Data with integrity

Numbers you will be quoted that are not true

Somebody is going to put one of these in front of you, in a proposal or a pitch deck or a trade article. Three of the four below were built by someone selling the solution to the problem the number describes.

While putting a book together I checked the most commonly repeated statistics in small business writing against their original sources, and a startling number of them do not have one. They are repeated because they are repeated. Some of them are in books you have on your shelf.

This matters more than being right for its own sake. A number that is not true does not just mislead you. It points you at a constraint you do not have — and you spend a quarter, and a budget, on a place your work was never waiting.

The six ways a number goes wrong

Almost everything below fails in one of these ways. Once you can name them you can check a number you have never seen before, which is the actual skill — and it is worth more than any of the corrections.

  • 1No study behind itThe trail ends at a blog, a press release, or a name with no publication attached.
  • 2The question changedTwo figures from the same publisher, years apart, are not the same measurement.
  • 3Wrong denominatorA subgroup figure presented against the whole sample, or two rates computed on different bases and compared.
  • 4A target sold as a measurementSomebody's goal, quoted as an average.
  • 5The vendor measured its own productReal data, no independent check, and a commercial interest in the answer.
  • 6A checklist read as a rankingPercentages that sum to far more than a hundred were a multi-select question, not a breakdown of causes.

Four that do not survive checking

Each carries its source, its sample size and its date — including the replacements, which have limits of their own. A better number is still a number with limits, and pretending otherwise would be the same mistake in the other direction.

82% of businesses fail because of poor cash flow management

Fails: no study behind it, a checklist read as a ranking

The problem

No retrievable study. The citation chain ends at one person's name and a former employer, with no publication, date, sample size or method behind it — every citation leads to another blog. The list it comes from sums to 802%, which means it was a multi-select question and never a breakdown of causes at all.

What is true instead

The Federal Reserve surveys this. 94% of small employer firms reported a financial challenge in the past year, and 57% describe their financial condition as fair or poor.

Small Business Credit Survey, Federal Reserve banks. Not a random sample — firms are recruited through chambers of commerce and development lenders, then weighted to Census figures. The Fed says so plainly. Good for direction and comparison; never describe it as nationally representative.

80 to 90% of new products fail

Fails: no study behind it

The problem

Trade-press repetition of an unsourced claim, examined and rejected in peer review in 2013. It traces to two 1961 articles — one in the Harvard Business Review, one in the Wall Street Journal — neither carrying data.

What is true instead

Nineteen peer-reviewed studies from 1945 to 2004 put post-launch failure at 30–49%. A panel of 83,719 consumer-packaged-goods SKUs launched between 2002 and 2009 found 25% gone within a year.

The SKU panel is consumer packaged goods, which is not every market. The range across nineteen studies is wide because they defined failure differently — which is the honest answer, not a defect.

76% of nearby searches lead to a store visit within 24 hours

Fails: wrong denominator

The problem

The source is real, still online, and does not say this. Google measured that 76% of people who search for something nearby visit a business within a day — any business, not the one they searched for.

What is true instead

For the decision itself: 75% of consumers choose a local business within 30 minutes of starting to look, and more than a quarter inside five minutes.

The original is a purchased diary study, 634 local searchers, 2016. The replacement is self-reported, 1,227 US consumers who searched for a local business in the previous three months, run by a company that sells local search tools.

The typical manufacturer runs at 60% OEE

Fails: no study behind it, the vendor measured its own product

The problem

An assertion on a monitoring vendor's website with no survey, sample or date behind it.

What is true instead

Across more than 3,500 machines the measured distribution peaks at 55–60%, and about 6% exceed 85%.

That distribution is a vendor's own customer base — firms that had already bought monitoring equipment. It is a real measurement of an unrepresentative group, which is a different problem from having no measurement at all.

There are about thirty of these. The rest are in the book, along with the sources for every figure on this page. If one of them is quoted at you, the fastest response is not to argue — it is to ask which study, what sample, and what year.

Data with integrity makes the best decisions

Bring me a number you are not sure about.

Twenty minutes, no obligation. If it holds up I will tell you that too.

Call (918) 200-9699 Book a call