The 2026 consensus scan ranks thirty-two reference markets by the dispersion of independent third-party benchmark estimates published in 2024–2026, normalised to a common base year, weighted by source-tier credibility, and reconciled into a single percentage spread. The headline finding is unsurprising: variance is high in young, fast-moving categories and low in mature, regulated ones. The finding worth reading is what the variance shape implies for how the buyer should approach each market.
How the scan was built
Each market in the scan has at least four published estimates from independent analyst firms, government statistical agencies, or trade associations of recognised standing. Estimates were normalised to a common base year and currency, adjusted where necessary for definitional drift between source firms, and weighted by source tier (regulator > statistical agency > tier-1 analyst > tier-2 analyst > industry trade body). The reported dispersion is the weighted percentage spread between the highest-credibility maximum and minimum after normalisation. Where two firms reported the same headline figure but with materially different definitions of the market boundary, those entries were treated as separate observations rather than collapsed into a false-consensus median.
The point of the scan is not to identify a "true" number for each market. It is to identify the markets where the analyst community visibly agrees and the markets where it visibly does not, so that the buyer treats those two cases differently when commissioning further work. A buyer who treats every market like every other market is a buyer who will overpay for sizing work in low-variance categories and underpay for scoping work in high-variance ones.
The high-variance band · where definition is the disagreement
Markets clustering at the high-variance end (>30% peer-estimate dispersion) share three properties: definition ambiguity, accelerating product launches, and unsettled regulatory boundaries. Quantum-computing services, AI-accelerator semiconductors for inference, cell-and-gene-therapy CDMO capacity, and digital-health remote-monitoring all sit in this band. The variance is real. It reflects genuine disagreement among credible analysts about what is in scope, how fast adoption is moving, and which revenue lines belong to which market.
In quantum-computing services, for example, the published estimates for 2025 spend run from roughly $750M to $4.2B depending on whether the analyst counts cloud-only access fees, on-premise hardware revenue, simulator software, or specialist consulting. Each definition is internally consistent. None of them is wrong. They are simply five different markets, sold under one label, by five firms with five publishing schedules.
The right response in this band is not to pick the median. It is to ask which scope definition is the one the buyer is actually buying. Once that is settled, the variance collapses sharply, sometimes to under 15%, because the disagreement was definitional rather than methodological. Buyers who proceed to commission work in this band without doing the scoping conversation first end up with a number that survives in their internal documents and dies in committee.
There is one operational corollary. The credibility weight of the source tier matters more in this band than anywhere else. A statistical-agency definition, where one exists, is not just one estimate among five. It is the definition the regulator will use when the market is eventually classified, and the buyer's framing should anchor on that definition rather than on whichever analyst's framing is closest to the buyer's instinct. Definition convergence is a forward-leading indicator of variance compression.
The low-variance band · where the median is well-anchored
Markets clustering at the low-variance end (<10% dispersion) share the opposite properties: stable definitions, slow product cycles, and authoritative source data from regulators or statistical agencies. Commercial property-and-casualty insurance premiums, hospital inpatient services in the US, retail mortgage origination volume, and utility-scale electricity generation in Western Europe all sit here. The median is well-anchored, the path-weighted estimate is stable to within a few percent across the major firms, and a buyer who treats one of these markets as risky because the headline number is large is misreading the underlying signal.
The operational point in this band is that further commissioned work should not focus on resizing the headline. It should focus on the sub-segments and on the forward trajectory. Where the headline is well-anchored, the marginal value of research is in the layers below the headline: the breakdown by line of business, the geographic dispersion under the national figure, the rank-ordering of the leading operators, and the forward sensitivity to the one or two policy variables that actually matter. A buyer who commissions a $20K resizing study on a market where the headline is published with a 6% spread has misallocated the budget by a factor of three or four.
The middle band · where the read becomes operational
The middle band, markets with 15–25% variance, is where the scan becomes a working tool. Most enterprise SaaS categories, most automotive subsectors, most renewable-energy installation types, and most pharmaceutical specialty subcategories sit here. A median figure with no variance band misrepresents the underlying disagreement. A figure published with the band intact, and with the cause of the band attributed (definitional drift versus growth-rate disagreement versus regulatory ambiguity), tells the buyer where the open questions are and what would resolve them.
This is where research methodology starts to matter most to the buyer. Two firms can both publish a $34B headline with a 21% spread, but if one firm's spread is driven by definitional drift and the other's is driven by growth-rate disagreement, the implications for the buyer are completely different. The first will resolve when the buyer picks a definition. The second will only resolve as the next two quarters of statistical-agency data arrive, and any commissioned work that ignores the cause will produce a memo that ages badly.
What the scan does not tell you
The scan ranks markets by published-estimate dispersion. It does not rank markets by underlying volatility, underlying opportunity, or underlying buyer suitability. A high-dispersion market is not a worse market to commission than a low-dispersion market. It is a market where the buyer needs a different kind of work, scoping and definition first, sizing second. A low-dispersion market is not a finished question. It is a market where the headline is settled and the work moves to the layers under it.
There is a second thing the scan does not capture, and it is worth surfacing. Sources that publish in different fiscal calendars sometimes appear to disagree because they are measuring different windows. The normalisation to a common base year removes the worst of this, but there is a residual band of the spread, usually 2–5 percentage points, that is calendar artefact rather than substantive disagreement. The corrections page of each source is the place to verify which is which when the spread sits near the band thresholds.
The published table behind this scan is updated quarterly, with the methodology, source-tier weights, and per-market commentary attached. Variance is the signal. The number under it is just the median.