Methodology, version 1

Oct 19, 2026

This page explains how the Value Model Institute (VMI) collects, counts and reports the data behind every VMI report and Index, whatever group a report covers. Each report states the methodology version and the taxonomy version it used. When either changes, VMI publishes a new version of this page and leaves earlier versions online for the reports that cite them.

Where the data comes from

Every number VMI reports comes from the Agency Model Map Diagnostic, an online diagnostic that an agency leader completes in about five minutes. Each respondent receives their own result. The questions do not change within a wave. The diagnostic asks how the agency defines its expertise, how new work arrives, how it scopes work, how it uses AI in client work and how it sets its prices. It asks what tensions its leaders feel and what they plan to change. It also records the respondent's role, the agency's type, ownership, headcount and country, and three financial questions that respondents answer as ranges.

Agency leaders reach the diagnostic through partner associations and networks and through VMI's own channels. Participation is voluntary. The results therefore describe the agencies that took part, and VMI does not weight them to represent the industry as a whole. VMI states the number of agencies behind every chart so readers can judge each result on its own count.

Who counts: one agency, one response

VMI counts one response per agency. When several people from one agency respond, the response from the most senior respondent (by the role each one selects) counts. When the same person responds more than once, the latest response counts. Each firm or office is a separate agency, so three offices of one network are three agencies.

Only firms whose primary business is agency services (advertising, media, digital, design, PR or communications consulting) enter the benchmarks. Firms in adjacent businesses receive their own diagnostic result, and VMI leaves them out of every report.

VMI reviews every response before counting it. A reviewer holds a response out of the counts when it duplicates another or when the respondent's firm is not an agency. A reviewer also holds out a response whose answers contradict one another, for example a headcount and a net revenue range that cannot both be true. VMI logs every exclusion with a one-line reason. Responses to earlier versions of the diagnostic, before 14 July 2026, are not in any published report.

How VMI groups agencies

The diagnostic records the country of the office the respondent answers for. VMI never asks a respondent for their region. VMI assigns every country to one of eleven regions and one of three super regions (Americas, EMEA and APAC) from a published mapping table. VMI also reports Europe as a combined region: UK & Ireland, Western & Northern Europe, Southern Europe and Central & Eastern Europe together. This page uses VMI geographic taxonomy version 1.1.

An association edition covers the agencies on a member list the association gives VMI. A country edition covers the agencies that chose that country.

Minimum counts

VMI counts minimums in agencies, never in responses. A published geographic group (global, super region, combined region, region or country) appears in a report only when at least 30 agencies in it responded. The three financial charts (net revenue, net revenue per full-time employee and recurring revenue) appear only when at least 20 agencies answered them. VMI produces an association edition only when at least 20 of the association's member agencies responded. Inside an association edition every chart needs at least 10 agencies, and the financial charts keep their minimum of 20.

When a group falls below its minimum, its responses count inside the next level up (a country inside its region, a region inside its super region) and the report does not name the group. Every chart states its agency count.

How to read the numbers

Each share is a percentage of the agencies in the group that answered the question, rounded to a whole number. Rounded shares can add to 99 or 101. On questions where respondents could choose more than one answer, the shares add to more than 100. On the financial questions, VMI reports shares of the agencies that gave a range and states separately how many preferred not to say.

A share is an estimate, and the margins below hold 19 times in 20. With 30 agencies, a share can be off by up to 18 percentage points; with 100 agencies, by up to 10; with 200, by up to 7. Readers should treat a difference between two groups as a pattern worth attention, and not as a measured gap, unless it is larger than those margins. The margins assume a random sample; because participation is voluntary, treat them as a guide to size, not a guarantee.

Respondents answer the financial questions as ranges in their own currency. VMI combines responses across currencies by range position, with the range edges in each currency fixed as equivalents at one reference date, and never converts an individual answer. Tables that mix currencies are approximate at the range edges.

VMI assigns each agency one of four archetypes: Busy by Design, Scaling with Strain, Expertly Undervalued or Distinctly Scalable. Three answers decide it: how the agency defines its expertise, how it scopes work and how it sets prices. Because those three answers define the archetype, VMI never reports them broken down by archetype; every other question can be.

A report compares a group only with the larger geography it sits in: a country with its region, a region with its super region or with all participating agencies. VMI does not compare or rank associations against one another.

Privacy

No report identifies an agency or a respondent. VMI publishes aggregates only, and the minimum counts above exist so that no reader can work back to a single agency's answers. A partner association sees which of its member firms completed the diagnostic, by firm name and completion status. It sees nothing else: no respondent names, no answers, no scores and no results. VMI does not share or sell response data. Questions about privacy go to privacy@lodestaragencyconsulting.com.

Versions

Version 1

Date 19 October 2026

Taxonomy 1.1

Data Wave 1 (14 July to 31 December 2026)

Change First version

How to cite

Source: Value Model Institute, State of Agency Value Models 2026: [Insert Group] Edition, methodology version 1.