Why do Gartner and IDC numbers differ for the same market?

Because each firm defines the market's boundaries independently, builds the estimate with its own method — top-down allocation, bottom-up vendor aggregation, or buyer surveys — and applies its own regional and currency assumptions. Same reality, different questions, different totals.

Which market-size number should I trust?

The one whose definition matches the claim you are making. For momentum arguments, cite the full range with both ends attributed; for conservative planning, use the narrowest defensible definition. Never quote the largest figure without stating its scope.

Why is the martech market sized anywhere from $189B to $585B?

The spread across firms such as Grand View Research, Precedence, and IMARC comes mainly from boundary decisions — whether services, data, agency work, adtech, and retail media count as martech — compounded by different construction methods and base years.

Is a wide range of estimates a red flag for a market?

No — it signals contested category boundaries, which is normal for fast-moving markets. Even global advertising, one of the most-measured markets anywhere, carries a roughly 25% spread between major 2026 forecasts. The range is a free sensitivity analysis, not a warning.

Why analyst market-size numbers disagree

Guide · Frameworks · 5 min read · last verified 2026-07-27

Reviewed before publication Editorial board Independent commercial review

In shortAnalyst market sizes differ because of definitions, methodology, and regional assumptions — not error. How to read the range instead of hunting for one true number.

Analyst market-size numbers disagree because different firms measure differently defined markets with different methods — not because one of them is wrong. The spread between estimates is information: it tells you how contested a category's boundaries are and which assumptions move the total most. Learning to read the range is far more useful than hunting for the one true number, because the one true number does not exist.

The spread is a feature of the method, not a failure

Consider marketing technology. Across research firms, the global martech market is sized anywhere from roughly $189 billion to $585 billion — Grand View Research, Precedence Research, and IMARC each publish materially different totals for what sounds like the same market. That is more than a threefold spread between firms whose business is producing exactly this kind of estimate.

The instinctive reading is that somebody is incompetent. The correct reading is that they answered different questions. Once you see which questions, the spread stops being noise and starts being a map of the category's fault lines.

Definitions decide most of the gap

The largest single driver of disagreement is scope: what counts as inside the market. For martech, does the figure cover software licences only, or also the services, data, and agency work wrapped around them? Does it include advertising technology, or treat adtech as a separate market? Each inclusion decision can move the total by tens of billions, and firms make these calls independently and often quietly.

Retail media is a live example of a classification fight. The budgets are real and large, but whether a given firm files them under advertising spend, e-commerce infrastructure, or marketing technology varies — and wherever those billions are filed, that market's total inflates relative to a firm that filed them elsewhere. Neither firm is wrong. They drew the boundary in different places.

The practical consequence: before comparing two market sizes, read each report's definition section first. If you cannot find a definition, that itself tells you which tier of source you are dealing with.

Methodology: how the number is actually built

The second driver is construction method. Top-down estimates start from macro spending data and allocate a share to the category. Bottom-up estimates aggregate vendor revenues — which depends entirely on which vendors made the list and how the firm estimated revenue for the private ones. Survey-based estimates project from what a sample of buyers reports spending, inheriting all the biases of sampling and self-reporting.

Each method is defensible. Each produces a different figure from the same reality. A bottom-up estimate misses spend flowing to vendors outside the list; a top-down estimate depends on an allocation percentage that is itself a judgement; a survey depends on who answered. Firms rarely reconcile their method against rivals, so the differences persist across publication cycles.

Regional and temporal assumptions

A quieter set of drivers: which regions are counted and how currencies are converted. A market sized in dollars during a strong-dollar year shrinks on paper without any underlying change. Base years differ — one firm's 2025 figure may be an actual estimate, another's a forecast made in 2023. Calendar versus fiscal years, and silent base-year restatements when a firm revises its model, add further drift. None of these are headline issues, but stacked together they comfortably explain double-digit percentage gaps between otherwise similar estimates.

A live example: 2026 global advertising spend

Advertising is one of the most-measured markets in the world, tracked continuously by multiple credible firms — and it still carries a spread. For 2026, Dentsu forecasts global ad spend at around $1.04 trillion, crossing the trillion mark for the first time (Dentsu), while WARC's forecast for the same year sits near $1.30 trillion (as reported by Marketing-Interactive; WARC's full dataset is subscriber-only, which is itself a lesson in how sourcing works).

Both are serious, methodical organisations. The quarter-trillion gap is scope: which channels and trade activities count as advertising, how self-serve platform spend is estimated, and which markets are included at what exchange rates. If the best-measured category in commercial research carries a 25% spread, expect wider spreads everywhere else — and stop treating any single figure as ground truth.

How to read a range like an analyst

A practical sequence for any market where estimates disagree. First, collect the estimates and place them on a line, lowest to highest. Second, read each source's definition and note what is in and out — this usually explains the ordering. Third, identify each firm's method (top-down, bottom-up, survey) and base year. Fourth, annotate the line: 'low end excludes services; high end includes adtech and agency spend'. That annotated range is the actual deliverable — it says what the market is, under which definition, which is what a sophisticated reader wanted to know all along.

This also matters for how AI systems describe your market. Answer engines synthesise across all of these sources without necessarily distinguishing definitions, so the range you find manually is the raw material of the answer a buyer receives when they ask an assistant how big your market is. Understanding the spread helps you predict — and, with the right published content, inform — that answer. When Magrios builds a market study it keeps each figure attached to its source, definition, and authority tier for exactly this reason: ranges should stay visible, not be silently averaged into false precision.

Which number should you use?

It depends on the claim you are making. If you are arguing category momentum, state the range and cite both ends — 'sized between X and Y depending on definition' reads as sophistication, not evasion. If you are building a conservative plan, use the narrowest defensible definition. What you should never do is quote the largest available number without its definition, because the first reader who checks will find the smaller ones, and will then wonder what else you rounded up.

The disagreement between analysts is not a problem to be solved. It is the closest thing you will get to a sensitivity analysis of the category, performed for free, by professionals, in public.