H

Harbor Lane

Customer Voice

Inbox23
Reviews
Complaints8
Returns5
Topics
Playbook
Team
Alerts6
Report
Sources

Data as of 10:12

  • Trendyol · product questions: read 4 min ago
  • Trendyol · order questions: read 4 min ago
  • Hepsiburada · product questions: read 8 min ago
  • N11 · product questions: read 22 min ago
  • Instagram · comments: read 11 min ago
  • Instagram · direct messages: read 3 min ago
  • WhatsApp: read 2 min ago
  • Mailbox: read 6 min ago

Reviews and two panels are not read on a clock; they arrive as an export.

Sample data
Omni*module 06 · demoSales Cockpit demo
H

Harbor Lane · Customer Voice

Data as of 10:12 · sample data

Omni*

Topics

7 Sep – 13 Sep · vs last year: 7 Sep 2025 – 13 Sep 2025

TodayYesterday7 daysThis monthLast month90 days12 months
Previous periodLast year

What the week's messages were about, how much of each topic reads as unhappy, and where it concentrates. The counts come from the inbox; the sentences come from the model, and each one carries the figure it was read from.

The period in short

  • The period brought 406 messages, +102 against the period before. What is in it is the largest topic at 24.6% and is almost entirely neutral.
  • Grind and brewing is the mover, up against the period before since the brewing guide of 8 September. The questions have one answer each; the reviews on the same subject do not, and that is the part to read twice.
  • Where an order is, and missing or wrong items, are 24.9% of the messages between them and the great majority of the unhappy ones. Neither is answered by writing better — they are answered by the parcel being right.

Messages in the period

Channel inboxes

406

▲ 33.6%vs last year

Every surface, questions and reviews together

Reads as unhappy

Analyst

27.1%

▼ 1.2 ptvs last year

The model's reading of the text, not a score anybody gave

Biggest topic: what is in it

Computed

24.6%

— no change vs last year

Spread across the catalogue

Median first reply

Reply log

1 h 42 m

Across the period's messages; reviews arrive weekly and are left out of it

Messages, and the unhappy part of them

Two lines rather than a dozen: the channel palette belongs to channels and the status colours are spoken for, so the topics are ranked below instead of coloured here.

Channel inboxes
0204060807 Sep8 Sep9 Sep10 Sep11 Sep12 Sep13 Sep

“Unhappy” is the model’s reading of what was written, weighted by topic. Nothing on these surfaces asks a customer whether they were satisfied, so the page never calls it satisfaction.

Every topic

Ranked by volume, with what each one is in the desk's own words. A topic exists because somebody has to do something different about it.

Channel inboxes
  • What is in it

    10024.6%
  • Where is my order

    6616.3%
  • Grind and brewing

    mostly House Blend 1 kg

    5513.5%
  • When is it back

    mostly House Blend 1 kg

    4310.6%
  • Missing, wrong or damaged

    358.6%
  • Returns and refunds

    327.9%
  • Payment

    235.7%
  • Subscription

    mostly Monthly Subscription · 2 × 250 g

    204.9%
  • Just saying thanks

    163.9%
  • Invoice

    71.7%
  • Everything else

    71.7%
  • Wholesale and office

    20.5%

A rising topic is not the same as a problem: what is in the bag is the biggest topic on the page and is almost entirely neutral. The column to read beside volume is the unhappy share.

Which product draws which question

The topics that attach to a product, against the lines that draw the most messages.

Computed
ProductWhat is in itGrind and brewingWhen is it backWhere is my orderMissing, wrong or damagedReturns and refundsSubscriptionTotal
House Blend 1 kg101515622151
Ethiopia Yirgacheffe 250 g832533125
House Blend 250 g722532122
Espresso Blend 1 kg621422118
V60 Starter Set621422118
Colombia Huila 250 g621422118
Hand Grinder521422117
Ethiopia Yirgacheffe 1 kg521322116
Monthly Subscription · 2 × 250 g521322419
Total58322538201912204
lessmore

Wholesale, invoices and applications are about the shop rather than a product, so they are not here; nor is the long tail of the catalogue. The table's total sits below the period's for both reasons, and is not padded to match it.

Topics worth a second look

Grind and brewing

Analyst

The brewing guide went out on 8 September and the questions followed it: which grind for which machine, on the blends rather than the single origins. The web store saw the same week on the V60 set it recommends, which is the cheerful half of this. The other half is on the Reviews page, where a run of low stars says the grind picked at checkout is not always the grind that shipped.

Grind and brewing, this period
55 messages, 13.5% of them all
The evidence is on the Reviews page

When is it back

Analyst

One line is out and people are asking when it comes back. The answer is only as good as the incoming batch date, and the draft quotes the warehouse rather than a guess — so if the batch slips, every waiting reply slips with it.

When is it back, this period
43 messages, 10.6% of them all
The evidence is on the Inbox page

Subscription

Analyst

Pausing, skipping, changing the grind, and one double charge. It grows with the subscription itself, and half of it is a question the playbook still cannot answer — whether a month can be paused rather than cancelled.

Subscription, this period
20 messages, 4.9% of them all
The evidence is on the Playbook page

Wholesale and office

Analyst

The smallest topic on the page and the one worth the most: cafes and offices asking for a price list. The desk cannot quote, so the whole job here is to catch it within the day and hand it to sales with the right detail attached.

Wholesale and office, this period
2 messages, 0.5% of them all
The evidence is on the Inbox page

Ask the inbox

The figures in an answer are counted from the messages themselves; the model only writes the sentence around them.

How this page is worked out

Where each figure comes from, and what none of them can say.

Analyst
  1. 1Every surface is read on its own clock and each message is filed under one topic, the desk's own list rather than a model's clustering. A topic exists because somebody has to do something different about it.
  2. 2Sentiment is the model's reading of what was written. Nothing on these surfaces asks a customer whether they were satisfied, so the page says “reads as unhappy” and never “satisfaction”.
  3. 3The comparison is the same length of days immediately before the period, or the same dates a year back. A day still running is only ever set against the same hours of another day.
  4. 4Reply times come from the reply log. Reviews arrive as a weekly export rather than message by message, so their waiting time is known to the day and is left out of the median.
InboxReviewsCasesAlerts