We went looking for one market size and found a report mill

We were building a page about rebound boots, the spring-loaded fitness boots used in bounce classes, and we wanted to open it with something ordinary: how popular is this? Not a precise figure. A defensible order of magnitude, with a source we could name.

We could not find one. Not because the answer is hidden behind a paywall, but because the question has been answered so many times, so confidently, by so many pages that contain no measurement at all, that the real answer no longer has anywhere to sit. This is what that looks like up close, and why we think it is a data-acquisition problem rather than a fitness one.

What the first page actually contained

Searching for the market size of this category returns, near the top: theinsightpartners.com, archivemarketresearch.com, futuremarketreport.com, imrmarketreports.com, a Google Sites page built for the keyword, and a GitHub repository of scraped report stubs. Every one of them offers a market size and a compound annual growth rate through 2032 or 2033. They disagree with each other, and none of them says where any figure came from.

The tell is not subtle. One of those reports is titled, verbatim:

"Kangoo Jump Shoes Report Probes the XXX million Size, Share, Growth Report and Future Analysis by 2033"

A second, from the same publisher, is titled "Professional Kangoo Jump Shoes Is Set To Reach XXX million By 2033, Growing At A CAGR Of XX". The XXX and the XX are not our redaction. They are the template's placeholders, published as-is, because nothing in the pipeline that produced these pages ever had a number to put there.

The body text underneath does carry figures, which makes it worse rather than better. It has a market value for 2025, a projection for 2033, and a growth rate. It also has a named research analyst with a LinkedIn profile, section headings, and this description of method: "Our rigorous research methodology combines multi-layered approaches with comprehensive quality assurance." No survey, no panel, no sample, no organisation, no year of collection. A human-looking byline over machine-filled text.

The pages are built to rank, not to inform

This is the part worth internalising if you work with data. These pages are not low-effort. They are high-effort in every dimension except measurement. They have the schema markup, the plausible section structure, the confident round numbers, the analyst headshot, the segment breakdowns by region and by end user. Everything a ranking system reads as authority is present. The only absent ingredient is the observation.

And they are mass-produced. The GitHub repository that ranked for our query has since 404'd, but the pattern is mirrored across accounts: we found repositories holding 161, 207, and 21 markdown files, each one a stub for a different market, each linking back to a report vendor with a "Request Sample Report PDF" call to action. A representative file opens "The 5D BIM Market represents a critical advancement in the digital transformation of the construction and infrastructure industries" and proceeds to segment a market it never measures. GitHub is being used as link surface. The search index still returns the dead ones.

We watched the number get laundered

Here is the step that turns a bad page into a bad fact. When we ran the search through a tool that summarises results, it returned a clean, confident answer: roughly $150 million in 2025, roughly $450 million by 2033, a growth rate somewhere between 9.4% and 12.5%. It even accounted for the disagreement between sources, attributing it to "different methodologies and market segments covered by various research reports."

None of those reports states a methodology. The summary is a reasonable-sounding sentence built entirely out of pages with nothing underneath them, and it now reads as sourced, because a system that appears to have checked has repeated it. We are describing one search we ran and observed, not making a claim about how any particular product behaves in general. But the mechanism is not hypothetical, and it only needs to happen once for the figure to start circulating with a citation attached.

What real research on this question looks like

For contrast, there is an organisation that answers questions of this shape properly. The Sports & Fitness Industry Association publishes an annual Topline Participation Report covering Americans aged six and up across 124 activities (126 in the most recent edition, after adding disc golf and padel), weighted to the US population by gender, age, income, ethnicity, household size, region, and population density. It costs about $700 for non-members.

Notice what that has that the free pages do not: a defined population, a stated weighting scheme, a fixed activity list, a year, and a price. We could not establish whether rebound boots appear among its tracked activities without buying the report, and we are not going to assert that they do not. What we can say is that not one of the pages ranking for our question cites the SFIA, or any other named methodology, or any measurement of any kind.

The verifiable data exists. It just loses.

The same category does have real published research. It is small, and it is specific, and it contradicts the marketing.

Two peer-reviewed studies of this equipment appear in the Journal of Bodywork and Movement Therapies. Rossato and colleagues (2017) put ten women through fitness exercises in rebound boots and measured greater hip and knee extension on landing, with less lateral gastrocnemius activation. De Britto and colleagues (2021) measured impact forces across fifteen women and found peak impact roughly 20% lower on jump landings and about 11% lower running, along with a longer time to peak force and slight asymmetries between legs.

Manufacturer marketing for this category commonly claims up to 80% less impact. The measured figures are 20% and 11%. Both studies are small, both used only female participants, and neither is the last word on anything. We say that plainly because the argument of this article collapses if we do not hold ourselves to it.

So the record is not empty. It contains two careful, checkable studies with sample sizes in the low teens, and it contains an unbounded quantity of synthetic pages that rank above them. That is the whole problem in one sentence: verifiable data exists and loses to fabricated data on search.

What we did instead

We refused the number. The page we were building says that no trustworthy public figure for this category's size exists, and explains why, rather than borrowing a plausible one from a report mill. That is the page where we worked this through: a breakdown of rebound-boot class formats on MyBootLab, a site we operate. It is written in Spanish. We are pointing at it as our own working, not as an independent source, because it is neither.

The general lesson is the one we keep running into building data products: for any question where the answer is commercially useful and expensive to measure, assume the free corpus has been filled in by something that did not measure it. The defensible move is to say what you could not establish and why. An unsourced number repeated confidently is worse than no number, because the next system to read it will treat the confidence as evidence.

Kangoo Jumps is a registered trademark of Etablissement AMRA. We name it only to identify the specific published pages and product category discussed here.