Drop, disregard, refer
Sort every alert into one of three buckets before you touch anything: cut the basket line that caused it, ignore it and carry on, or escalate it to a written query. The bucket is decided by what the alert points at rather than by how alarming the wording is, and our reading record covers 85 selected entries with 23 flags raised across them.
1. Read which line the alert names, because a restriction attached to one seller is not a restriction on your account. 2. Check whether the named line is still in the basket and whether it was there when the alert appeared. 3. Sort into the bucket that matches the named line, then act once rather than re-reading the alert five times.
Enter the scale figure, the carton dimensions and the divisor into one note before you read the alert a second time, because a weight warning only ever moves the freight line. On our own £120 parcel the chargeable weight was 3.8 kg against a £20.00 freight charge, which is £5.26 per kg, so half a kilo of rounding is worth about £2.63 there; the £6.13 service fee, set at 6% of goods plus domestic freight, and the £3.50 packaging charge did not move with it (Recorded). 1. Compare the two weights: if the gap falls inside one 0.5 kg rounding step, ignore the alert and keep the note. 2. If the gap crosses a boundary at 1, 2, 3, 5, 8 or 10 kg, escalate in writing with the scale photograph attached. 3. Cut the line when the quoted figure came from the listing description and no scale photograph exists, since nothing you send can correct a number nobody measured (Recorded).
What triggers it
Four things generate almost every alert we have seen. A listing whose category changed after you added it; a basket value above the point where the platform wants a second look; an image the automatic check could not read; and a variant that sold out between the moment you added it and the moment you settled. Three of those four are about the listing and only one is about you.
1. Note the exact time the alert appeared. 2. Note the exact time you last changed the basket. 3. If the gap is under 60 seconds, the alert is almost certainly reacting to your own edit rather than to anything on the account.
Divide the frame into a 5 by 5 grid before you judge the flagged area, because each cell is 4% of the picture and three cells make 12%. 1. Count the cells the flagged region touches, then compare the share: 6 of the 23 flags on our 85 curated entries sat under that 12% figure (Recorded). 2. Check where those cells fall, since 5 of those 6 flags proved to be lighting artefacts on the border of the photograph rather than marks on the item (Recorded). 3. If the band sits on an outer edge, request a re-shoot that names the angle, the distance in centimetres and the light source; our 6 recorded re-shoot cases came back inside 6 hours when the request was explicit and in about 2 days when it was vague (Recorded). 4. Write the reply date beside the image, so a second flag on the same line is answered from the file rather than from memory.
Run the volumetric figure three times before accepting a size warning, because the divisor decides the answer: chargeable weight is the greater of the scale reading and length x width x height in centimetres divided by 5000, 6000 or 8000, rounded up to 0.5 kg, and our tier table holds 6 lines across 9 bands from 0.5 kg to 10 kg (Recorded). 1. Enter all three results and keep the largest. 2. Compare it with the prices we recorded: £8.40 for the first weight, £13.20 at 2 kg and £31.80 at 5.1 kg (Recorded). 3. Check the plateaus, since 8 kg and 8.5 kg both came in at £43.60 and 10 kg and 11.5 kg both came in at £54.60, so a warning inside either one changes nothing. 4. Mark a quote without a divisor label as Not verified: 3 of our 12 rate reviews left it unstated, and 4 of the 12 raised the first weight on an unchanged line by up to £1.60 (Recorded).
Azure perimeter 12%
The false alarms share a signature if you look at where the flagged area sits in the image. On our record, 6 of the 23 flags were raised on a region covering under 12% of the frame, and 5 of those 6 turned out to be a lighting artefact on the border of the photograph rather than a defect on the item. A flag on a corner or an edge is weaker evidence than a flag in the middle of a seam.
1. Open the image at full size rather than in the list preview. 2. Measure roughly what share of the frame the flagged area covers. 3. If it is under about an eighth of the frame and sits at the border, treat the alert as a photographic artefact and ask for a retake before you decide anything.
Compare the two inner positions with the label reading before you answer the alert, and treat the comparison as a measurement job rather than a matter of taste. 1. Photograph the stitching run and the hardware fitting from one distance and one angle, so both images share a scale in millimetres. 2. Write both readings beside the label size and the measurement the listing carried, then check the gap against one whole size step. 3. Escalate when either inner point disagrees by more than that step, naming one position rather than the full set. Of the 23 flags on our 85 curated entries, 11 sat on the label, 6 on edges, 4 on hardware and 2 on stitching, and the label and hardware groups were the only ones that changed an outcome; our 6 recorded re-shoots show the cost of a badly framed question, with explicit requests answered inside 6 hours and vague ones in about 2 days (Recorded).
Two interior spots
Check the two weighted positions before you accept a flagged image as a defect. Stitching density and hardware carry a weight of 1.2 each in our scorecard against 1.0 for the main photograph, so a flag on either of those two is worth more attention than a flag anywhere else. A garment that scored 80.8% overall sat in the middle band purely because both weighted positions were marked as questions.
1. Score all five positions rather than only the flagged one. 2. If the score is 85% or above, the flag did not change the verdict and the alert was cosmetic. 3. If the score sits between 60% and 84.9%, request exactly the positions that were scored as questions and nothing else.
Write the settlement total, the chargeable weight and the divisor into one note the day the parcel clears; our three recorded cases closed at £108.34, £48.75 and £167.93 on baskets of £120, £40 and £210 (Recorded). 1. Compare the invoice within 24 hours, while the carton is open. 2. Give the line its own time before calling a gap a fault: across 24 UK routes the median ran from 8 days on a 4-parcel line to 19 days on a 3-parcel line (Recorded). 3. Expect a credit inside 3 days when a query lands, as 9 of our 14 credit events arrived that fast and the slowest took 11 days; 2 codes in the window were refused for reasons we could not establish (Recorded / Not verified). 4. Read the storage clock from the inbound scan, not the dispatch date: 8 of 24 items ran over by 1 to 14 days, the longest case worth about £2.80 at the community rate of £0.20 per item per day (Recorded / Community-reported).
Reply within 48 hours
A false alarm has a short shelf life. The retake request that works first time names the position, the distance and the lighting, and it arrives while the item is still on the warehouse bench; a request sent after the parcel is packed becomes a different and slower conversation. Our recorded retake turnaround was 6 hours on a well-formed request against 2 days on a vague one.
1. Send the request within 48 hours of the alert. 2. Name one position per request rather than three. 3. Give the photographer a distance in centimetres and a lighting condition, because both of those remove a decision from their side.
Grade the cues
Order the signals by what they predict rather than by how they read. A missing item predicts a refund; a substitute model predicts a return; a lighting artefact predicts nothing at all. On our 23 recorded flags the split was 11 on the label position, 6 on borders and edges, 4 on hardware and 2 on stitching, and only the label and hardware groups produced a real change of outcome.
1. Write the predicted consequence beside each signal before you act. 2. Handle the signals that predict a refund or a return first. 3. Leave cosmetic signals until the parcel is otherwise ready to travel, because they resolve themselves with a single photograph.
After carrying on
Proceeding past an alert is a recorded decision rather than a click, and the record is what protects you later. Keep the alert text, the photograph it referred to, the request you sent and the reply with its timestamp; on our own orders that bundle took under two minutes to assemble and answered every later question without reopening the platform.
If the parcel arrives with a defect that the alert warned about, that bundle is the difference between a documented claim and an argument. If it arrives clean, the same bundle tells you which alerts on your account are noise, and that is how the next decision gets faster.
The data point behind this note
Of 23 recorded flags, 6 covered under 12% of the image frame and 5 of those 6 proved to be lighting artefacts at the border, while the label and hardware groups produced the only real changes of outcome (Recorded).
Rates last checked 2026-10-02. Where a figure is community-reported we say so; where we could not verify it, we write Not verified instead of estimating.