Your support queue is a margin report that nobody reads as one. The tags are already there.
Eight weeks before the CL-7732 Trail Boot showed up in a margin report, your support team had already written the finding. Seventy-four percent of return-related tickets on that SKU said some version of "runs small." The tag existed. The pattern was visible. Nobody was reading support as a financial system, so it sat in Gorgias until the ad spend made it a finance problem.
Support is the earliest signal you have. It is the only place a customer tells you in words what went wrong, and it happens weeks before the same information reaches finance as a return, a refund, or a margin variance.
Follow one defect through your stack:
| Week | Where it shows up | What it looks like there |
|---|---|---|
| 0 | Gorgias | Tickets: "runs small," "sizing is off" |
| 1 to 2 | Returns | Return requests on one SKU rising |
| 3 to 5 | Refunds | Refund volume up, reason codes unread |
| 6 to 8 | Margin reports | SKU contribution margin down, cause unclear |
| 8+ | Ad account | Still spending, ROAS still looks fine |
By the time finance can see it, you have paid for eight weeks of returns, eight weeks of return shipping, eight weeks of restocking labor, and eight weeks of ad spend acquiring customers for a product that disappoints them. All of it was preventable at week zero, in plain language, in a system you already own.
The reason nobody catches it is organizational. Support is measured on response time and CSAT. Nobody asks the support team what a ticket costs, so nobody counts.
Raw ticket volume tracks sales volume and tells you nothing. Normalize it. A SKU generating 4 tickets per hundred orders against a brand average of 1.6 has a problem, regardless of how well it sells.
This is the single most useful number in this guide, and almost nobody computes it. It is a straightforward join between ticket tags and order line items.
Tickets containing "arrived damaged," "box was crushed," "broken in transit" are recoverable money, not just service events. Every one is a potential carrier claim. Crestline had 287 such tickets in thirty days and had filed claims on 109 of the 396 eligible, about 38 percent. The remaining 62 percent were absorbed as brand-funded refunds.
At roughly $46 average claim value, that is $13,200 a month written off as a cost of doing business, when it was a filing problem. The reason is mundane: each carrier claim takes four to six minutes in a portal, and a support agent under a response-time SLA will always choose the refund.
"Runs small," "sizing chart is wrong," "ordered my usual size." When this concentrates on one SKU it is a product data problem with an unusually cheap fix, updating a sizing guide, and an unusually expensive consequence if left alone, because you keep advertising into it.
The tell is concentration. Sizing complaints spread evenly across a catalog are normal. Seventy-four percent of one SKU's return reasons is a defect.
A customer with $2,940 lifetime value and a 4 percent return rate contacting you three times in a month is a churn event forming. The cost of losing them is their remaining lifetime value, not the cost of the resolution, and support systems are not built to surface that distinction. An agent sees a ticket. They do not see a $1,800 asset at risk.
Crestline's thirty days of support data, read as a margin report:
| Signal | Volume | Value | Type |
|---|---|---|---|
| Unfiled carrier claims | 287 tickets | $13,200 | Recoverable now |
| CL-7732 sizing complaints | 74% of returns | $7,200 | Ad spend on a defect |
| High-LTV repeat contacts | 203 customers | $24,100 | LTV at risk |
| Total, visible in support first | $44,500 |
Crestline Co. is a simulated DTC brand used for demonstration. These figures are illustrative and are not a client result.
Every one of those was legible in the support queue before it was legible anywhere else. None required new data collection. The tags already existed.
The reframe. A ticket is not just a service interaction, it is a customer telling you something is costing you money. Your support queue is the highest-resolution operational data in the business and it is almost universally read only for response time.
None of this works without tagging discipline, and most tagging schemes fail by being too elaborate. Keep it small enough that a busy agent applies it correctly during a shift.
The barrier is that this requires four systems at once. Ticket tags live in your helpdesk, order and SKU data in your commerce platform, spend in your ad account, and margin in finance. The signal only appears when all four are read together, which is why it usually is not. Postlytix correlates them continuously, surfaces the patterns with dollar values attached, and prepares the carrier claims for you to approve.
This is one of eight places the numbers come apart. Why your e-commerce revenue numbers never match explains why fragmentation happens across the whole stack, and links to the rest.