Ask The Winner Who Was First

The ninety-seventh article was about what happens when you cut a continuum into categories. This one is about what happens when you let the people being studied decide which category they are in.

The Verdict, Stated First

Five claims, in descending order of confidence.

One. The abstract states that almost half of market pioneers fail, and that their mean market share is much lower than earlier studies found.

Two. It names three limitations in that earlier work, and the third, self-classification by the firms themselves, is the one this article is about.

Three. Early market leaders enter an average of 13 years after pioneers and have much greater long-term success, which is in the abstract and inverts the practical advice completely.

Four. On our own arithmetic, excluding failures inflates a mean by a factor of about 1.9 if roughly half fail, which is a large but not sufficient effect.

Five. And on our own arithmetic, letting surviving firms classify themselves can generate the entire finding, with no real advantage present at all.

The fourth and fifth together are the point, ours. Two separate design faults each capable of producing this result, running in the same direction, which is why the earlier literature was so consistent.

Two of the five come from a verified abstract and three from arithmetic, ours. The fifth is the one we would most want a reader to keep, because it names a fault that recurs in material a business owner reads every week.

Our Grades For These Claims

Applying the scheme from the first article in this series, and the grading here is unusually split.

Grade A for the abstract, obtained verbatim and confirmed identically against four independent records.

Grade D for the specific numbers, since the abstract contains none of them. The figures of 47 percent, 10 percent and 30 percent that circulate come from summaries we grade poorly, and we mark every calculation that uses them.

Grade A for our own arithmetic, which is checkable and which we have separated into calculations that need only the abstract and calculations that need the weaker figures.

Grade D for the prior literature, which we did not obtain and which we describe only through the reanalysis's characterisation of it.

That second grade shapes the whole article, ours. The most quoted numbers about this paper are not in its abstract, and we have built the argument so it survives their being wrong.

We could have quietly used the figures and nobody would have noticed, ours. Marking them is the difference between reporting a paper and reporting what people say about a paper, and after ninety-eight articles we know how often those diverge.

A Note On Method

Everything here is verified to August 2026.

We obtained the abstract verbatim from the journal's own record[1] and confirmed it identically against a working paper repository[2], an academic index[3], and a research network[4].

We did not obtain the paper's full text, so we cannot describe its historical method, its categories, or its statistics beyond what the abstract states.

The specific percentages come from an automated summary on a paper-sharing site[5] and a student study aid[6], both flagged as weak wherever they appear.

We obtained none of the prior studies whose limitations are at issue, so we take the reanalysis's characterisation of them on trust.

All arithmetic is ours. The self-report model and every figure in it are invented to demonstrate a structure.

We separate two kinds of calculation throughout, ours. Those that need only the verified abstract, and those that need figures we could not check, and each is labelled where it appears.

This article discusses research on market entry and is not business, strategy or investment advice.

The Claim As Used

What the idea does before we look at what supports it.

The abstract describes the prior position: "Several studies have shown that pioneers have long-lived market share advantages and are likely to be market leaders in their product categories."[1]

Four observations, ours.

In business it becomes a reason to hurry. Ship before you are ready, because the window closes and the first entrant keeps the category.

It is expensive advice, which is why it matters. Speed is bought with quality, cash and staff, and a firm rushing a launch is spending real money on the strength of this claim.

Two things get spent that never appear on the invoice, ours. Scope, which is cut to hit a date, and staff goodwill, which is drawn down by the schedule and repaid later or not at all.

And it has an unusual property for a business belief. It is supported by named academic studies rather than by anecdote, which is what gives it standing in a boardroom.

Which makes the reanalysis worth reading closely, since it is about those studies rather than about the belief.

And that is an unusual target, ours. Most of this series examines a finding; this paper examines a method, which is why its three named limitations are more useful than its headline result.

The Paper

The source.

Golder, P. N., and Tellis, G. J. (1993), Pioneer Advantage: Marketing Logic or Marketing Legend?, Journal of Marketing Research, 30(2), 158–170, May 1993, DOI 10.1177/002224379303000203[1][3].

The abstract states the design: "The authors of this study use an alternate method, historical analysis, to avoid these limitations. Approximately 500 brands in 50 product categories are analyzed."[1]

Four observations, ours.

The title asks a question rather than asserting an answer, which is unusual and which we read as a sign the authors expected an argument.

Roughly 500 brands across 50 categories is a substantial undertaking for historical work, which is slow and cannot be automated.

The method is the interesting part. Historical analysis means going and finding out who was actually first, rather than asking anybody, and the paper's design is a response to a measurement problem rather than to a theoretical one.

And we did not obtain the paper, so we cannot describe how categories were chosen, how first entry was established, or how failures were identified.

That gap is larger here than in most of our articles, ours. A historical method lives or dies on its procedural detail, and we have the claim that it was used and none of the detail that would let anybody assess it.

Three Limitations, Not One

The sentence that organises this article, quoted in full.

The abstract states that the prior research "has potential limitations: the reliance on a few established databases, the exclusion of nonsurvivors, and the use of single-informant self-reports for data collection."[1]

Four observations, ours.

The second is survivorship, which this series has treated at length in its fifty-eighth article and which we will not re-derive here.

The first is a sampling problem. A few databases means a particular set of firms, assembled for a particular purpose, and whatever is systematic about who is in them is systematic in the finding.

The purpose matters more than the size here, ours. A database assembled to support strategic benchmarking contains firms that bought strategic benchmarking, which is a specific and prosperous population.

The third is the one we have not seen discussed anywhere, and it is the subject of most of this article: the classification came from the firms.

And the three are not alternatives. They are three separate faults in the same body of work, each capable of producing the reported result on its own.

One observation on how they are usually reported, ours. Almost every account of this paper we have encountered mentions only survivorship, which is the familiar one, and drops the other two.

That is the pattern this series has now found ninety-eight times. The part of a source that is already understood travels, and the novel part does not, even when the novel part is in the same sentence.

The Numbers, And Their Sourcing

A section on evidence quality, because the figures everybody quotes are not where people think.

The abstract says only that "almost half of market pioneers fail and their mean market share is much lower than that found in other studies."[1] It gives no percentages at all.

The widely repeated figures are a 47 percent failure rate, a mean pioneer market share of 10 percent, and a prior figure of 30 percent. These come from an automated summary[5] and a student study aid[6].

Four observations, ours.

We have not verified any of the three against the paper, and a reader should treat them accordingly.

We tried, ours, and the paper sits behind a paywall we did not pass. That is an ordinary limit on this kind of work and we would rather report it than imply a reading we did not do.

They are consistent with the abstract, which counts for something. Forty-seven percent is "almost half", and ten against thirty is "much lower."

Consistency is not verification. A summary written from the paper would be consistent; so would one written from another summary, and we cannot tell which we have.

So we have split our arithmetic in two. Some of it needs only the abstract and some needs these figures, and we say which at the head of each calculation.

The split has a further use, ours. A reader who distrusts the weak figures can discard one section and keep the rest, which is not usually possible with an argument and is worth designing for.

What Excluding Failures Does

Our own arithmetic, using only the abstract's "almost half."

Suppose failed pioneers end with roughly zero market share, which is close to what failing means. A study that measures only survivors then reports the survivors' mean, while the true mean across all pioneers is lower by the surviving fraction.

The inflation factor is one divided by the surviving share. If 40 percent fail, the factor is 1.67. At 47 percent, it is 1.89. At 50 percent, it is 2.00.

Four observations.

So excluding failures roughly doubles the apparent reward, at the failure rate the abstract implies.

The effect is entirely mechanical and requires no bad faith. A database of operating firms contains operating firms, and nobody has to decide to exclude anybody.

Our assumption that failed pioneers end at zero share is ours and is favourable to the earlier studies. If failures retained some share, the inflation would be smaller.

Being wrong in that direction is the safe error, ours. An assumption that understates our own case is one a critic cannot use against us, which is why we chose it.

And this much of the story is familiar, which is why we have kept it short. The interesting question is whether it is enough, and we can check.

We would note the size of the effect anyway, ours, because it is easy to underrate. A factor of two on a headline strategic finding is enormous, and would on its own change what a firm should be willing to spend.

Does Survivorship Explain It All

Our own arithmetic, and this one relies on the weakly sourced figures, which we flag before giving the result.

If the reanalysis found a mean of 10 percent including failures, and roughly 47 percent of pioneers failed, then a study of survivors alone should report about 10 divided by 0.53, which is 18.9 percent.

But the earlier figure was 30 percent, not 18.9.

So survivorship accounts for a factor of 1.89, and a further factor of 1.59 remains unexplained by it. In logarithmic terms, survivorship is about 58 percent of the total gap.

Four observations.

Survivorship is the largest single contributor and it is not the whole story, on these figures, which is the result we did not expect when we started.

The remainder has to come from somewhere, and the abstract names two other candidates. The databases and the self-reports are the available explanations.

We would state the caveat clearly. This decomposition rests on three numbers we could not verify, and if any is wrong the split moves.

What does not move is the direction of the argument. The convenient story, that this was all survivorship, is not supported even by the figures its proponents quote, and that holds under any reasonable variation of them.

Two ways to see that it is robust, ours. Survivorship would need a failure rate near 67 percent to close the whole gap, which is well above "almost half."

And the failure rate would have to be higher still if failed pioneers retained any market share at all, which they surely did on the way out.

The Limitation Nobody Discusses

The third fault, stated plainly. Ours.

The abstract's phrase is "single-informant self-reports for data collection."[1]

Four observations.

Unpacked, that means somebody at the firm was asked, and their answer became the data. One person, at one company, describing that company.

For most variables that is merely noisy. For this variable it is structurally fatal, because the thing being asked about is a historical fact the respondent has an interest in.

And the interest runs in a specific direction. A firm that leads its category has every reason to describe itself as the one that created the category, and no reason to research the question carefully.

So the classification is partly determined by the outcome being predicted, which is the condition under which a study cannot fail to find what it found.

Two things distinguish this from ordinary measurement error, ours. Ordinary error is symmetric and weakens a finding, pushing an observed association toward zero.

This error is asymmetric and strengthens one, because it moves cases in one direction only, which is why it can create an association rather than merely obscure one.

Manufacturing An Advantage From Zero

Our own arithmetic on an invented model. Every figure is ours, and the model exists to demonstrate a structure rather than to estimate anything.

Take 1,000 surviving firms. Suppose 10 percent are true pioneers and 20 percent are current market leaders, and suppose pioneering has no effect whatever on leading: the two are independent, and the true association is exactly zero.

Now let leaders who did not pioneer claim pioneer status at some rate, because the story is flattering and the definition is loose.

At an over-claim rate of zero, claimed pioneers lead at 20.0 percent and others at 20.0 percent: a lift of 1.00, which confirms the model behaves. At 5 percent: 26.6 against 19.2, a lift of 1.39. At 10 percent: 32.2 against 18.4, a lift of 1.75. At 20 percent: 41.2 against 16.7, a lift of 2.47. At 30 percent: 48.1 against 14.9, a lift of 3.23.

Four observations.

At a 20 percent over-claim rate, pioneers appear two and a half times more likely to lead, from a true association of exactly zero.

The mechanism works from both ends at once, which is why it is so powerful. Every false claim adds a leader to the pioneer group and removes one from the comparison group, so both rates move.

That double action is what distinguishes it from ordinary contamination, ours. A misclassification that only added cases would dilute the comparison group far more slowly, and would produce a much weaker artefact.

Twenty percent is not an extreme assumption. It means one leader in five, among those who did not actually pioneer, describes itself as having done so, which anybody who has read company histories will recognise.

And the lower rows are not reassuring either, ours. Even a five percent over-claim, one leader in twenty, produces an apparent thirty-nine percent lift from an association of exactly zero.

And the figures are invented while the structure is not. Any self-classification correlated with the outcome produces this, at a magnitude set by how correlated it is.

One check we ran on our own model, ours. At a zero over-claim rate it returns exactly 1.00, which is what a correct construction must do when the contamination is removed, and is why we trust the rest of the column.

Why A Firm Would Over-Claim

Not dishonesty, mostly. Ours.

Four observations.

The person answering usually does not know. A marketing director in 1985 answering about a category founded in 1930 is recalling company folklore, not consulting an archive.

And company folklore is written by the survivors, so the internal history of a leading firm genuinely says it created the market, and the respondent is reporting accurately what they believe.

The definition is also genuinely ambiguous, which we come to next. First to sell, first at scale, first with the modern form, and first to name the category are four different firms in many markets.

So the over-claim is best understood as a sincere answer to an underspecified question, which is worse than dishonesty because no amount of integrity fixes it.

Which suggests where to look for this fault generally, ours. Not at questions people might lie about, but at questions people cannot answer accurately, and where the inaccuracy has a direction.

Who Counts As First

The definitional problem underneath. Ours, with one weakly sourced note.

A study aid records that the definition of pioneer used in the main database was inconsistent with the term's use by researchers[6]. A student study aid, our weakest source, flagged.

Four observations, ours.

The candidates for first entrant are numerous and defensible. First to have the idea, first to sell one, first to sell at volume, first with the form the market settled on, and each picks out a different firm.

Which one you choose determines the answer to the research question. Defining the pioneer as the firm that established the category's modern form nearly guarantees a pioneer advantage, because establishing the form is most of what leading is.

That is circularity rather than bias, and it is worse. A biased measurement can be corrected; a circular definition cannot be, because there is no independent quantity to correct toward.

And the remedy is the one the reanalysis took. Fix the definition in advance and go find out who met it, which is slow, unglamorous, and the only thing that works.

There is a version of this available to any firm, ours. Write the definition down before you look at the data, which costs nothing and prevents the definition drifting toward the answer you find.

Thirteen Years Late

The finding that changes the advice, and it is in the abstract.

It states: "Also, early market leaders have much greater long-term success and enter an average of 13 years after pioneers."[1]

Four observations, ours.

So the successful position on this finding is not first, and not late either, but a category the paper calls early market leaders who arrive more than a decade behind.

Thirteen years is much longer than any business planning horizon, which makes the finding hard to act on directly and easy to misread as an argument for delay.

We would read it more carefully than that, ours. The finding describes when successful firms happened to enter, not a recipe for when to enter, and those differ.

And it is a strong hint about mechanism. If arriving thirteen years late is compatible with dominating a category, then whatever produces dominance is not timing, which is the paper's real argument.

Two candidates the paper's abstract does not name, ours and offered as reasoning. A later entrant sees a market that already exists, so it is solving a demonstrated problem rather than a hypothesised one.

And it can copy what worked and skip what did not, which is a real advantage the pioneer paid to generate.

If that is the mechanism, the strategic implication is uncomfortable and clear, ours. The pioneer is funding the market research that its competitors will use, and doing so at full price.

What They Did Instead

The design response, and it deserves credit. Ours.

Four observations.

The abstract calls the method historical analysis, used explicitly "to avoid these limitations"[1], so the method was chosen to fix a measurement problem.

That is the right response to all three faults at once. Archival work includes firms that no longer exist, is not limited to a database's membership, and does not ask anybody to classify themselves.

It is also enormously more expensive, which is why it is rare, and why the earlier literature took the route it did.

And we would note the general principle, ours. When a variable is contested, the measurement has to be done by somebody with no stake in the answer, and that person is usually not in the industry.

Which is uncomfortable for a great deal of commercial research, ours. Most industry data is collected by parties within the industry, for reasons of access and cost, and the access is exactly what creates the stake.

Two Faults Pointing The Same Way

Why the earlier literature was so consistent, which needs explaining. Ours.

Four observations.

A field producing a wrong answer repeatedly is usually taken as evidence the answer is right, and here the consistency is itself the thing to explain.

Both faults push in the same direction. Excluding failures raises the apparent reward, and self-classification raises the apparent association, so studies sharing either would agree.

And studies sharing a database, which is the first limitation named, share both faults at once. Independent research groups analysing the same source are not independent tests.

So replication offered no protection here, ours. Repeating an analysis on the same data with the same design reproduces the artefact faithfully, which is what agreement between such studies demonstrates.

That is worth generalising, ours. Agreement between studies is evidence only to the extent that their designs differ, and several studies sharing a database and a method are closer to one study than to several.

What An Honest Study Requires

Because it is fair to say what would satisfy us. Ours.

Four features.

It would define first entry in advance, precisely enough that two researchers applying the definition to the same category would name the same firm.

It would identify entrants from records rather than from respondents, so no firm's account of itself enters the classification.

It would include firms that no longer exist, which requires archival work because no current database contains them.

And it would be slow and expensive, which is the recurring finding of this whole series. The reanalysis appears to have done exactly these things across roughly 500 brands, and that is why it took the form it did.

One further thing follows, ours. Nobody would fund such a study to confirm an existing belief, so this kind of work only ever gets done by somebody who suspects the belief is wrong, which introduces its own selection.

What Hurrying Actually Costs

Our own arithmetic on an invented firm, because the belief has a price and nobody computes it.

Suppose a launch can be brought forward six months by adding staff and cutting scope, at an additional cost of $180,000, and that doing so raises the chance of being first from 30 percent to 70 percent. Every figure here is ours.

The extra 40 percentage points of pioneer probability cost $4,500 per point.

Four observations.

For that to pay, being first must be worth at least $450,000 in expected terms, which is a number the firm has never estimated.

And on the abstract's own finding, almost half of pioneers fail, so the position being purchased carries its own large downside.

The rushed scope cut is not free either, and does not appear in our figure. Whatever was removed to hit the date is absent from the product a customer sees, and that cost lands later.

So the honest framing is a wager rather than a strategy, ours. The firm is paying a known amount for a probability of a position whose value it has not established, which is a defensible thing to do only once the numbers are written down.

And writing them down usually changes the decision, ours. A four hundred and fifty thousand dollar threshold is a specific claim a management team can argue about, where "we need to be first" is not.

What Actually Survives

Our reading, stated directly.

Five statements.

Almost half of market pioneers fail, on the abstract, and their mean share is much lower than earlier studies reported.

Three limitations were named, not one, and two of them concern who got counted rather than how they performed.

Early market leaders enter about 13 years after pioneers and do much better, which is in the abstract and is the practically important sentence.

On our own arithmetic, excluding failures inflates the reward by roughly a factor of two at the implied failure rate.

And on our own arithmetic, self-classification alone can produce a two-and-a-half-fold apparent advantage from nothing, which is a mechanism we have not seen discussed.

Those five are what we would defend, ours, and the first three come from an abstract confirmed against four records, which is the sourcing position we prefer to be in.

Not An Argument For Being Slow

The obvious misreading, addressed. Ours.

Four observations.

Nothing here says entering late is better. The finding is that pioneering does not confer the advantage it was credited with, which leaves the question of timing open rather than reversing it.

There are real reasons a first entrant can do well, and the paper does not deny them. Learning, supplier relationships and a head start on scale are genuine, and the argument is about magnitude rather than existence.

What changes is what a firm should be willing to pay for speed, which was the practical content of the belief.

And that is a number rather than a posture, ours, which is the useful shift. A firm can decide it will spend up to some amount to be first, and that decision can be examined.

And the failure rate is the number to hold on to. If almost half of pioneers fail, being first is a risk position as much as an advantage, and it should be priced like one.

Who Is Not In The Room

The two faults restated as one question a firm can ask. Ours.

Four observations.

Survivorship and self-classification are both answered by asking who is not in this dataset, and would their absence change the answer?

For the pioneer literature the answer was concrete. Roughly half the pioneers were absent, and their absence changed the reported reward by about a factor of two.

The question generalises to any evidence a firm relies on. Which customers, staff, competitors or periods are missing from this, and are they missing for a reason connected to what I am trying to learn?

The second half of that question does the work, ours. Random absence is merely a smaller sample; systematic absence is a different sample, and only the second changes the answer.

And it is answerable more often than people expect, ours. A firm usually knows roughly how many of its own customers churned, so it can estimate what a survey of current customers is leaving out.

Two habits make this routine rather than heroic, ours. Record who was invited to respond, not only who did, which most survey tools already capture and most firms never look at.

And keep a list of the accounts that left, which is the population every satisfaction measure silently excludes.

That list is the cheapest research asset a small firm owns, ours, and almost nobody keeps it. The customers who left have already decided, so they have nothing to gain by being polite, which makes them the most informative people you could ask.

The General Form

The transferable lesson, ours, and it is the one worth keeping.

Four observations.

Ask who assigned the categories. If the subjects assigned themselves, and the assignment could be influenced by the outcome, the study cannot distinguish the finding from the artefact.

This is a different fault from survivorship and is frequently present alongside it. Survivorship is about who is missing; self-classification is about who is mislabelled, and a study can suffer both.

The test is a single question. Could a respondent's answer have been different if their outcome had been different? If yes, the classification is downstream of the result.

And it is remarkably common outside academia. Any survey asking firms about their own strategy has it, since a successful firm describes its past differently from a struggling one.

Two more places it appears, ours. Employee engagement surveys, where people who are leaving answer differently or not at all.

And customer satisfaction data, where the customers who left are the ones whose views would have been most informative and are precisely the ones not asked.

Awards, Lists And Rankings

The commercial form most owners meet weekly. Ours.

Four observations.

A best-workplaces list, a fastest-growing ranking or an industry award is compiled from firms that entered, and entering is a decision correlated with expecting to do well.

So the population is doubly filtered. The firm must still exist and must have chosen to be measured, which is survivorship and self-selection stacked in the same direction.

The consequences run both ways and are worth separating. Reading such a list to see what good firms do is unsound, because the comparison group is missing entirely.

And appearing on one is weak evidence about your own firm, since the achievement is partly a fact about who else entered, which is a thing you cannot see.

We would not tell anybody to refuse the award, ours. It has real value as a signal to customers and staff, and that value does not depend on it being a valid measurement.

What This Article Cannot Tell You

Stated plainly, because the gap here is larger than usual. Ours.

Four things.

Whether the reanalysis is right. We obtained its abstract and confirmed it four times, which establishes what it claims and not whether the claim holds.

Whether the prior studies were as described. We have one side's characterisation of another body of work and did not read that work.

Whether any replication exists. We did not search for one, and a 1993 paper has had thirty years in which to be tested.

That gap is the one we would most like closed, ours, since a reanalysis that has stood unchallenged for three decades means something quite different from one that has been disputed.

And whether the specific percentages are correct, which we have said repeatedly and repeat here because the decomposition section depends on them entirely.

We list these because the alternative is a confident article, ours. Four unanswered questions is what an honest reading of one abstract leaves, and pretending otherwise would be the failure this series was built to examine.

Your Own Benchmarks

Where this reaches a small firm directly. Ours, and not business advice.

Four points.

Industry benchmark surveys are self-reported, almost without exception, and they are answered by firms that still exist and chose to respond.

Both faults apply at once. The failures are absent and the respondents are describing themselves, which is the exact pair the reanalysis named.

The direction is predictable and worth stating. Reported margins, growth rates and win rates will run above the true distribution, so a firm comparing itself to them will conclude it is behind when it may not be.

And the remedy is available. Ask what the response rate was and whether failed firms are represented, and if the survey cannot say, treat the figures as a description of respondents rather than of your industry.

One number makes this concrete, ours. A benchmark survey with a twenty percent response rate is describing a fifth of an industry, chosen by themselves, and that is usually disclosed in a footnote nobody reads.

Every Case Study You Read

The same fault in its most common form. Ours.

Four observations.

A business case study is a successful firm describing its own past, which is single-informant self-report with the outcome already known.

The narrative is assembled backward from the result, so whichever decisions preceded success become the strategy, and the ones that preceded nothing are not recalled.

We are not accusing anybody of dishonesty, and would emphasise that. Memory genuinely reorganises around outcomes, and the executive telling the story believes it.

Two features make the reorganisation invisible to the person doing it, ours. The decisions that preceded success are the ones that get retold, so they are rehearsed and the others are not.

And the retelling is refined by audience response over years of conference talks, which selects for a coherent narrative rather than an accurate one.

And the practical consequence is the same as for the benchmark surveys. Read a case study as a description of what one firm believes about itself, which is interesting, and not as evidence about what works.

We would add one thing in their defence, ours. A case study is often the only account of an unusual situation that exists, and reading it for detail rather than for inference is a legitimate use.

Bibliographic Note

The series keeps a count, and this article produced three.

A repository lists the authors as Tellis then Golder[2], where the journal and three other records give Golder then Tellis[1][3][4].

The same repository gives the location as page 158 alone[2], where an index gives 158 to 170[3].

And that index gives volume 30 with no issue number[3], where the journal's record is 30(2)[1].

Three observations, ours.

The reversed author order is the most consequential variant of this kind we have recorded, since first authorship is a claim about contribution and a reader searching by first author would miss the paper.

The other two are ordinary, and the start-page-only form is the more troublesome of them, ours, since a reader cannot tell from it how long the paper is and therefore cannot tell whether a summary they are reading covers it.

That brings the running count of bibliographic variants across this series to fifty-two.

Fifty-two across ninety-eight articles is roughly one every two, ours, and we did not go looking for them. They are simply what turns up when you check more than one record for the same paper, which is a finding in itself about how citations travel.

What To Do

Do not pay a premium for being first on the strength of this literature. The abstract reports that almost half of pioneers fail and that their mean share is much lower than earlier work found.

Note the three limitations, not just survivorship. The databases, the exclusion of failures, and the self-reports are three separate faults each capable of producing the result.

Ask who assigned the categories in any study you rely on. If the subjects classified themselves and the classification could depend on the outcome, the finding cannot be separated from the artefact.

Treat almost-half failure as a risk figure. Being first is a position with a large downside, and should be priced rather than assumed to be an advantage.

Remember the thirteen years. Early market leaders on this finding entered more than a decade after the pioneers and did much better, which means timing is not the mechanism.

Discount self-reported industry benchmarks. They exclude the failures and are answered by interested parties, which is the same pair of faults.

Read case studies as self-description. A successful firm narrating its own past is the exact design the reanalysis was written to avoid.

And check whether the numbers you are quoting are in the abstract. The three most repeated figures about this paper are not, and we could not verify any of them.

The Limits Of This Analysis

Several caveats matter. This article discusses research on market entry and is not business, strategy or investment advice. Everything is verified to August 2026. We did not obtain the paper's full text, only its abstract, which we confirmed verbatim against four independent records; we therefore cannot describe its historical method, how it defined or established first entry, how it identified failures, how categories were selected, or any statistic it reports. The specific figures of 47 percent, 10 percent and 30 percent appear nowhere in the abstract and reach us from an automated summary on a paper-sharing site and a student study aid, both of which we grade poorly; we have not verified any of them, and our decomposition section rests entirely on all three being correct. We obtained none of the prior studies whose limitations are at issue, so we describe them only through this paper's characterisation, and a reader should note that this is one side's account of another body of work. The note that the main database's definition of pioneer was inconsistent with researchers' usage comes from our weakest source, flagged. All arithmetic is ours. The survivorship inflation calculation assumes failed pioneers end with zero market share, which is our assumption and is favourable to the earlier studies. The self-report model is entirely invented: the thousand firms, the ten percent pioneer rate, the twenty percent leader rate, the independence assumption and every over-claim rate are ours, chosen to demonstrate a structure rather than to estimate any real quantity, and no reader should take the 2.5-fold figure as an estimate of what occurred. Our claim that self-classification has not been discussed elsewhere reflects our own reading and no systematic search. And our sections on benchmark surveys and case studies are our own reasoning rather than findings we obtained.

Related Reading

No. 99 They Stopped Talking To Each Other Two field studies measured face-to-face interaction before and after firms went open plan, using wearable sensors rather than surveys. It fell by about... Behavioural Finance Fourteen Percent Come Back Someone Else What else goes wrong when a category is assigned rather than measured. Behavioural Finance The Author Wrote It Down Another case where the design, not the data, produced the finding.

Frequently Asked Questions

Is first-mover advantage real?

On the abstract we obtained, almost half of market pioneers fail and their mean market share is much lower than earlier studies found. That does not say entering late is better; it says pioneering does not confer the advantage it was credited with, and the failure rate makes being first a risk position as well as an opportunity.

What were the three limitations?

In the abstract's words: the reliance on a few established databases, the exclusion of nonsurvivors, and the use of single-informant self-reports for data collection. The second is survivorship, which is well known. The third means firms were asked to classify themselves.

Why does self-classification matter so much?

Because the classification can depend on the outcome. On our own invented model, if one leader in five who did not pioneer describes itself as having done so, pioneers appear two and a half times more likely to lead, from a true association of exactly zero. Each false claim adds a leader to one group and removes one from the other.

Does survivorship explain the whole difference?

On our own arithmetic using figures we could not verify, no. Survivorship accounts for roughly 58 percent of the gap in logarithmic terms, leaving a further factor of about 1.6 to be explained by the other two limitations. That was not the result we expected.

What is the thirteen-year finding?

The abstract states that early market leaders have much greater long-term success and enter an average of 13 years after pioneers. If arriving more than a decade late is compatible with dominating a category, then whatever produces dominance is not timing.

Are the 47 and 10 percent figures reliable?

We could not verify them. They appear nowhere in the abstract and reach us from an automated summary and a student study aid. They are consistent with the abstract's wording, but consistency is not verification, and we have marked every calculation that depends on them.

How does this affect industry benchmarks?

On our own reasoning, industry benchmark surveys have both faults at once: the failed firms are absent and the respondents are describing themselves. Reported margins and growth rates will therefore run above the true distribution, so a firm comparing itself to them may conclude it is behind when it is not.

References

1. Journal record for Golder, P. N., & Tellis, G. J. (1993), Pioneer Advantage: Marketing Logic or Marketing Legend?, Journal of Marketing Research, 30(2), 158–170, DOI 10.1177/002224379303000203, reproducing the abstract: that several studies have shown that pioneers have long-lived market share advantages and are likely to be market leaders in their product categories; that this research has potential limitations, being the reliance on a few established databases, the exclusion of nonsurvivors, and the use of single-informant self-reports for data collection; that the authors use an alternate method, historical analysis, to avoid these limitations; that approximately 500 brands in 50 product categories are analyzed; that the results show that almost half of market pioneers fail and their mean market share is much lower than that found in other studies; and that early market leaders have much greater long-term success and enter an average of 13 years after pioneers. Note: the journal's own record and the source of every quotation in this article. We obtained the abstract only, not the full text. journals.sagepub.com
2. Working paper repository record for the same article, giving the citation as Journal of Marketing Research, Vol. 30, No. 2, p. 158, May 1993, and reproducing the abstract identically. Note: a repository record, used to confirm the abstract independently. Recorded also as the source of two bibliographic variants: it lists the authors in the order Tellis then Golder, and gives the location as a single start page. papers.ssrn.com
3. Academic index record for the same article, giving Journal of Marketing Research, 1993, volume 30, pages 158 to 170, and reproducing the abstract identically. Note: an indexing service, used as a second independent confirmation of the abstract and as our source for the full page range. Recorded also for giving a volume with no issue number. semanticscholar.org
4. Research network record for the same article, listing both authors and reproducing the abstract identically, including the three named limitations and the finding on early market leaders. Note: a research network, used as a third independent confirmation of the abstract. researchgate.net
5. Paper-sharing platform page for the same article, carrying an automatically generated summary stating that pioneers have a 47 percent failure rate, that the mean market share of pioneers is 10 percent, substantially lower than the 30 percent reported in previous studies utilising a particular database, and that early leaders entering markets almost 13 years after pioneers tend to exhibit higher market share and success rates. Note: an automatically generated summary on a paper-sharing platform, NOT the paper itself and NOT an academic source, flagged at every use. The source of the specific percentages quoted in this article, none of which we verified against the paper and none of which appear in the abstract. academia.edu
6. Student study aid summarising the paper, stating that the failure rate of market pioneers is shown to be 47 percent, that the failure rate is twice as high for durable as for non-durable goods, that the definition of pioneer in the main database is inconsistent with the term's use by researchers, and that the study's objectives included estimating the rewards of pioneers after controlling for survival bias by studying both successful and unsuccessful pioneers, demonstrating historical analysis as a method, and providing an objective measure of the true pioneer or first entrant in each product category. Note: a commercial student study aid. This is the WEAKEST SOURCE IN THIS ARTICLE, flagged at every use. Cited only for the database definition point and as a second occurrence of the 47 percent figure. Nothing load-bearing rests on it. studysmart.ai

This article discusses research on market entry and is not business, strategy or investment advice. The paper's full text was not obtained, only its abstract, confirmed verbatim against four records. The specific percentages quoted appear nowhere in that abstract and come from sources graded poorly here. All arithmetic is the authors' own, and the self-report model is entirely invented to demonstrate a structure rather than to estimate any real quantity.