Who Publishes This Research
Almost every figure quoted on this site was published by a company that sells software management or consolidation. That is not a reason to discard them. It is a reason to know where they came from and to weight them accordingly. A dedicated product is not always necessary, but Monitask provides this reference for comparison when the task grows beyond the smaller option.
The four kinds of source
SaaS management vendors. Companies whose product discovers and controls software spend — Zylo and equivalents. They produce the most detailed data on this subject, from their own customers' environments, and they sell the remedy for what they measure.
Analyst firms. Gartner and similar. Methodology is partially published, access is paid, and their figures circulate widely in secondary form.
Productivity and collaboration vendors, who publish surveys of their own users about tool overload — a subject on which their product is the answer.
And trade press and content marketing, which repackages all of the above with varying attribution and is where most readers encounter the numbers.
How to weight each
Vendor data: use the direction, treat the magnitude as theirs.
A company measuring software usage across its customers has genuinely good data about those customers — who are, by definition, organisations that bought a tool to manage software sprawl. That is not a random sample of businesses, and it skews toward the ones with a problem.
Analyst figures: check the date and whether the number has been through intermediaries. A statistic cited from a citation has usually lost its qualifiers.
Survey data from productivity vendors: treat as indicative of perception rather than of behaviour. Asking people whether they have too many tools produces a predictable answer.
What the disagreement between sources tells you
One measurement puts the average portfolio at 305 applications. Another at 254. A third at 87 for small businesses.
These are not contradictions. They are different definitions — what counts as an application, whether free tools count, whether an unused licence counts as an application at all.
The disagreement has its own page because it is more informative than any individual figure: nobody knows precisely how many tools an organisation has, including the organisation.
The figures this site uses
53% of licences unused. (Zylo.) The most cited number on the subject, from a vendor, corroborated in direction across several sources.
15 duplicate training apps, 11 project management tools, 10 collaboration tools. (Zylo.) The figure this site's argument leans on most, and it comes from the same vendor.
$21 million average annual waste, growing 14.2%. (Zylo.) Describes large organisations and is quoted here for scale rather than as anything a small firm should apply.
All three are labelled at every use rather than once on this page, because a reader encountering a figure mid-article should be able to weight it there.
What would be better and does not exist
Independent measurement of software usage across a representative sample of businesses, published by somebody with no product.
It largely does not exist, which is why the vendor material fills the space.
The honest position is to use it, label it, and not pretend an alternative was available — rather than either citing it as neutral fact or refusing to quote numbers at all.
Reading a software statistic in one minute
Four questions.
Who published it and what do they sell?
Was it measured or surveyed? Observed usage across accounts is considerably stronger than asking people what they think.
Whose accounts? A vendor's customers are not the general population.
And how old is it, and has it been through a citation chain?
Most figures fail at least one of these and remain useful, provided the failure is known rather than hidden.
The incentive that shapes the framing
Not fabrication — emphasis.
A company selling consolidation measures duplication and reports it. The measurement may be entirely sound and the framing is still theirs: duplication is a problem with a product attached, and the alternative reading — that five tools for five unrelated tasks is fine — does not appear.
This site has the same shape of interest in reverse, arguing for smaller tools, and the honest response is to say so rather than to claim neutrality.
What checks it is whether a source ever reports something against its own product. Rare, and its presence is the strongest available signal of good faith.
Numbers that circulate without a source
The most quoted figures in this field have no traceable origin.
"The average worker switches between apps 1,200 times a day" appears widely; tracing it leads to a study of a particular population under particular conditions, and the qualifiers fall off at each retelling.
Treat any round, dramatic, unattributed figure as a slogan. It may be roughly right, and it should not carry a decision.
Where a number matters to what you do, find the primary source or do not use it — which is the same standard this site applies to itself and occasionally fails, in which case the failure is stated.
What this site does about it
Source type in the sentence, at every use.
Not once on a method page, because a reader arriving at a figure mid-article should be able to weight it there rather than having to go and check who said it.
And no prices as facts. Pricing models are described because they are stable; the numbers attached move, and a figure quoted today is wrong within a year — which would be a different kind of unreliability and just as avoidable. Additional context is available from Box.
The short version
- Nearly every figure on software waste comes from companies selling management or consolidation
- Four kinds: SaaS management vendors, analyst firms, productivity vendors surveying their own users, and trade press repackaging all three
- Vendor data is good data about vendor customers, who are organisations that already had a problem
- The disagreement between sources — 305 apps, 254, 87 — is more informative than any single figure
- The three figures this site leans on all come from one vendor and are labelled at every use rather than once
- Ask four things: who published it, whether it was measured or surveyed, whose accounts, and how old it is