The why behind the change

What's included

1Returns rising2Conversion falling

Your numbers

Linen shirt, navy

14%returned in May

Up from 6% before the March restock.

6%

Nov

5%

Dec

6%

Jan

6%

Feb

9%

Mar

12%

Apr

14%

May

Your customers' words

"Lovely fabric but I had to send it back, it runs a size small."Review, 4 Aprruns small

"Tighter across the shoulders than my last one."Return reason, 9 Aprruns small

"Sized up and it's perfect now."Review, 15 Aprruns small

AI analyst: returns rose after the March restock. 9 of the 12 reviews since then say it runs small, so the new batch is likely cut smaller.

Your numbers

Bundle landing page

1.8%conversion

Down from 3.2% before the new Meta ad went live.

3.2%

W30

3.1%

W31

3.3%

W32

3.0%

W33

2.4%

W34

2.0%

W35

1.8%

W36

Your customers' words

"Is the 20% off still on? The page says full price."Facebook comment, 26 Augoffer mismatch

"Ad says free delivery, checkout says £4.95."Facebook comment, 29 Augoffer mismatch

"Clicked through three times and couldn't find the bundle deal."Facebook comment, 2 Sepoffer mismatch

AI analyst: conversion fell from the week the new ad went live. 14 of the 31 comments on it say the offer in the ad isn't on the page, so the ad and the landing page are out of step.

Illustrative: example questions and figures.

1. 1

   Every feedback source in your warehouse

   Product reviews, post-purchase and customer surveys, ad comments and your team's docs, in one place.

2. 2

   Tagged as it arrives

   Every review, answer and comment is tagged by theme, sentiment and the product it mentions.

3. 3

   Joined to your numbers

   Each piece is keyed to the order, customer, product or ad it came from, so it sits beside your revenue.

4. 4

   Counts and quotes

   Your AI agent can count, such as reviews that mention taste by month, and pull the customer's own words behind the count.

5. 5

   Themes agreed with you

   A fixed theme list set with your team, so a theme means the same thing every month.

6. 6

   Personal data kept out

   Names and emails never reach what your AI reads.

The questions it answers

1. What do customers who buy again say that one-time buyers don't?
   • Increase retention
   • Increase LTV
2. Which complaints in reviews come before a product's returns rise?
   • Reduce refunds
   • Increase CM2
3. Why did people buy, by channel, in their own words?
   • Increase conversion rate
   • Reduce CAC
4. Sort our recent reviews and Facebook comments by sentiment, and write 20 Meta ad hooks from the most positive.
   • Increase CTR
   • Reduce CAC
5. Why do people abandon, or nearly not buy, in our survey answers? Draft a landing page test whose sections answer each main reason.
   • Increase conversion rate
   • Reduce CAC
6. Review our last five Meta ad tests and propose five CRM tests that build on what they found.
   • Increase repeat rate
   • Increase LTV

Plus more: see all the workflows →

AI analyst · in Claude● Live

What do customers who buy again say that one-time buyers don't?

Fit. Repeat buyers mention fit in 38% of reviews, one-time buyers in 12%. Price comes up about equally.

Fit, repeat38%

Fit, one-time12%

Price, repeat15%

Price, one-time17%

Tables used: reviews, customers, orders

Illustrative: example questions and figures.

What goes in

• Product reviews

  What customers say about each product, matched to what they bought.

  • Trustpilot
  • Reviews.io
  • Okendo

• Survey answers

  Post-purchase and customer survey answers, including why people bought and why they left.

  • Okendo
  • Fairing
  • Kno

• Ad comments

  What people write under your Meta and TikTok ads.

  • Meta
  • TikTok

• Your team's docs

  Business docs, meeting notes and email from Notion, Google Drive or Gmail.

  • Notion
  • Google Drive
  • Gmail

• Add-ons

  More sources can be added as Context sources, such as return reasons or support tickets.