago therefore belongs to that earlier period, not to this week. This keeps periods comparable, but it also means the most recent weeks look better than they will turn out: their orders have not had time to be refunded yet. The "Cohorts" view shows that build-up.
Headline figures
- "Refund rate (by value)": the refunded amount divided by the total charged, for the orders placed in the period.
- "Refund rate (by count)": orders that got at least one refund divided by orders placed in the period. An order refunded twice counts once.
- "AI share of refunds": refunds the AI issued as a share of all refunds, by value or by count.
- "Time to refund (median)": the number of days between an order and its refund, for the middle refund. The median is shown because a few very late refunds pull the average up; the average, the figure nine out of ten refunds stay under and the number of refunds are under it. Only the refunds the period counts are in it. Refunds made more than 60 days after their order, or recorded before their order, are excluded, so the figure cannot exceed 60 days.
- "Total refunded": the refunded amount, or the number of refunds, on the orders placed in the period. The tooltip calls it all refunds in the period; what is counted is the refunds on the period's orders, as explained above.
Under the refund rate, a line says whether cancellations are included. In "Value", it gives their contribution in percentage points to the rate with cancellations included. In "Count", it gives the share of orders cancelled, rather than their contribution to the refund rate.
Refund rate over time
The refund rate per day, week or month. Each point is the refund rate of the orders placed in that step. The grey bars are the sales of that step on their own scale, for context only: where the line sits against the bars means nothing.
Who issued the refunds
Splits refunds three ways:
- "AI": refunds the AI processed automatically.
- "Human agent": refunds that came from a request a teammate approved in the Refund Hub.
- "Outside the helpdesk": every other refund on your orders: made in Shopify admin, by a payment provider or by another app. This group is always there and is not an error.
Refunds made with the manual refund form in the inbox count as "Outside the helpdesk". A refund is recognised as Squid's by matching it to an automatic AI refund or to a Refund Hub request. The manual refund form is neither, so a refund a teammate makes with it lands in "Outside the helpdesk", and under "Why the money went back" it shows as "Unknown (external)". "Human agent" is therefore lower than what your team actually refunded when the form is in use.
A Shopify refund without a matching AI or Refund Hub record counts as "Outside the helpdesk". The page also warns about Squid refund records with no matching Shopify refund. Those records do not themselves add money to the report: if their Shopify refund has not synced yet, it is absent from the totals.
Under the chart, a sentence says on what share of Refund Hub requests a teammate changed the amount the AI proposed, and, where it is recorded, by how much in total. This sentence counts the requests created in the period, not the refunds on the period's orders.
Refund size
Refunded orders grouped by how large the refund was compared with the order, from under 10% of the order to the full amount and more. Only orders that got a refund are counted, never all orders. Refunded orders with a total of zero have no percentage and are mentioned separately.
Why the money went back
The ten refund reasons with the most refunded value or the most refunds; remaining nonzero reasons are added up as "Other". "Unknown (external)" stays separate even when it falls outside the top ten. The reason comes from Squid's record of the refund. A refund made outside Squid has no reason and is shown as "Unknown (external)" so that it is not left out of the picture. A refund from Squid that was recorded without a reason is shown as "Unspecified".
Return resolutions
What customers chose in the return portal, counted per returned item: "Refund", "Partial refund", "Exchange", "Store credit", "Keep the item", "Other", or "Unspecified" when no choice was recorded. It counts the return requests submitted in the period, whenever the order was placed. Of the filters, only the period and "Stores" apply to it. This panel is not part of the refund rate: an exchange or a store credit settles a return without money going back.
Return requests
- "Return request rate": return requests submitted through the return portal in the period, divided by orders placed in the period. A request is not a refund, and a request submitted in the period can be about an older order.
- "Return → refund conversion": of the requests that could be matched to a Shopify order, the share whose order has at least one refund. Any refund on that order counts, whenever it was made: also one made before the request, or for another reason than the return. The tooltip says the request "ended with money going back"; what is counted is that the order was refunded at some point. Requests that could not be matched are left out, not counted as "no refund". The line under the figure gives the numbers and the share of requests that could be matched.
- The amount under the tiles is every refund, of any date, on the orders that have a return request submitted in the period.
The request counts, conversion and refunded amount use only the period and "Stores" filters. "Refunded by" does not change them. However, "Return request rate" divides by the headline order count, so "Exclude cancellation refunds" reduces its denominator without removing any return requests from its numerator.
"Full returns analytics" opens Reports → Returns Analytics.
Refund rate by order cohort
The "Cohorts" view. Each row is a month in which orders were placed. Each column is the number of months since the order: M0 is the month of the order, M1 the month after, and so on. A cell shows the share of that month's order value that had been refunded by then, added up from the start. Reading along a row shows how long a month's orders keep being refunded.
- "Latest cohort month" and "Cohort months" choose the last row and how many months to show: 6, 12, 18 or 24, with 12 by default. "Stores" and "Exclude cancellation refunds" apply too. "Refunded by" has no effect on the cohort table, and "Value" / "Count" is disabled because it measures value only.
- An unfinished cell shows its measured cumulative rate once at least 20% of its window has passed. This window runs from the start of the order month through the end of the column month; it is not just the latest month.
- "Show projected cells" is on by default, but has no effect today: this report receives measured figures only, so it adds no estimates. The projection example in the legend does not mean projections are available.
- A dash means there is not enough data yet, or no orders in that month.
- Click a cell for its details: orders in the month, cumulative refunded value, the rate and how much of its window has passed.
With only a few months of order history connected, the table shows what exists so far and fills in as the months pass.
Export CSV
"Export CSV" downloads the refunds behind the report for the current period and filters, one row per refund: the refund and order, the store, the order date and refund date in your report timezone, the amounts in the shop's currency and in your reporting currency, who issued it, the reason and whether the order was cancelled.
One file holds at most 5,000 refunds. When there are more, the file holds the 5,000 most recent ones and stops there. Neither the file nor the page tells you it was cut off. A file with exactly 5,000 rows is therefore most likely incomplete: choose a shorter period and export again.
The button is disabled in the "Cohorts" view, with the reason as its tooltip: the export covers the period of the "Overview".