Data and Research Methods
Where the numbers on this site come from: the tools, the dates, the sampling, what counts as a source, and how each original pull is stated precisely enough for someone else to repeat or contradict.
Original figures on this site come from named tools on stated dates, with the query, country, and threshold recorded so the pull can be repeated. Where a figure comes from a third party, it carries a deep link to the specific table or series rather than to the publisher.
A figure is only as good as the account of where it came from. This page is that account for the data published here, including the parts of it that are weak.
What a pull records
Every original pull states the tool, the endpoint, the country, the date retrieved, the seed set, and the threshold applied. Where the tool meters usage, the cost is recorded too. The number of calls is how much of the topic a pull actually covers, and a reader cannot judge coverage without it.
The question-demand measurement is the worked example. Tool: Ahrefs Keywords Explorer, matching-terms endpoint. Country: United States. Retrieved: 2026-08-06, at four API calls totalling 5,902 units. Seed set: manufacturing and industrial head terms plus process, vertical, company-type, certification, cost, ROI, RFQ, and directory modifiers. Threshold: 20 monthly searches.
That is stated at a level someone with a seat can repeat in about ten minutes. If a repeat returns something different, the page is wrong and gets corrected rather than defended.
Where the weaknesses are, stated rather than smoothed
Reported volume is an estimate from a clickstream panel. Panels are weighted toward consumer behaviour, and a query issued forty times a month by forty different procurement engineers generates almost no panel signal. Manufacturing SEO is characterised by low search volume and high transaction value, so the demand that matters most here is the demand the instrument sees worst.
Absence of reported volume is not absence of demand. Every conclusion drawn from a null result on this site is stated as a measurement of the tool's output rather than of the world.
Null SERP data means the tool holds no result set for the term, which it reports as null difficulty and null parent topic. The volume on those rows is low confidence, and pages carrying such figures say which rows are affected. Several headline terms in the question-demand pull are in that state.
That last point generalises into the standard this site applies to itself: a figure that survives a check is not thereby closer to correct, it is only harder to disprove.
Sampling, and what a small sample can carry
Measurements here are frequently small. A three-term, 24-position SERP sample is enough to refute a claim that directories own industrial head terms and not enough to describe the market, and the page carrying it says both.
A sample is reported with its size, its date, and the specific claim it can support. A measurement that refutes something is doing different work from one that estimates something, and small samples are much better at the first.
Manufacturing keyword research should incorporate CRM and request-for-quote data rather than search volume alone. That is the same limitation stated as advice, and it is why the keyword research treatment exists.
Third-party figures
A third-party figure carries a deep link to the actual table, page, or series identifier, the publication date, and the figure as the source states it. Where a source updates in place with no year stamp, the retrieval date is recorded instead.
Attribution to an organisation without a location inside its output does not qualify. That form of citation is what the methodology page describes failing across this site's own legacy corpus, where an 11-figure sample returned zero verifiable from the URL cited.
An absence claim records three things: the figure, who it is commonly attributed to, and what tracing returned. It is what a widely repeated figure becomes when no primary source survives tracing. The benchmarks page is built that way, and it is deliberately less satisfying than a page of round numbers.
Dates, and why every figure carries one
A figure without a date is an assertion about the present that quietly becomes an assertion about whenever it is read.
Every figure here carries a retrieval or publication date, and every page carries a lastReviewed date meaning it was read against its sources on that date. Figures attached to standards, certification revisions, and compliance deadlines are swept on a schedule, because those change without announcing it.
What is not done here
No survey data is collected. Nothing on this site is based on a proprietary panel, and no figure is presented as one.
No client data is used. The site is not an agency and has no engagements to draw on. Where a benchmark would ordinarily come from aggregated account data, this site does not publish one.
No figure here is estimated to fill a gap. Where a number would be useful and cannot be obtained, the gap is stated. That is why several pages here name a figure they decline to give and explain why the published versions do not survive tracing.
Common questions
Which tools are used?
Ahrefs for keyword, SERP, and backlink data, and search console data where a site owner provides it. Tool and endpoint are named per figure rather than in general, because different endpoints of the same tool return different things.
Why is so much of the data United States only?
The pulls are scoped to the US market and say so. A figure scoped to one country and read as global is a common failure, so the scope travels with the number.
Can a pull be re-run without an Ahrefs seat?
Not identically. The seed set and thresholds are published so a different tool can be used, and a different tool returning substantially different results is a stated way to falsify the conclusions drawn here.
How are conflicting figures handled?
Both are recorded with their sources and the disagreement is stated. The published number is not silently replaced, because a number changing without a visible reason makes the site's history unreliable from outside.
What happens to a figure that later proves wrong?
It is corrected with the correction visible, or the page is withdrawn where the premise rather than the figure was the problem. The withdrawal precedent and its reasoning are on the methodology page.
Sources
- 01
Last reviewed . Published by ManufacturingSEO.ai.