Case study · Customer-experience analytics

AI Journey Insight - A Walk-through

A satisfaction score only tells you how you compare to the competition overall — a single number with nothing to grab hold of. This dashboard shows how you line up against each competitor, stage by stage across the customer's purchase journey — and why, in customers' own words. AI Journey Insight is that dashboard, built for eyewear retailer S Brand.

Shengyang “Jonas” Sun — Data Scientist LLM verbatim classification Interactive linked views
All company names and identifying figures in this case study have been redacted to protect client confidentiality — the client appears as S Brand and its competitors as A–D.
Scroll to follow one investigation

The idea

Beyond a single satisfaction score

A standard benchmark tells a brand how it compares to its competitors in customer satisfaction and stops there. It never says where in the experience the brand falls behind, or why.

This tool classifies every survey response and social post into the stage of the customer journey it belongs to — from research and booking, through the eye test, to delivery and post-purchase — and extracts the themes customers raise in their own words.

The output isn't just a ranking. It's a map — exactly where the brand leads or trails each competitor, and the customer language behind every number.

A real investigation

Following one gap, end to end

Here is how an analyst actually uses the dashboard — the path from a number that looks bad to a finding the client can act on.

01

Spot the gap

S Brand is trailing Competitor A at the Sight Test

The overview compares the brand against a competitor at every touchpoint on the journey. S brand leads or track closely in most stages — but at Store Visit → Sight Test, Competitor A is clearly ahead. That single gap is worth understanding.

Overview · scores across the journey
S Brand · AI Journey Insight
Touchpoint score comparison across the customer journey

Green = S Brand, coral = Competitor A, across the eleven journey touchpoints.

02

Investigate on the journey map

One theme jumps out — and it belongs to the competitor

Zooming into Sight Test on the journey map, each bubble is a theme customers raised. Height is its impact on NPS; size is how often it comes up. Hovering the standout bubble reveals it: for Competitor A, customers mentioning a staff member by name is both frequent and strongly positive.

Hover a bubble → live detail
S Brand · AI Journey Insight
Journey map bubble chart with the Mentions Name theme tooltip open
Theme: Mentions Name Impact +5.65 NPS Prevalence 45.37% Sample 4,322
03

Read what customers said

Click through, and the raw voice of the customer loads

Clicking the bubble pulls up the underlying verbatims — the actual customer language behind the number, with the matched theme highlighted. Competitor A's Sight Test reviews name the optometrist again and again: a signal that their staff build genuine personal rapport with customers.

Click a bubble → linked verbatims
S Brand · AI Journey Insight
Verbatim panel showing customer reviews naming staff members

Every metric links to its source text — one click from a score to the words behind it.

04

Quantify the gap

Same rapport when it happens — it just happens less often

The comparison table settles the question. When a customer does name their optometrist, S Brand scores almost identically to Competitor A — the experience is just as good. The difference is frequency: only about a third of S Brand's Sight Test feedback mentions a name, against nearly half of Competitor A's.

Filter to one theme → benchmark all brands
S Brand · AI Journey Insight
Tag metric comparison table for the Mentions Name theme
NPS when named — S Brand 93.65 vs Comp A 93.78 Prevalence — S Brand ~31% vs Comp A ~45%

The table also works in reverse: start from a theme you care about and see where you stand.

The finding

A gap the headline score would never reveal

S Brand's optometrists build rapport just as well as the competition — they simply do it less often. That's a specific, fixable coaching problem, surfaced from a benchmark that started out only saying the brand ranked second.

Under the hood

Built end to end

Access runs on authentication I implemented in-house — no third-party identity provider and no external dependency. Everything past this screen, from the data pipeline through the interactive charts, is a system I designed and built myself.

Every brand and figure on the following screens is anonymised — the real client appears as S Brand and its competitors as A–D — to protect client confidentiality while showing exactly how the tool works.

What it took to build

One compact tool, several moving parts

Verbatim classification

LLM tagging of hundreds of thousands of survey and social verbatims to journey touchpoints and underlying themes.

Impact metric

Each theme's effect on NPS, measured as segment score against baseline and weighted by how often it appears.

Linked, interactive views

Bubble map, touchpoint scores, verbatims and comparison table stay in sync — click one, the rest respond.

Built in-house

Custom authentication and application, no third-party dependencies — designed and shipped end to end.

Shengyang “Jonas” Sun · Data Scientist Client data anonymised throughout · S Brand and Competitors A–D