CS economics report
In the app: Reports → CS economics
Needs the 'View Reports' permission.
Shows what customer service costs per order and as a share of your margin, where that money goes, and how well the AI and your team resolve conversations: repeat contact, first-contact resolution and how often an offer saves a refund. The cost inputs are set under Settings → Costs.
When the report is available
The tab is only in the Reports strip for organizations that have Squid Agent switched on. For other organizations the page shows "Available with Squid Agent" with a link back to Reports, and no figures.
Period and who sees what
"Last 7 days", "Last 30 days" and "Last 90 days" choose the period. The default is the last 30 days. The period ends today and is cut in whole UTC days, not in your organization's timezone. There are no date fields and no store filter on this page.
Two groups of figures need different permissions:
- Repeat contact, first-contact resolution, "AI vs people" and offers are shown to everyone with the 'View Reports' permission.
- The cost tiles, "Where the money goes" and "Cost per order by store" hold wages and money. They also need the 'View customer service costs' permission. Without it, a notice explains that cost lines are shown to Owners, Admins and roles with that permission.
- "Examples of repeat contact" also needs the permission to view conversations.
A teammate whose access is limited to certain stores only sees those stores. Shared costs, such as staff and tools, are then counted for the share of the organization's conversations that those stores had.
A "Finish cost setup" row appears while inputs are missing: "Standard COGS % is not set, so margin impact can't be shown", "Worked hours are not uploaded, so labour isn't included", or "Labour is only partly included" when some weeks of the period have no hours. Each links to Settings → Costs.
Cost tiles
Needs the 'View customer service costs' permission.
- "CS cost per order": the "CS cost" total divided by the orders placed in the period. Orders are the Shopify orders placed in the period in the stores you can see; test orders are left out and cancelled orders are counted. Under the figure are the number of orders and, when worked hours are uploaded, the staff cost per order on its own.
- "CS cost": the total of the cost lines below for the period, in your reporting currency.
- "Share of gross margin": CS cost divided by gross margin, where gross margin is the revenue of the orders placed in the period minus the cost of goods at your standard COGS %. The percentage in force when each order was placed is used, including any store override. It reads "Needs your standard COGS %" when an order in the period has no percentage to use.
- "Worked hours": the hours uploaded for the period for the whole organization, with the number of weeks that have none. Hours are not split by store, so a teammate limited to certain stores sees the same hours.
When no worked hours can be priced, the totals leave staff cost out and say "excl. labour". Uploading hours is not enough: at least some need an hourly rate and, where needed, an exchange rate. "Labour partly included" means a priced staff cost exists but some weeks have no uploaded hours. Squid never estimates staff cost from handling time: no hours, no staff cost.
Where the money goes
Needs the 'View customer service costs' permission.
The cost lines that add up to "CS cost":
- "Refunds": every refund on your Shopify orders that was issued in the period, counted on the day the refund was made, whichever period the order was placed in. Refunds made in Shopify itself, outside Squid, are included. Refunds on test orders are not. A store whose orders are not synced has no refund figure; it is left out and named, and the line is marked "Partial".
- "Staff": worked hours times the hourly cost. The Owner sees hours priced at a person's own rate where one is set, and otherwise at the rate of their team. Everyone else sees every hour priced at the team rate, because rates per person are shown to the Owner only; the two views can therefore show a different staff cost. Hours without a rate are left out and counted separately.
- "Vouchers and discounts": the face value of the vouchers in offers customers accepted in the period, multiplied by your voucher redemption percentage when one is set under Settings → Costs. Without that percentage the full face value counts.
- "AI and tools": each monthly tool cost from Settings → Costs, including your Squid plan, spread evenly over the days of the month.
- "Replacements and reships": reshipment tickets filed through Squid in the period, priced at what the customer paid for the reshipped items times the store's standard COGS %. This line is always marked "Partial": a reship arranged outside Squid is not in it, and a reship without a known item value or COGS % is left out.
A line without input shows why it is empty, such as "Staff cost: not set", and is never shown as zero. Amounts in another currency are converted at the exchange rate of their own day. "What is left out" lists what the total does not include. Repeat-contact cost, when shown, is part of the staff and AI lines and is not added on top.
Cost per order by store
Needs the 'View customer service costs' permission.
One row per store: "Store", "Orders" (orders placed in the period), "Cost per order" and "Standard COGS" (the percentage in force for that store at the end of the period). Refunds, vouchers and reships are counted for the store they belong to. Shared costs, staff and tools, are divided over the stores by each store's share of all the conversations the organization created in the period. A store without synced orders shows a dash or "Sync off" instead of a cost per order.
Under this table the page shows "Why refund cost moved": what the period's refunds consist of and how that changed against the period before. It is explained under Reports → Investigations, which can also split CS cost by contact reason, store or country.
AI vs people
Compares conversations that were closed for the first time in the period and that only the AI answered ("AI alone") with those only people answered ("People"). Who answered is judged by who wrote the customer-facing replies; campaign sends, internal notes, saved FAQ answers and templated photo-review verdict emails do not count. An AI draft that a teammate approved and sent counts as an AI reply here, so a conversation answered only with approved drafts is under "AI alone". The "AI" tab of Reports counts such a reply for the teammate instead. Conversations that both the AI and a person answered are in neither column of the rates, and conversations nobody answered are left out.
- "Repeat contact within 7 days" and "First-contact resolution": as defined in the next two sections.
- "CSAT (4 or 5 stars)": the share of the ratings given on those conversations that were 4 or 5 stars. The rating can have been given after the period.
- "Refunds per 100 conversations": refunds made through Squid on those conversations, per 100 of those conversations. A Refund Hub request counts when it was approved or marked as refunded by hand; a refund from the manual refund form counts unless it failed, was rejected or was cancelled. Refunds made directly in Shopify are not counted here.
- "Cost per conversation": only shown when some worked hours can be priced for the period and you may see costs. For "AI alone" it is the "AI and tools" line divided by the conversations only the AI answered. For "People" it is the "Staff" line divided by the conversations a person answered, including the ones the AI answered first.
The AI and people handle different kinds of questions, so a difference between the two columns does not prove that who answered caused it.
Repeat contact
"Repeat contact within 7 days" is the share of closed conversations after which the same customer contacted the same store again about the same thing within the follow-up time. The follow-up time is 7 days unless "Count a new conversation as a repeat contact within (days)" was changed under Settings → Costs.
- "About the same thing" means the two conversations share an order, or else have the same first contact reason. When neither is known for one of them, the same customer coming back is enough.
- The later conversation must be with the same store and must have been created after the first one was closed. Deleted, merged and example tickets, and campaign tickets the customer never answered, do not count as a repeat.
- A conversation is counted in the period of its first recorded close, also when it was reopened later. However, repeat matching uses its current close time: an open conversation that was reopened has no close time to match against, and a re-closed conversation is matched after its latest close. Repeat contact after an earlier close can therefore be missed. Older conversations without recorded close history can be measured from their current close time only while closed.
- Only closes whose follow-up time has fully passed are counted, because "no repeat yet" is not "no repeat". The card says how many recent closes are not measured yet, and until which day closes are measured. With none measured it reads "Not measured yet".
- The card gives the rate overall and split by who answered: "AI only", "People", and "AI then a person" for conversations both answered.
- For a period longer than the report can measure, the card asks you to pick a shorter one.
"Examples of repeat contact" lists up to 10 pairs of a first conversation and the one that followed, with links to the tickets.
First-contact resolution
The share of closed conversations that were resolved the first time: the conversation was not reopened and the customer did not make repeat contact within the follow-up time. It combines the repeat-contact check described above with recorded reopens; that repeat-contact check can miss contact after an earlier close.
A reopen counts against it when it happened within the follow-up time after the first close. The card shows the rate for "AI only" and for "People"; it does not show one overall figure.
It is only measured from the day Squid started recording reopens; the card names that day as "Measured from …". Like repeat contact, it only counts closes whose follow-up time has passed, and shows "… resolved first time" with the number not measured yet.
Offers instead of refunds
How offers perform, such as a partial refund or a voucher offered instead of a full refund.
- "Offer take rate": accepted offers divided by presented offers, shown as a percentage with "… of … presented". An offer is counted in the period in which it was presented, also when it was accepted later. One offer on several items counts once. It reads "No offers presented in this period" when there were none.
- Current offers count as presented. Squid keeps the current offer for each unit of an item and each reason. Units on the same order with the same offer are counted together; different offers on one order can count separately. An accepted unit keeps its offer, and some final offers are retained when the next step is presented. When a pending offer is replaced by the next step, its earlier presentation is no longer counted and the replacement is counted on its new presentation date. When a customer accepts an offer, open offers for the same items made for another reason are removed. The take rate measures the offers still retained this way, rather than every offer ever shown.
- Two tables under the tiles split the take rate. The first has one row per reason the offer was made for, under the heading "Contact reason", such as "Damaged item"; it lists at most 20 reasons, the most presented first. The second has one row per step, under "Offer step": "Offer 1", "Offer 2" and so on. A row counts retained offers at that step, so "Offer 1" does not include earlier offers replaced by "Offer 2", and the rows are not a funnel. Each row shows "Take rate" and "Accepted" as the number accepted of the number presented.
- "Offer save rate": over the accepted offers, how much of the items' value was kept: 1 minus the refund and voucher value given, divided by what the customer paid for the items the offer was about. Vouchers count at their full face value here, whatever redemption percentage is set. A higher rate means the offer cost less than a full refund would have.
- Accepted offers with a missing item value or an amount that cannot be converted count in the take rate only. The card gives their number, although it describes all of them as missing item values. The date shown is the earliest presentation among the accepted offers with usable values in this period; it is not necessarily the day recording began.