Team
1 Aug – 31 Aug · vs previous period: 1 Jul – 31 Jul
Who answered what, how fast, and how often the draft went out unchanged. The last figure is about the company brain rather than the person: when it falls, something in the playbook has gone stale.
Messages handled
Reply log1,366
Replies that went out, across every surface
Drafts kept as they were
Company brain89.1%
A reading of the playbook, not of the people
Edited before sending
Company brain10.9%
The desk records that a draft changed, never why
Drafts the checks stopped
Company brain40
Held back before they could publish
Median first reply
Reply log1 h 47 m
From the moment the customer wrote, not from the moment the desk opened
What the period says
- The desk handled 1,366 messages, 89% of them going out exactly as the playbook wrote them. That figure is the playbook's, and it is the one to watch.
- 44% of the period's messages — 659 of them — arrived in hours when nobody was on the desk. The evening peak is the largest part of that, and no rota covers it today.
- DY carries the worst median at 210 min, which is a staffing fact rather than a performance one: one seat covers Saturday and none covers Sunday, so everything that arrives in that gap starts its clock late.
- DY needs an edit on 19% of drafts, against 7% for AY. The two cover different surfaces, so the gap points at which templates are stale rather than at who is writing.
When messages arrive, and when anyone is on
Every hour of the day against the period's own total. A bar in blue is an hour the desk covers; a grey bar is an hour when the messages arrive and nobody is there.
Seats on shift are the week's average for that hour, so Sunday's nothing is part of every figure in the column and Saturday's single seat counts once. The table behind the switch carries the seats beside the messages, so the split is never colour alone. 44% of the period's messages — 659 of them — arrive in an hour with no cover, the largest single one being 21:00. The busiest hour of the day is 10:00.
The desk
One row a seat. “Kept as they were” is a reading of the playbook rather than of the person: when it falls for a seat, the templates that seat leans on have gone stale, and the fix is a change to the brain rather than a word with anybody.
| Seat | Handled | vs previous period | Median first reply | Kept as they were | Edited | Checks stopped | Covers |
|---|---|---|---|---|---|---|---|
| AYDesk lead | 519 | ▼ 11.0% | 1 h 7 m | 93.0% | 7.0% | 7 | Trendyol Q&ATrendyol ordersHepsiburada Q&AN11 Q&A |
| MKCustomer relations | 423 | ▲ 0.2% | 1 h 30 m | 86.0% | 14.0% | 18 | WhatsAppInstagram DME-mailComplaints |
| SBSocial and reviews | 301 | ▲ 6.0% | 2 h 37 m | 90.0% | 10.0% | 4 | InstagramTrendyol reviewsHepsiburada reviewsPazarama Q&A |
| DYWeekend cover | 123 | ▼ 5.4% | 3 h 30 m | 81.0% | 19.0% | 11 | Trendyol Q&AWhatsAppInstagram |
A median is the middle reply, so one message left over a weekend does not move it the way an average would. The desk records that a draft was edited before it went; it does not record why, and nothing on this page should be read as though it does.
Inside the channel's promise
Each surface's own promise, and the share of the period's replies that landed inside it. The clock starts when the customer wrote, and it runs through the weekend.
| Surface | Promise | Answered | Inside it | vs previous period |
|---|---|---|---|---|
| Trendyol · product questions | 24 h | 348 | 97.9% | ▼ 0.2 pt |
| Trendyol · order questions | 24 h | 161 | 97.9% | ▼ 0.2 pt |
| Hepsiburada · product questions | 24 h | 128 | 97.9% | ▼ 0.2 pt |
| N11 · product questions | 48 h | 54 | 99.1% | ▼ 0.1 pt |
| Instagram · comments | 12 h | 126 | 95.0% | ▼ 0.4 pt |
| Instagram · direct messages | 4 h | 82 | 82.5% | ▼ 1.7 pt |
| 2 h | 97 | 66.9% | ▼ 1.6 pt | |
| Mailbox | 24 h | 56 | 97.8% | ▼ 0.2 pt |
| Every surface above | 1,052 | 93.5% | ▼ 0.5 pt |
The short promises are the hard ones: a two-hour promise on WhatsApp is broken by any gap in the rota, where a day on a marketplace question is broken only by a whole Sunday. 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
One question, added to the closing reply where the desk writes the last word.
It was first asked on 1 Sep, and this period ends before that. There is no score to show, and the model's reading of the messages does not stand in for one.
How the desk is staffed
What the rota says, against channels whose clocks do not stop for it.
Weekdays 09:00 – 13:00
AY · MK
The heaviest two hours of the day sit at the front of this one.
Weekdays 13:00 – 18:00
AY · MK · SB
Reviews and social are worked through in the afternoon, when the queue allows.
Saturday 10:00 – 16:00
DY
One seat. Anything that needs the warehouse or finance waits for Monday.
Sunday
—
Nobody. Sunday's messages are what Monday morning opens with, and the marketplace clocks ran all the way through.
The rota is the shop's to set. Omni does not change it; it shows what the rota costs in messages that start their clock before anybody reads them.
What the checks stopped
Every draft held back or rewritten before it could go out, and what the rule was.
- Trendyol Q&ADYHeld backSun 13 Sep, 14:22
A phone number in a reply that publishes under the listing
…you can reach us on 0212 *** ** ** and we will sort it
- InstagramMKRewrittenSat 12 Sep, 16:48
Sending a public comment to another channel by name
…write to us on WhatsApp and we'll help you there
- Trendyol reviewsDYHeld backSat 12 Sep, 11:09
An order number in a public review reply
…your order TY-4**-6390 was sent on Tuesday…
- WhatsAppMKRewrittenFri 11 Sep, 10:35
A health claim about decaf, softened but still a claim
…it will not upset your stomach at all…
- Hepsiburada Q&AAYWent outThu 10 Sep, 15:57
A variable left unfilled and sent
…the batch arrives on {eta} and the listing opens…
The rules behind these sit in the playbook, and changing one there changes every draft written afterwards. Channel rules