Measuring where you are on an adoption curve
Cluster F holds four models of how a new thing spreads and no instrument for locating yourself on one while it is happening. This page is the measurement side of that cluster: what an observable would have to do, what one candidate instrument does, and where it fails. Written 2026-08-08 from a question about the cross-spoke seam with search-marketing-wiki; the instrument it examines is emerging-category-search-signals (Taylor, Search Engine Land, T2), which lives in that spoke because its purpose is commercial positioning.
The gap the models leave
The cluster's four members all describe shape:
- technology-adoption-curve — Rogers' five adopter groups under an S-curve.
- crossing-the-chasm — Moore's discontinuity between visionaries and pragmatists.
- gartner-hype-cycle — expectations over time, a separate axis from adoption.
- bass-diffusion-model — the maths,
dF/dt = (1 − F)(p + qF).
None of them tells you where a particular technology sits today without waiting for the history that would let you fit the curve. The corpus says this about itself already: Bass's parameters are "fit retrospectively" and its forecasts are "sensitive to early-data noise"; the hype cycle is "not empirically validated"; tech-adoption-curve-twenty-years found most significant post-2000 technologies were never flagged early. A model fitted after the fact is not a measurement.
What the search signals observe, mapped onto the models
emerging-category-search-signals proposes five signals of a forming category. Three of them are the theory's own concepts becoming visible; two are not in the theory at all.
| Signal | What it corresponds to |
|---|---|
| Buyer phrasing absent from keyword tools | Rogers' first 2–3% — too few adopters for language to aggregate. The instrument's blindness is the reading |
| Formal vocabulary first (standards, statutes, job titles) | Institutional rather than individual adoption — see the power gap below |
| Contested labels at similar volume | Before critical mass, the point where Rogers says diffusion becomes self-sustaining. Moore's version: visionaries each coin their own framing; the pragmatist majority needs one name |
| Difficulty below demand | Not in any of the four models — a supply-side count |
| Small firms outranking incumbents | Not in any of the four models — market structure before consolidation |
Two consequences worth keeping.
The supply side is a genuinely new axis. Rogers, Moore and Bass all count adopters. Keyword difficulty and results-page composition count competitors. A category can be read from either side, and the models only have one of them.
It splits Bass's two coefficients. The social listening the article pairs with keyword data — TikTok
and YouTube search, Reddit, autocomplete — detects word-of-mouth (q) igniting months before
aggregate search volume registers it, while "formal vocabulary first" behaves like external influence
(p). Bass fits both from one adoption series after the event; two different observables, read live,
is a different kind of evidence.
It also lands on the cluster's admitted hole
technology-adoption-curve records that Rogers has no term for power — no state, no regulator, no organised opposition — and quotes the claim that "states ultimately decide how fast technology is adopted," filed there as an unresolved tension.
Searches for ISO 42001, the Colorado AI Act, AI regulation are that missing variable arriving as data. They are not consumers adopting anything; they are an institutional layer moving first and demand following it. If the power critique is right, this is the part of the curve the models cannot explain and the one the search signals see earliest.
Where the instrument fails
Four limits, and they are not small.
- Search is interest, not adoption. Someone searching "AI governance consultant" has adopted nothing. The whole method is a proxy, and this page said nobody had measured how good a proxy it is. Retracted 2026-09-04 (absence-claim drain): people have, repeatedly, in public health rather than technology — Google Trends against national vaccination series gives correlations of roughly 0.46 to 0.90 depending on the lag, and comparable work exists for dengue and COVID case counts. The famous negative result is the sharper one: Google Flu Trends failed because search volume tracked media coverage rather than infection, which is this bullet's own worry, confirmed by a measured collapse rather than asserted. So the proxy has been validated where a real denominator exists to validate it against, and technology adoption is precisely the case with no such series — that, not the absence of any validation work, is the limit here.
- No denominator. Rogers' curve and Bass's equation both need a market potential; Bass assumes a fixed one. Search volume supplies a numerator with no total, so it cannot say what fraction of the eventual market has arrived — which is the one thing the curve is for.
- It cannot separate hype from adoption. A rising line fits the Peak of Inflated Expectations as well as it fits real uptake. The article's own rules — a twelve-month window, and commercial-intent terms before believing there is budget — are an ad-hoc attempt at exactly that separation, which is the sentiment-versus-adoption distinction gartner-hype-cycle already says needs both axes measured.
- No false-positive rate. Every case in the source is told forward from a category that emerged, so the signals have no measured precision. Search-marketing-wiki carries this as its own growth edge.
The standing that leaves it with
Strongest exactly where the models are weakest — the far-left tail, before there is enough history to fit anything — and weakest where the models are strong, in describing the overall shape and the eventual share. That is a useful complement rather than a competitor, and it is one unvalidated instrument, not a method the corpus can vouch for.
What would change that: a study that takes a fixed set of candidate categories, records the five signals at time zero, and reports how many became categories. That is the missing piece on both sides of this seam — a false-positive rate for the practice, and the corpus's first live-fitted diffusion measurement.
A second instrument hits the same wall (added 2026-08-14)
org-mode-adoption tried to measure one technology's uptake with three instruments that have nothing to do with search signals: a community survey (n=7,344), Q&A tag shares on two Stack Exchange sites, and package-manager download counts. It ran into the identical no-denominator failure. Every figure it produced is a ratio inside a population — Emacs users — whose size has never been measured (re-checked 2026-09-15: still true — the 2025 Stack Overflow survey gives Emacs ~8–9% of respondents, a share of a self-selected sample, not a population count), so the result locates Org precisely relative to its niche and not at all on a curve.
Two instruments this different failing the same way promotes the no-denominator problem from a limit of the search method to a property of the measurement problem itself. Both can say this is spreading; neither can say what fraction has arrived, which is what Rogers' curve and Bass's equation both need.
The Org case also adds a mechanism none of the four models name. Org entered Emacs core in 2012, and a bundled technology is invisible to install telemetry at exactly the moment bundling makes its adoption near-universal — MELPA can count six Org add-ons and cannot count Org. Distribution decisions of that kind sit alongside the missing power variable above: another actor deciding uptake, with no term for it in the theory.
Related
technology-adoption-curve · bass-diffusion-model · gartner-hype-cycle · crossing-the-chasm · tech-adoption-curve-twenty-years · everett-rogers · emerging-category-search-signals · org-mode-adoption · emacs-user-survey-2020 · synthesis