Pricing
Reading costs nothing and always will. What costs money is the record underneath — as data you can query yourself, or as a question answered properly.
Last reviewed September 4, 2026
The four tiers
Prices in US dollars, excluding any tax. Published rather than quoted, so nobody gets a different number depending on how they look.
Read
Everybody. There is no account, no wall and no email required to read anything.
- All 51 pages of original analysis
- The full archive of 14,951 items, browsable and searchable
- A plain-text version of every page, at the same address with .md appended
- Every figure computed at build time, with its source and period stated
This tier is not a trial and does not expire. It is what makes the rest worth buying.
Academic
Universities, non-commercial research, and anybody publishing what they find.
- The corpus as structured data, with its documented defects
- A written description of how it was collected and where it is thin
- Help framing a question the record can actually answer
Attribution and a copy of what you publish. Cited work makes the corpus more useful, which is the whole return.
Commercial data
Competitive intelligence, market analysis, and product teams testing a claim against a decade and more of coverage.
- 14,951 items from 969 publications, collected since 2014 and still growing
- Headline, date, attribution, subject tags and summary where one exists
- The defects documented in writing, not discovered later
- Updates for the term, and a named contact for questions about the data
A one-time snapshot without updates is $2,500.
Commissioned analysis
A question you want answered properly, to the same standard as everything else here.
- One question, scoped in writing before anything starts
- Answered under the seven checks every page passes
- Every figure computed and reproducible from the named source
- An explicit statement of what the record could not support
Kept private instead of published: $9,000. The reason that costs more is below.
Why the numbers are what they are
A rate somebody has to ask for is a rate that moves depending on who is asking. These do not, so here is what each one is built around.
$4,000 sits below where procurement starts. Above five thousand, most companies add a second approver, a vendor form and a security review, and a purchase that would have taken a week takes a quarter. The price is set at the point where you can still approve it yourself.
$6,000 buys a question taken all the way. Scoped so the record can actually answer it, computed from source, checked against it, and put through the same seven checks as everything else published here. Anything cheaper would have to drop the checks, and the checks are what you are paying for.
Academic use is free. The conditions are attribution and a copy of whatever you publish. Cited work makes the record more useful to everybody licensing it, including the people paying.
Why private analysis costs more than published
Every other publisher charges the other way round, so it is worth saying plainly: letting the work be published saves you $3,000.
A published piece answers your question and also shows the method working on a question we did not choose, which is better evidence than anything we could write about ourselves, and it keeps earning attention for years. Part of the fee is taken in that form. A private deliverable produces none of it, so it is paid in full.
What publication does not buy is influence over the answer. A commissioned piece states what the record shows, including when that contradicts whoever paid for it, and the scope agreed at the start is a question rather than a conclusion. Anybody wanting a predetermined finding is better served elsewhere and will be told so on the first call.
What the record can and cannot answer
Worth establishing before money changes hands, because it decides whether any of this is useful to you.
It answers questions about coverage. What the trade press wrote about, when, how much, from which publications, and how that has changed over more than a decade. Whether attention tracked events or deadlines. Which subjects arrived early, which arrived late, and which have not arrived yet.
It does not answer questions about the world. Incidence, cost, whether any control worked, what a market was worth. Those need data this record does not contain, and a pitch resting on them gets told so in the first reply rather than after an invoice.
Its defects are published in advance. Three thin months in 2016, a collection that began in two different modes, under half the items carrying a written summary, and a trailing edge the most recent weeks sit behind. How this site checks itself sets all of that out with the arithmetic attached. A buyer who reads that page and still wants the data is a buyer who will be satisfied.
What is not for sale
- Coverage, placement, or any influence over what gets written.
- A link inside editorial text, at any price.
- The mailing list, or anything identifying a reader.
- Analysis with a conclusion agreed in advance.
- Access to anything before it is published.
The same list, from the two other directions: placements are not for sale at any price, and sponsorship buys adjacency and never influence.
What would change these numbers
An audited audience. Sponsorship is not priced on this page because there is no independently verified traffic figure to price it against, and quoting one we produced ourselves would be worth nothing to a buyer. Data and analysis do not depend on audience, which is why they can be priced today.
A larger corpus. The record is still being added to, and a bigger one is worth more. Any change to these rates applies to new terms rather than to one already running, so a licence bought today keeps its price for its year while the data underneath grows.
Nothing else. Rates do not move because of who is asking, how large they are, or how the conversation is going.
Getting started
Write to [email protected] with the question you are trying to settle. Not the tier you think you want — the question. Which tier fits, or whether none of them does, falls out of that in one exchange.
Expect a straight answer about whether the record can support what you need. Being told no early costs us a sale and saves you a quarter, and a corpus sold into the wrong question comes back as a complaint that the data was thin.