This chapter documents exactly how — from the proprietary Earliness Index scoring framework and the 41-source data fusion layer, through named-analyst review and contrarian counter-evidence, to seven years of backtested directional accuracy against Nielsen and IRI baselines.
What follows is a research-grade disclosure rather than a marketing brief. It documents the data sources Earliestown monitors, the scoring framework used to assign lead time, the mandatory human review layer that signs off on each call, and the historical backtesting protocol that anchors every directional forecast to independently measured outcomes.
Where claims are made, citations are provided in footnotes. Where a forecast carries uncertainty, the confidence interval and the contrarian counter-evidence are published alongside the call. Readers who want to interrogate the methodology further can request the full State of Earliness companion report or book a working session with a senior analyst.
The methodology is not a single model. It is a sequence of four independent checks, each of which is reviewed against its own evidence base. A forecast cannot reach publication unless it passes all four.
Every emerging signal is assigned a score between 0 and 100, derived from velocity across the source layer, divergence from baseline behaviour, and qualitative analyst weighting. The score is calibrated against seven years of backtested trend data, so a signal at 70 today is empirically comparable to a signal at 70 in 2019.
The data layer combines 41 live signals — TikTok creative velocity, Reddit subculture dynamics, niche Discord communities, indie podcast charts, and 14 proprietary panel communities of more than 22,000 early adopters across 19 countries. Fused, not averaged: divergent sources trigger a review, not a discard.
No forecast ships without sign-off from a named senior analyst — currently 19 on staff, including the founding team. Each published call carries a confidence interval and a contrarian counter-case: the strongest evidence against the directional forecast, written by a second analyst. There is no black-box AI in the publication path.
Every directional forecast has been retroactively checked against Nielsen and IRI baselines for the period 2017–2024. The result: 81% directional accuracy at the 6-month horizon. The backtest itself is published, with horizon breakdowns and a full audit available to enterprise clients under NDA.
Earliestown fuses public social and cultural telemetry with 14 proprietary panel communities. Public sources are weighted by velocity and originality, not by raw volume. Panel sources are refreshed quarterly and stratified by age, geography, and category affinity.
Earliestown has been retrospectively checked against Nielsen and IRI measurement baselines over a continuous seven-year window. The figures below report directional agreement — whether Earliestown's published call and the eventual measured outcome pointed in the same direction — at each forecasting horizon.
| Forecasting horizon | Earliestown directional accuracy | Nielsen baseline (matched categories) | IRI baseline (matched categories) | Confidence interval (95%) | Sample size |
|---|---|---|---|---|---|
| 3-month | 89.2% | 71.4% | 68.9% | ± 2.1 pp | 1,284 calls |
| 6-month | 81.0% | 62.7% | 60.3% | ± 2.6 pp | 1,108 calls |
| 12-month | 72.5% | 54.0% | 51.8% | ± 3.1 pp | 962 calls |
| Gartner Hype Cycle hit rate, 2021–2024 | 83.0% | n/a | n/a | ± 4.4 pp | 114 trends |
1 Method: each Earliestown directional forecast was paired with the matched-category Nielsen and IRI outcome at the same horizon. Agreement counted as both pointing up or both pointing down on the same category metric. Full audit available to enterprise clients under NDA. 2 Gartner Hype Cycle hits scored where an Earliestown call preceded the trend's entry onto the relevant Hype Cycle between 2021 and 2024. Median lead time across the 114 matched trends: 9.4 months.
Every figure below is taken from the seven-year backtest record or the 2021–2024 Gartner Hype Cycle hit-rate audit, both of which are reproducible on request.
A note on what the methodology does not claim.
Earliestown publishes directional forecasts, not deterministic predictions. Accuracy figures are probabilistic and apply to the matched-category sample, not to every individual product or campaign. Financial returns for any specific client deployment are not guaranteed and have not been independently audited beyond the directional backtest record.
Book a Trend Briefing