Nobody has published a measurement of how much of a downloaded document gets read. Not a survey of what people say they read. An actual measurement, with a method, of what proportion of a gated PDF is opened, scrolled or finished.

That absence is the most useful finding available on this subject, because the entire economics of gated content rests on an assumption about it. The tools that could measure it exist, sit inside document-sharing platforms, and their operators do not publish aggregate figures with methodology.

So the honest starting point is that the format has no measurement of its finished product. What follows is what happened when the other statistics used to justify one were traced back to their sources.

The measurement that does not exist

Before any of the traced statistics, the gap at the centre of the format is worth stating precisely, because it is unusual for a mature industry to have one this size.

What was searched. Academic literature, and the operators of document-sharing platforms that instrument reading behaviour page by page.

What was found. No published study, with a stated methodology, measuring what proportion of a downloaded business document is opened, scrolled, or completed.

The nearest thing available is self-reported. One survey has buyers saying they spend between five and thirty minutes on almost every content format, which is a claim about intention rather than a measurement of behaviour.

And the tools that could answer it do not publish. Platforms that track per-page reading exist and are widely used. None of them publishes aggregate figures with a method attached.

Why that matters more than it sounds. Every argument for gating rests on the document being consumed. If it is not, the gate has traded a contact detail for a file that sits in a downloads folder, and both parties have recorded a success.

One consequence worth accepting. You can measure this yourself, on your own document, in a fortnight. Host it as a page rather than a file, or use a viewer that reports scroll depth, and look. Your own number will be more useful than any benchmark, and nobody else has published one to compare it against.

The absence of any published measurement of how much of a downloaded document is readThe absence of any published measurement establishing what proportion of a downloaded business document is actually read, and the consequence of that absence for the economics of gated content. A search of academic literature and of publications by operators of document sharing platforms that instrument reading behaviour on a page by page basis found no published study, carrying a stated methodology, measuring what proportion of a downloaded business document is opened, scrolled through or completed. The nearest available material is self reported rather than measured, one survey recording that buyers say they spend between five and thirty minutes with almost every content format, which is a statement of intention rather than an observation of behaviour. The platforms capable of answering the question exist and are widely deployed, since document sharing services routinely track per page reading depth, but none publishes aggregate figures accompanied by a methodology. The significance of this gap is that every argument for placing a document behind a form assumes the document is subsequently consumed, and if it is not then the transaction has exchanged a contact detail for a file that remains unopened in a downloads folder, while both parties record the exchange as a success. One practical consequence follows, which is that a company can measure this for its own document within a fortnight, either by publishing the material as a web page rather than as a file, or by using a viewer that reports scroll depth, and that the resulting figure will be more useful than any published benchmark precisely because no comparable benchmark has been published by anyone.The number nobody has publishedWhat proportion of a downloaded document gets read?No published study with a stated methodology answers this. Not in academic literature, and not from theplatforms that instrument reading page by page.The nearest thing availableBuyers saying they spend “five to thirtyminutes” on most formats. Self-reported.And the tools that could answerThey exist and are widely used. Nonepublishes aggregate data with a method.Why the gap mattersGating trades a contact detail for a file. If nobody opens the file, both sides record a success anyway.You can measure your own in a fortnight. Nobody has published one to compare it against.
Every argument for gating assumes the document is read. No published study establishes that it is. Source : Search of academic literature and document platform publications; Demand Gen Report 2022 for the self-reported figure (2022)

The number everyone quotes, and what the original document says

The claim that most of the buying decision happens before a supplier is contacted is the foundational statistic of modern B2B content. It has a real source, and the source says something more interesting than the claim.

The document exists and is public. A 2012 report produced by a research council in partnership with a search company, drawing on a survey conducted in 2011.

The sample is stated. 1,500 customer contacts, described as decision makers and influencers in a recent major business purchase, across 22 large B2B organisations spanning ten industries.

The body text does not say 57 percent. It says “nearly 60%” and “more than one-half”. The 57 is a chart label.

And the chart behind it is a distribution, not a point. Averages by supplier run from under 45 percent to 100 percent. The headline number is the middle of a very wide spread.

Which explains the other figure people quote. The same document contains the sentence: “At the upper limit, that number ran as high as 70%.” The 70 percent that circulates as a competing statistic is the top of that distribution, from the same chart.

Two limitations belong with any use of it. The survey is from 2011 and covers large organisations only. And the measurement is a respondent’s retrospective self-estimate of how far through a process they were before contacting a supplier, which is not a thing people can observe about themselves.

None of that makes the underlying idea wrong. Buyers do research before they call. But a number with a 55-point spread, from a self-estimate collected fifteen years ago, cannot carry the weight of a content strategy.

The distribution behind the most cited business to business marketing statisticWhat the original 2012 research document actually contains, compared with the statistics quoted from it. The document is a report produced by a research council’s marketing leadership arm in partnership with a search company, drawing on a customer purchase research survey conducted in 2011. Its stated sample is one thousand five hundred customer contacts, described as decision makers and influencers in a recent major business purchase, drawn from twenty two large business to business organisations spanning all major industry classification categories and ten industries. The body text of the report does not state fifty seven percent. It states nearly sixty percent, and separately more than one half. The figure of fifty seven percent appears only as a label on the first chart. The second chart in the document presents a distribution of averages by supplier rather than a single value, and that distribution extends from under forty five percent at the lower end to one hundred percent at the upper end, meaning the headline figure represents the middle of a spread of more than fifty five percentage points. The same document contains the sentence stating that at the upper limit that number ran as high as seventy percent, which establishes that the seventy percent figure circulating as a competing statistic is in fact the upper limit of the same distribution rather than an alternative measurement. Two limitations attach to any use of these figures. The survey was conducted in 2011 and covers large organisations exclusively, drawn from the client bases of twenty two supplier organisations that were members of the publishing council, making it a convenience sample with no published response rate. And the underlying measurement is a respondent’s retrospective self estimate of how far through a purchase process they had progressed before contacting a supplier, which is not a quantity a person is able to observe about their own behaviour.One chart, two famous numbersThe distribution of supplier averages in the original documentunder 45%100%57%the chart label70%“the upper limit”The sample1,500 contacts across 22 large B2Borganisations, ten industries.Survey conducted in 2011.What the body text says”Nearly 60%” and “more than one-half”.The 57 is a chart label, not a sentence.And the measurement itselfA respondent’s retrospective self-estimate of how far through a process they were before contacting a supplier.Which is not something a person can observe about themselves.
Both figures in circulation come from the same chart. One is its midpoint, the other is its upper limit. Source : The Digital Evolution in B2B Marketing, CEB Marketing Leadership Council with Google, 2012 (2012)

The buying committee number, and its quiet withdrawal

The second pillar of the gated content argument is that many people are involved, so you must reach them at scale. That figure has a stranger history.

The claim, verbatim as published. “The typical buying group for a complex B2B solution involves six to 10 decision makers, each armed with four or five pieces of information they’ve gathered independently and must deconflict with the group.”

What accompanied it. No study name. No sample size. No date. No methodology. It appeared on a marketing page.

And it is no longer there. The sentence has been removed from the page that published it. The same organisation now describes buying groups “ranging from five to 16 people across as many as four functions.”

Which is a wider and less convenient range. Five to sixteen does not support the same argument as six to ten, because it admits that a great many purchases involve a handful of people rather than a committee.

A companion figure went the same way. The claim that buyers spend only 17 percent of their time meeting potential suppliers appeared on the same page, without methodology, and is also gone.

One related figure does have a sample. A 2025 release reports that 61 percent of B2B buyers prefer a purchasing experience without a sales representative, from “a survey of 632 B2B buyers conducted in August through September 2024.” Sample size and field window are given; recruitment and geography are not.

The pattern is worth naming. The most repeated statistics in B2B marketing were published without methodology, and the organisation that published two of them has quietly stopped saying them.

Traceability status of three statistics commonly used to justify producing gated business contentThe traceability status of three statistics commonly used to justify the production of gated business content, established by locating and reading the primary sources. The first states that a majority of the purchase decision is complete before a supplier is contacted. Its primary source exists and is publicly accessible, being a 2012 report drawing on a 2011 survey of one thousand five hundred contacts across twenty two large organisations. However the figure quoted is a chart label rather than a statement in the body text, which says nearly sixty percent, and the underlying chart presents a distribution of supplier averages ranging from under forty five percent to one hundred percent, with the alternative figure of seventy percent appearing in the same document as the upper limit of that distribution. The second states that a typical business buying group contains six to ten decision makers each carrying four or five pieces of independently gathered information. Its published source carried no study name, no sample size, no date and no methodology, appearing on a marketing page, and the sentence has subsequently been removed from that page, with the same organisation now describing buying groups ranging from five to sixteen people across as many as four functions. A companion statistic from the same page, stating that buyers spend only seventeen percent of their time meeting potential suppliers, was published on the same basis and has likewise been removed. The third states that sixty one percent of business buyers prefer a purchasing experience without a sales representative. This figure is accompanied by a stated sample of six hundred and thirty two business buyers surveyed between August and September 2024, making it the only one of the three published with a sample size and field window, although the recruitment method, sampling frame and geography remain undisclosed. The pattern across all three is that the most repeated statistics in business marketing were published without methodology, and that the organisation responsible for two of them has quietly ceased to state them.Three statistics, traced to source”57% of the decision is done before contact”source foundReal document, real sample of 1,500. But the figure is a chart label, the body says “nearly 60%“,and the chart runs from under 45% to 100%.”Six to ten decision makers”withdrawnNo study name, no sample, no date, no method. The sentence has been removed from the page.The same publisher now says “five to 16 people”.“61% prefer a rep-free experience”has a sample”A survey of 632 B2B buyers conducted in August through September 2024.”Sample and field window given. Recruitment, sampling frame and geography not.Only the third can be quoted as published. The first needs its spread attached. The second no longer exists.
One is a chart midpoint, one was withdrawn by its publisher, and one has a sample. Only the third is quotable as published. Source : CEB and Google 2012; archived Gartner page 2023 and its current version; Gartner press release, 25 June 2025 (2025)

The surveys behind “buyers consume X pieces of content”

This family of statistics justifies the volume argument, and it does not hold up either.

The most-cited annual survey publishes its field dates and its count. The most recent edition states: “The survey was fielded between June 24 and August 14, 2025 … The survey drew 1,229 global responses. This article reports on the 1,015 B2B marketers, mostly from North America.”

What it does not publish. Recruitment method, sampling frame, response rate. It is a self-selected sample drawn from the publisher’s own lists.

And it measures the wrong population for the claim. It surveys marketers about what they do. Any figure from it describing the “best performing format” is an opinion held by marketers, not a measurement of audience behaviour.

A second survey supplies the content-count figures, and its bands overlap. One edition reports 1 to 3 at 28 percent, 3 to 5 at 42 percent, 5 to 7 at 19 percent, more than 7 at 11 percent. The values 3 and 5 each appear in two bands, so any aggregate built from them is not reproducible.

And its most recent edition contains no methodology at all. No sample size, no field dates, no demographics. It is sponsored by a content agency.

One frequently quoted infographic deserves a specific warning. The claim that content marketing costs 62 percent less and generates three times the leads sits on a gated infographic with no date, no methodology, no sample, and no source given for any of the statistics it contains.

Methodological disclosure in the annual surveys used to justify content production volumeThe extent of methodological disclosure in the annual surveys most commonly used to justify decisions about content production volume and format, together with the population each actually measures. The most cited annual content marketing survey publishes its field window and its response count, stating for its most recent edition that the survey was fielded between the twenty fourth of June and the fourteenth of August 2025, that it drew one thousand two hundred and twenty nine global responses, and that the reported analysis covers one thousand and fifteen business to business marketers mostly located in North America. The preceding edition disclosed equivalent detail for its own field window and count. What neither edition publishes is the recruitment method, the sampling frame or the response rate, making the sample self selected from the publisher’s own distribution lists rather than drawn probabilistically from a defined population. More significantly, the survey measures marketers reporting on their own activity, so any statement derived from it about which content format performs best expresses an opinion held by marketers rather than a measurement of audience behaviour. A second annual survey supplies the figures describing how many pieces of content a buyer consumes before purchase, and its published response bands overlap, one edition reporting one to three at twenty eight percent, three to five at forty two percent, five to seven at nineteen percent and more than seven at eleven percent, so that the values three and five each fall within two bands and any aggregate constructed from them cannot be reproduced. The most recent edition of that same survey contains no methodological disclosure whatsoever, providing no sample size, no field dates and no respondent demographics, and is sponsored by a content production agency. Separately, a widely circulated infographic claiming that content marketing costs sixty two percent less than outbound marketing while generating three times as many leads is gated behind a registration form, carries no date and no methodology, and provides no source for any of the statistics it presents.What the surveys disclose, and what they measurePublishedField window: 24 June to 14 August 2025Responses: 1,229 globalAnalysed: 1,015 B2B marketersNot publishedRecruitment methodSampling frameResponse rateAnd the population is the wrong one for the claimIt surveys marketers about what they do. “Best performing format” from it is a marketer’s opinion, not audience behaviour.The content-count bands overlap1-3, 3-5, 5-7. The values 3 and 5 appearin two bands each. Not reproducible.And one to avoid entirely”62% cheaper, 3x the leads”: gated, undated,no method, no source for any of its stats.The most recent edition of the second survey contains no methodology at all. Not a reduced one. None.
Field dates and counts, yes. Recruitment and response rates, no. And the population surveyed is marketers, not buyers. Source : Content Marketing Institute annual B2B research, and Demand Gen Report content preferences surveys (2026)

The origin story is off by at least thirty-two years

Since the format is usually justified partly by its pedigree, the pedigree is worth checking too.

The common claim. That the white paper originates with a British government paper of 1922.

What the parliamentary record shows. The term is in ordinary use decades earlier. In a Commons debate of 4 August 1890: “In the White Paper Report there is a reference to a transfer of territory.”

And by 1907 the papers were a numbered series. A member asks whether the President of the Board of Trade “will have White Paper 340, Condition of Trade and People, reprinted with figures included to 1906.”

The etymology is literal, and visible in the record. In 1881 a chairman distinguishes documents “delivered this morning with the blue paper Votes” from “the white paper copy of the Votes”. It was the colour of the paper, as against the Blue Books.

The 1922 paper exists, and is not an origin. It is a command paper of June 1922 concerning correspondence on Palestine. The exact day is not stated in the institutional sources consulted, so it should not be given.

And the scholarly position goes further. A 2017 article in a technical communication journal argues that most accounts of the format’s history “assume that marketing white papers evolved from government white papers. They conflate genre history with etymology.”

Which is the useful correction. The marketing document did not descend from the parliamentary one. It borrowed a name that had been in use for a different kind of document for over a century, and the borrowing carries no authority with it.

So what is a white paper actually for

Strip out the unsourced justifications and something usable remains, but it is smaller and more specific.

It is a filter, not a reach mechanism. Somebody willing to exchange a work email address for a twenty-page document about a narrow problem has told you something about their situation. That signal is real, and it does not depend on whether they read it.

It is an argument you can hand to someone else. In a purchase involving several people, one participant needs something to forward. A document is a portable version of your position, and its job is to survive being read by someone you never spoke to.

It is a forcing function on your own thinking. Writing twenty defensible pages about a problem reveals which parts of your position you cannot support. That value accrues whether or not anyone downloads it.

And it is a durable asset in a way a post is not. It can be cited, updated, and referenced in a proposal three years later.

What it is not is a lead volume instrument. The statistics that make that case do not survive tracing, and the download is not evidence of reading.

Which suggests the honest measurement. Not downloads. Whether the document appears in conversations you can observe: mentioned on a call, forwarded internally, referenced in a reply. Those are countable, they are few, and they are the actual outcome. They are also the measurement nobody owns, since the document is written by one party, gated by another and promoted by a third, and it only gets counted when brand, site and campaigns sit with the same team.

The functions a gated business document performs and the measurement appropriate to eachThe functions a gated business document actually performs once the unsourced justifications are removed, and the measurement appropriate to the outcome. The first function is filtering rather than reach. A person willing to exchange a work email address for a twenty page document about a narrow problem has disclosed something about their situation, and that signal exists independently of whether the document is subsequently read. The second function is portability of an argument. In a purchase involving several participants, one of them needs something they can forward internally, and a document constitutes a portable version of the seller’s position whose task is to survive being read by a person the seller has never spoken to. The third function is as a forcing function on the seller’s own thinking, since writing twenty defensible pages about a problem reveals which parts of a position cannot be supported, and this value accrues whether or not anybody downloads the result. The fourth function is durability, since a document can be cited, updated and referenced in a proposal several years after publication in a way that a social post cannot. What the format is not is an instrument for generating lead volume, because the statistics constructed to support that case do not survive being traced to their sources, and because a download is not evidence of reading. The measurement that follows is not the download count but the number of occasions on which the document appears in an observable conversation, whether mentioned during a call, forwarded within a prospect organisation, or referenced in a reply. That number is smaller, it is countable, and it corresponds to the outcome the document was produced to achieve, whereas the download count measures only that a file transfer occurred.Four jobs, none of them lead volume1. It filtersTrading a work address for twenty pages on anarrow problem is itself a signal.2. It travelsSomebody internal needs something to forward.Its job is to survive a reader you never met.3. It forces your own thinkingTwenty defensible pages reveal which parts ofyour position you cannot support.4. It lastsCitable, updatable, and referenced in a proposalthree years later.Notice that three of the four work whether or not anybody reads itWhich is convenient, given that nobody has published a measurement of whether anybody does.So measure the thing that corresponds to the jobNot downloads. How often it is mentioned on a call, forwarded internally, or referenced in a reply.
Three of them work whether or not it is read. The fourth is what to measure instead of downloads. Source : Method (2026)

What to do with this

Stop reporting downloads as a result. Count instead the number of times the document is mentioned by someone you are talking to, which is a smaller number and a real one.

Ask your own sales conversations whether anyone has read it. That takes a week and produces better evidence than any published benchmark, because it measures your document rather than an average.

If you are deciding whether to produce one, decide on the three jobs above: does it filter, does it travel, and does writing it improve your position. If the answer to all three is no, the statistics were never going to save it.

And when a proposal or an agency cites the 57 percent or the six-to-ten committee, ask for the source. One of them has a spread of 55 points behind it, and the other was deleted by the organisation that published it.

The related pieces are editorial line for B2B and the real cost of social selling.