“You cannot take a bounce rate to a repair shop and explain the invoice.”

Measuring what the customer actually pays for

An advertiser pays for position and gets value in calls. So we measured both ends, and everything between.

Get in touch

A US marketplace for truck services · technical leadership · 2022–2026 · same engagement

~180

event types in the taxonomy

stages: position to contact

phone numbers tracked separately

Summary

We tracked around 180 event types. That number sounds like sprawl and isn’t — the taxonomy is a map of how the business makes money, from the position an advertiser paid for through to the phone call that resulted.

Designing it was a product decision far more than a technical one, and it’s the part of the analytics work we would point at first.

The problem

In a directory business the advertiser buys position and receives calls. Everything between those two points is assumption.

The constraint

Standard analytics measures neither. It counts sessions, pageviews and bounce rates — quantities that describe a website rather than the value delivered to one customer of that website. You cannot take a bounce rate to a repair shop and explain why the invoice is what it is.

What we did

Measured position as a first-class event
top1, top3, top5, top10 — recorded at the moment a listing appears, not derived afterwards. A premium listing is sold as better placement, so placement is the thing that must be demonstrable. Reconstructing it later from something else is not the same as recording it.
Recorded, not derived
Interest is granular by design
Opening hours, services, amenities, rates, reviews. Those are different questions, and a system that lumps them together as “engagement” throws away the answer.
Five questions, not one
Separated intent from contact
Revealing a phone number and actually placing a call are two events, not one. The gap between them is a conversion rate no general-purpose tool would ever produce, because no general-purpose tool knows that revealing a number is the meaningful moment in this business.
Reveal → call
Distinguished the phone numbers
Primary, secondary and after-hours are tracked separately. In truck repair that is not pedantry: a call at three in the morning is a different kind of urgency and a different kind of job.
Three a.m. is not noon
Measured the sales funnel too
Which package a visitor chose, and which tier of listing they created. So the system describes not only what an advertiser receives but how they decided what to buy.
What they get and how they chose

THE VALUE CHAIN — FIVE STAGES

Position — what was paid for

top1, top3, top5, top10 — recorded at the moment a listing appears. A premium listing is sold as better placement, so placement must be demonstrable rather than reconstructed afterwards from something else.

The taxonomy is built around the value chain rather than around the pages. That is why it measures what the advertiser buys and what they receive.

The result

~180 event types mapping the full chain from placement to call
Yearly reports that answer the only question an advertiser has at renewal
A conversion rate — reveal to call — that no off-the-shelf tool would have produced

What we’d do differently

180 event types across four years, and no scheduled review of which ones anyone still looked at. Some certainly stopped being used and kept being written. A taxonomy needs a pruning habit as much as a naming convention, and we only ever had the second.

2M writes a day, a report under a secondThe architecture that carries these 180 event types.Four re-platformings, zero downtimeThe engagement the taxonomy was designed inside.

All 18 case studies