Two software companies have taken Gartner to court over a Magic Quadrant placement. Both lost, and both lost for the same reason: the courts looked at the quadrant and named it something buyers rarely stop to name. It is an expression of opinion, constitutionally protected, and legally incapable of being proven true or false.
I think those rulings are the most under-read documents in enterprise procurement. Not because they expose wrongdoing. They establish the opposite, and that is the interesting part. Gartner defended itself, twice, by arguing that its ratings are subjective judgments no reasonable reader should treat as verifiable fact. It won on that argument, twice. Meanwhile, procurement teams keep pasting quadrant graphics into decision records as if they were audited measurements.
What did the courts actually decide about the Magic Quadrant?
Start with ZL Technologies. In May 2009 the email-archiving vendor sued Gartner in federal court in San Jose over its placement in the "Niche Players" quadrant, bringing five claims: defamation, trade libel, false advertising, unfair competition, and negligent interference with prospective economic advantage. Judge Jeremy Fogel dismissed all five that November. His reasoning cut to the bone: "the general tenor of the MQ Report negates the impression that Gartner is asserting an objective fact." ZL amended its complaint and tried again. In May 2010 the court dismissed the amended complaint as well, holding once more that the placement was non-actionable opinion. That was the end of it.
NetScout went much further. In August 2014 it filed a 58-page complaint in Connecticut state court alleging that Gartner runs a pay-to-play scheme: favorable quadrant positions for vendors who buy its consulting services, punishment for those who decline. NetScout had declined, and it had just been rated a "Challenger" rather than a "Leader" in the 2014 Magic Quadrant for network performance monitoring and diagnostics. It sued under Connecticut's Unfair Trade Practices Act and for defamation. This was no fringe grievance; per the court record, NetScout ran roughly $400 million in annual revenue at the time.
Here is the outcome, stated plainly, because reporting the allegation without it would be malpractice. Gartner won in the trial court in 2017, on summary judgment. NetScout could not substantiate the pay-to-play claim; its own expert witness was unable to conclude that Gartner's ratings correlated with the dollar volume of consulting services vendors purchased. In January 2020 the Connecticut Supreme Court affirmed, holding that every challenged statement was a nonactionable expression of opinion.
Why does Gartner winning matter more than the allegation?
The lazy reading of NetScout v. Gartner is "bitter vendor sues, loses." The useful reading is in how Gartner won.
It did not win by proving the quadrant accurate. Accuracy was never tested. It won by persuading the court that vendor ratings are "inherently subjective", that the criteria behind Completeness of Vision and Ability to Execute, and the weight assigned to them, "could not be proven true or false," and that the report's language was "abstract, unquantifiable, and comparative, such that the statements were insusceptible of being verified." The court went one step further than most people remember. Gartner's own marketing described its research as objective and impartial, and the justices held that such claims of objectivity, "like puffery, are insusceptible of being proven true or false" and unlikely to induce reliance in a reasonable person.
Fairness requires the next part too. The courts found no corruption. The trial court found no evidence that quadrant position correlated with consulting spend. What the summary-judgment record did describe, read in the light most favorable to NetScout, was a structure in which the analysts who provide Gartner's paid advisory services also sit on the teams that place vendors on the graphic, and are evaluated in part on the revenue they generate. Defamation law found that structure irrelevant once the statements were classified as opinion. My view, offered as opinion resting on those stated facts: a buyer making an eight-figure platform decision should at least know the classification question exists.
Both plaintiffs made the same category error the rest of the market makes daily. ZL and NetScout litigated the quadrant as if it were a factual, auditable measurement that could be shown false. Two courts said it is not that kind of object. Follow the logic through: if a placement cannot be false, it cannot be verified, and if it cannot be verified, it belongs in a different column of the decision record than the inputs that can.
How a decision record should weight opinion versus evidence
An opinion-class input is a synthesized judgment whose method you cannot re-run and whose conclusion cannot be proven true or false. An evidence-class input is an artifact you can check: an audit report, a published pricing page, a court filing, an SEC disclosure, a benchmark with a reproducible method, a status-page history. The distinction is not about prestige or usefulness. It is about verifiability, the exact axis the courts ruled on.
| Input | Class | Can it be checked? | Proper weight |
|---|---|---|---|
| Quadrant or wave placement | Opinion | No, by court ruling | Orientation and context |
| Vendor capability claim | Claim | Not until graded | Zero until evidenced |
| Audit artifacts (e.g. SOC 2) | Evidence | Yes | Factual |
| Court records, SEC filings | Evidence | Yes | Factual |
| Benchmarks with published method | Evidence | Yes | Factual |
A hypothetical makes it concrete. You are choosing between two endpoint vendors, CrowdHaven and Sentrix, both invented here for illustration. CrowdHaven sits one box higher on the quadrant. Sentrix publishes independent test results, a current audit report, and a status-page history you can scroll yourself. A decision memo that cites the quadrant as the tiebreaker has just decided on an input two courts classified as unverifiable opinion, while skipping the inputs that could actually be checked. CrowdHaven might still be the right choice. The quadrant simply cannot tell you that.
None of this makes analyst judgment worthless. A good analyst compresses hundreds of market conversations into a single picture, and as orientation that picture has genuine value. Orientation and proof are different jobs. The failure mode is jurisdictional: the graphic migrates from the context column into the evidence column, carrying weight it explicitly disclaims. The report at the center of the NetScout case stated on its face that it consisted of Gartner's own opinions and should not be construed as statements of fact. The courts took Gartner at its word. Most procurement processes do not.
Where OmniAxis fits
The evidence column the courts pointed at is the one OmniAxis was built to run. Not to unseat analyst opinion, but to give the factual-class inputs a system as disciplined as the opinion-class ones already have. OmniAxis takes what a vendor asserts about itself and tests each assertion against independent evidence, then grades the assertion by how well that evidence holds up. The marketing-claim score and the evidence-backed score are shown together, so the gap between a vendor's story and the checkable record reads as data rather than suspicion. OmniAxis refreshes those grades on a managed schedule, weekly on the top plan, with on-demand refreshes on higher plans, because the documented record moves faster than an annual cycle and a decision memo deserves the current version. Every profile carries its last-refresh date, so you always know how current the working is.
The design principle comes straight out of these rulings. If an input cannot be proven true or false, label it as opinion and weight it as opinion. If it can be checked, check it, grade it, and show the working. A quadrant tells you where one firm's judgment landed. An evidence grade tells you what would survive cross-examination. A decision record worth the name holds both, in their proper columns. The courts settled which is which years ago. The rest is up to your process.