A video-only account could never diagnose itself — so we built the measurement that turned guesses into signal.

Intro

This is the Data & Analytics cut of the Pilates Hunter engagement — and it’s worth being precise about what it is and isn’t. This is a measurement and diagnosis story, not an attribution rebuild. Phase I didn’t touch the tracking stack: no Conversions API rebuild, no CRM, no closed loop. What it built was theread— the benchmarks, the awareness-stage reads, and the diagnostics that made a stuck account legible.

The account had run video-only for as long as it existed, which created a quiet but fundamental problem: there was never a format comparison to read performance against. When performance hit a wall, the cause stayed hidden — because the data to diagnose it had never been created. We fixed that by benchmarking each creative type on the right metric, mapping every creative to an awareness stage, and surfacing the patterns underneath the individual wins. That’s the layer themediaandcreativecuts stand on.

Meet Pilates Hunter.

Four clinical-and-fitness pilates studios across Jakarta, certified instructors, condition-specific programming. The full brand story lives on theoverview case. This page is about the measurement the numbers run through — the part nobody sees, that lets every other claim be read honestly.

The challenge — an account with no way to read itself.

It wasn’t a tracking failure in the usual sense — purchases weren’t vanishing into a booking system somewhere. It was simpler, and just as limiting: the account had only ever run one format, to one kind of audience, judged on one kind of number.

1. No format baseline existed. Video-only means nothing to compare. A static’s CPL is only meaningful next to a video CPL — and that comparison had never been generated, so “is video the right format here?” was unanswerable.
2. Mixed creative types, one yardstick. TOFU reach videos and lead-gen creatives do completely different jobs — reading them on a single metric hides which is actually working and which is just cheap.
3. Wins weren’t read for patterns. Individual creative results without a structural read are anecdotes. The account had wins; it didn’t have the read that turns wins into a repeatable thesis.
4. Open questions had no instrumentation. Mid-video drop-off, CPL movement week to week, branch-vs-creative — none could be answered without deliberately reading the data for them.

The read, before → after — what the account could actually see:

Our approach — build the read, then surface the patterns.

We didn’t add a dashboard. We built the measurement discipline the account was missing: the right benchmark per creative job, an awareness-stage read on every creative, and a structural pass that separates pattern from noise.

1. Two benchmarks, matched to the job

Strategic reasoning

A reach video and a lead-gen static are not doing the same thing, so they can’t be judged on the same number. We benchmarked TOFU creatives on cost-per-profile-visit and lead-gen creatives on CPL plus click-to-lead rate. The moment the metric matches the job, “which creative is working” stops being an argument and becomes a number.

2. Read every creative by awareness stage

Strategic reasoning

Each creative was tagged by the awareness stage it actually spoke to — cold, aware, most-aware. That read is what exposed the blind spot: a whole product-aware segment with nothing built for it. The static thesis came directly out of this analysis; without the awareness read, the static is a guess instead of a diagnosis.

3. Surface the patterns, then name the open questions

Strategic reasoning

Individual wins matter less than the patterns underneath them. We ran a structural pass across Weeks VI–VII and kept only the reads that held across both — format, audience, distribution. And where the data ran out, we named the question instead of inventing an answer. A diagnosis that admits what it can’t yet see is worth more than one that pretends.

The benchmark framework — one read, three creative jobs:

A note on what’s shown: the working read is roughly10× more nuanced— each creative type splits by audience temperature, placement, and conversion event, with its own benchmark per cell. The framework above is the floor, not the ceiling.

Three patterns the measurement surfaced.

These appeared consistently enough across both weeks to be structural, not noise — and each one is owned in depth by another cut.

1. Format — static beat video on CPL, by a wide margin. — The static’s Week VI CPL came in 38.3% below the best branch video’s. Not a verdict on video — a verdict on audience match. A primed, product-aware segment needs a clear offer, not a hook and a story. (Owned by the Creative cut.)
2. Audience — named conditions beat generic fitness. — Skoliosis 38.70% hook vs the abstract Cara Duduk at 23.19%. When the creative names the exact condition the audience already has, both hook rate and click-to-lead rise. (Owned by the Creative cut.)
3. Distribution — one cross-branch creative beat four branch-specific ones. — 32 leads from one static vs 6–19 per branch video, at lower CPL. The condition-aware message travels across branch geography; the silo was limiting reach, not protecting relevance. (Owned by the Marketing cut.)

The shift, before → after.

What the data still can’t answer yet.

An honest read names its limits. Three questions Phase I surfaced but didn’t resolve — each one instrumented for Phase II rather than guessed at.

What building Pilates Hunter’s read taught us.

1. You can’t diagnose what you never measured against. — A single format means no baseline. The most important thing Phase I created wasn’t a creative — it was the comparison that lets every future creative be judged.
2. Different creative jobs need different yardsticks. — Judging a reach video and a lead static on one number hides which is working. Match the metric to the job and the account stops arguing and starts reading.
3. Patterns beat anecdotes. — Three reads that held across two weeks outrank seven one-off wins. The structural pass is what turns a good fortnight into a repeatable thesis.
4. Name the open question; don’t paper over it. — CPL creep, mid-video drop-off, Menteng — a diagnosis that admits its limits is more useful than one that guesses. The open questions are the Phase II roadmap.

The rest of the Pilates Hunter engagement.