Weekly moderation report
7 Sep – 13 Sep · against 31 Aug – 6 Sep
- Put together
- Mon 14 Sep, 09:00
- Goes out
- Mondays at 09:00, for the week that ended on Sunday
- To
- Management · Desk lead · Roastery
The five lines
- 23 messages were waiting when this went out, 5 of them already past the channel's own promise. The weekend is the shop's gap, not the customers'.
- The week carried 381 messages across the surfaces that can be counted, up 1.1% the week before.
- Median first reply was 102 minutes, up 3.0% on the week before.
- 9 cases were opened and 8 are still open; the oldest has been open 8 days.
- The checks stopped 2 drafts before they could publish, and 89% of drafts went out unedited.
Missing, wrong and damaged
A buyer who is short an item is asked to open the claim in the marketplace's own flow, because that is the route that gets the item replaced. The desk then checks it was opened in the right category — the wrong one is rejected and has to be raised again, which is the part that costs a customer their patience.
Messages about a missing, wrong or damaged item
35▲ 6.1%Channel inboxesReports raised as a case
4▲ 300.0%Case logOpened in the right category
4▲ newCase logSent back for the wrong category
0▼ 100.0%Case log
WhatsApp and Instagram
All three arrive through the desk tool the shop already uses, so they are counted as Omni received them.
WhatsApp messages
28+7.7%Channel inboxesInstagram direct messages
23−4.2%Channel inboxesInstagram comments
36+5.9%Channel inboxesCases raised on those three
2▲ 100.0%Case log
Phone
Nothing about the phone is available week by week. The monthly report carries the switchboard's own total.
Calls taken
The switchboard reports a total once a month and nothing in between. There is no weekly figure to give, so none is given.
not measured—by handEntered by handCases raised on a call
0—by handCase log
Messages by surface
Thirteen ways into the shop. Two of them cannot be counted at all and are left out of the total rather than folded into it.
Trendyol · product questions
106−1.9%Channel inboxesTrendyol · order questions
52+8.3%Channel inboxesTrendyol · reviews
55−3.5%weekly exportReview exportHepsiburada · product questions
37+2.8%Channel inboxesHepsiburada · reviews
Hepsiburada publishes no review export and no review API. Only what somebody pastes in is counted, so a total would be a guess at the rest.
not measured—by handReview exportN11 · product questions
17+6.3%Channel inboxesPazarama · product questions
7no changeby handChannel inboxesInstagram · comments
36+5.9%Channel inboxesInstagram · direct messages
23−4.2%Channel inboxesWhatsApp
28+7.7%Channel inboxesMailbox
15no changeChannel inboxesComplaint site
5−16.7%by handChannel inboxesPhone
The switchboard reports a total once a month and nothing in between. There is no weekly figure to give, so none is given.
not measured—by handEntered by handCounted across every surface that can be counted
381+1.1%Computed
Reply times
The clock starts when the customer writes, not when the desk opens. A marketplace does not stop counting at the weekend and neither does this.
Median first reply, minutes
102▲ 3.0%Reply logMessages answered
386▲ 3.8%Reply logWaiting in the queue this morning
23—Channel inboxesOf those, past the channel's promise
5—Channel inboxes
Promises kept, and the closing question
Each channel's promise is its own, and the clock is the customer's. Reviews arrive as an export or by hand and know the day, not the hour; Pazarama and the complaint site are pasted in and know only when somebody pasted them; the phone has no promise to keep. None of them is in these figures. The closing question is asked only where the desk writes the last message — e-mail, WhatsApp, Instagram messages — so its score stands for those conversations and is never spread over the whole desk.
Trendyol · product questions, inside 24 h
97.8%−0.3 ptReply logTrendyol · order questions, inside 24 h
97.9%−0.2 ptReply logHepsiburada · product questions, inside 24 h
97.9%−0.2 ptReply logN11 · product questions, inside 48 h
99.1%−0.1 ptReply logInstagram · comments, inside 12 h
94.9%−0.5 ptReply logInstagram · direct messages, inside 4 h
82.5%−1.2 ptReply logWhatsApp, inside 2 h
66.2%−2.1 ptReply logMailbox, inside 24 h
98.0%−0.1 ptReply logInside the promise, every surface that has one
93.6%−0.5 ptComputedClosing question, average score out of five
4.2—Reply logAsked, where the desk wrote the closing message
52—Reply logAnswered
19—Reply logShare of the desk's answered conversations with a score
4.9%—Computed
Quality
A draft that goes out unedited is a measure of the playbook rather than of the person who approved it. When the figure falls, something in it has gone stale.
Drafts the checks stopped
2—Company brainDrafts rewritten before they went
2—Company brainShare of drafts sent unedited
89.1%—Reply log
The case log by kind
A case is a message that needed something done rather than something said, and it closes when the customer has what they were owed — not when the warehouse closed its ticket.
Missing item
1▼ 50.0%Case logWrong item
2▲ newCase logDamaged
2▲ newCase logLate or not delivered
2▲ newCase logRefund
1— no changeCase logQuality
1— no changeCase logOpened in all
9▲ 125.0%Case logStill open this morning
8—Case logMedian days to close one
1 d—Case log
Returns and exchanges
Read from the marketplaces' claims APIs and the web store's return form. Since the desk moved onto the platform, every decision is drafted by Omni and approved by a person before it is written back; the money itself moves on the marketplace or on finance's Thursday list, not here.
Requests received
32+3.2%Claims APIsRefunded
18no changeClaims APIsReplaced or sent again
8+14.3%Claims APIsRefused
6no changeClaims APIsMedian hours to a decision
15.1 h+0.5 hClaims APIsDecided by the channel because nobody did
1+newClaims APIsWaiting on a decision this morning
5—Claims APIs