# Cadence Relaunch Proposition — Sellable Now

*11 June 2026 · Synthesised from full project review: briefs (Apr 19 + Apr 28), Phase 1 results, mockups v1–v6 + verticals, analysis code, KARLA infrastructure*

## What the review found (one paragraph)

The science was run honestly and the predictive bridge failed at country/CCI scale: 0/32 features pass Bonferroni, music *degrades* CCI forecasts by 4–15% RMSE, and the multi-signal stack (Netflix, GDELT, Wikipedia, weather) adds nothing at monthly national granularity either. But the descriptive proxy is real (tempo × savings intent r=−0.515 TIER 1; mode-major × CCI r=+0.433 TIER 2), the mockup craft is genuinely better than incumbent output, and the data infrastructure works — except **every fetcher has been stalled since ~20 April** and Phase 2 (the brand-level comeback test) was locked but never executed. The homepage still sells the claim that failed ("30–90 days before the macro indicators confirm it").

## The strategic error to correct

The pivot bet the product on an econometric claim no buyer was asking for. WGSN, Stylus, Mintel sell **confidence, freshness, and narrative** — none of them backtest anything, and none of their buyers ask them to. Cadence tried to win on prediction and lost; meanwhile its actual advantages — behavioural data, daily cadence, multi-signal triangulation, editorial voice, LLM-native delivery — were never the headline.

**Stop selling prophecy. Sell the fastest, most honest cultural read on any market.**

## The proposition

> **Cadence — the behavioural cultural read.**
> Incumbents ask people what they feel, quarterly. We measure what markets actually play, watch, search, and read — daily, across 25+ countries — and publish the read monthly, with every number traceable to its source. Plug it into your own AI and interrogate it yourself.

Three pillars, each one verifiably true today:

1. **Behavioural, not surveyed.** Spotify + Apple Music charts (audio-feature mood read), Netflix Top 10, Wikipedia attention, Google Trends intent, GDELT news tone, macro context. Revealed preference, not panel claims. The Amazon purchase-side chart is the only commerce-side music chart of any DSP — keep that exclusivity in the pitch.
2. **Receipts-grade.** Every figure in a Cadence report traces to a named public source and date. We pre-register our methods and publish our backtests — including the failures. No other cultural-intelligence vendor can say this; it converts the Phase 1 "failure" into the brand's trust asset. Tagline candidate: *"The only cultural intelligence firm that shows its working."*
3. **LLM-native.** Every subscription ships the Cadence Skill (Claude Skill / MCP / prompt packs): the buyer's team queries the country data inside the tools they already use. Incumbents ship PDFs; this is the "companies want it now" hook in 2026.

What we claim about mood: *"Music charts are a validated mood proxy — tempo and mode track French consumer-mood sub-indices with the strongest correlations in our published validation."* What we no longer claim: 30–90 day forecasts, until Phase 2 passes.

## Product line (sellable online, descriptive-tier pricing)

| SKU | Price | What it is |
|---|---|---|
| **Cadence Report** | £700 one-off | Monthly country read (v6 design, real data, no concept-stage filler). Monthly, not quarterly — cadence *is* the moat. |
| **Cadence Pro** | £2,400/yr | 1 user · all markets · all monthly reports · Cadence Skill |
| **Cadence Team** | £6,000/yr | 5 users + shared prompt packs |
| **Cadence Music Industry Edition** | £1,200/report or bundled | The labels/A&R/sync vertical — KWL's existing logos (UMG, Sony, Def Jam, PRS) live here; fastest first revenue |
| **Cadence Enterprise** | £30–100K/yr | Private instance, white-label, bespoke markets |

Free flagship: the France report (v6, honest edition) publicly readable each quarter — it is the content-marketing engine and the proof of craft. Field Notes (fortnightly email) feeds the funnel.

## What is blocking "start selling online" — in order

1. **Turn the pipeline back on.** All nine fetchers last ran 19–20 April (Amazon 1 May). You cannot sell "the rhythm of the data, not the pace of the survey" on a pipeline that has been off for eight weeks. Cron all fetchers; add incremental checkpoints so daily runs don't re-pull 2019–2025.
2. **Fix the homepage claim.** Remove "30 to 90 days before the macro indicators confirm it" and the locked "90-day forecast" teaser. Replace with the three pillars above. Cut all Salle/auction content from selling materials until it has real data.
3. **De-paramaterise France.** Google Trends (5 terms), Wikipedia (7 articles), GDELT queries are hard-coded to France. Make term/article lists per-country config; launch market #2 and #3 (UK, Germany — Germany is no longer a holdout if Phase 2 is shelved; if Phase 2 stays live, pick UK + Spain).
4. **Productionise the report generator.** `extract_real_values_v6.py` is the seed: data → values JSON → v6 HTML template → one editorial pass per report. Target: one country report in half a day of human time.
5. **Stripe + delivery.** Gumroad/Stripe payment link, PDF + HTML delivery, Pro tier via simple signed URL portal. No platform build.

## Phase 2 — run it, don't bet on it

The brand-level backtest is pre-registered, locked, and ~4 days of engineering from a result. Run it in parallel with launch because (a) if it passes, the £4,800 Pro tier and predictive premium reopen with receipts, and (b) either outcome is itself publishable content under the "shows its working" brand. Do **not** gate launch or positioning on it.

## Why this beats what's out there

| | WGSN / Stylus | Mintel / Euromonitor | Social listening (Sprinklr, Brandwatch) | **Cadence** |
|---|---|---|---|---|
| Signal | Surveys, semiotics, expert panels | Surveys, trade data | Social text sentiment | **Behavioural: played/watched/searched/read** |
| Cadence | Quarterly+ | Annual/quarterly | Real-time but noisy | **Daily data, monthly read** |
| Verifiability | None | Methodology appendix | Black-box sentiment | **Every number traceable; backtests published** |
| Delivery | Portal + PDF | PDF | Dashboard | **Report + LLM skill in buyer's own tools** |
| Entry price | £15K+/seat | £1,500+/report | £20K+/yr | **£700/report, £2,400/yr** |

The wedge is not prediction. It is **behavioural + fresh + receipts + LLM-native at a card-swipe price**. Nobody offers that combination.

## Two-week launch sequence

**Week 1:** cron all fetchers + checkpointing · strip homepage/v6 of failed claims and Salle filler · parameterise country configs · UK data backfill.
**Week 2:** report generator pipeline → France June edition + UK first edition · Stripe links + landing page · publish France free edition + first Field Note · kick off Phase 2 backtest in parallel.

Revenue posture: descriptive economics (£100–200K Y1) with the Phase 2 option as upside — unchanged from brief v2, but now with a launch motion instead of a waiting-room.