How to assess programmatic advertising benchmarks

What to take away

Why the definition matters more than the number

A benchmark is only as good as the definition behind it. Two dashboards can both report a median CPM: one counts impressions after invalid traffic filtering, the other counts everything served.

That is why the programmatic advertising measurement and reporting pillar starts with definitions. Read it before quoting a figure.

Ask which impressions were counted and which were removed before counting. The answers usually explain why the figure differs from your own reporting.

Industry working groups are where those definitions get argued over. The IAB UK guide to its communities and groups lists which groups exist and what each covers.

The ISBA and PwC Programmatic Supply Chain Transparency Study (2020) covered 15 advertisers, 8 agencies and 5 ad tech vendors.

The compliance line inside every measurement plan

Benchmark comparisons often pull together audience data from several platforms, which brings obligations that sit alongside the reporting.

Agencies that process personal data for advertising must pay the data protection fee unless they fall within a narrow exemption. The exemption covers processing only for staff administration, advertising, marketing and public relations, accounts and records, or similar purposes. The ICO sets the bands by size and turnover.

The UK General Data Protection Regulation is the retained version in force here. Consent records and lawful-basis notes belong beside the campaign data.

What a benchmark comparison should hold

Read the table left to right: the source tells you how much weight the number deserves. A platform figure is fine for pacing your spend but useless for a cross-market claim.

A benchmark that survives those checks can carry weight in a forecast. One that does not should stay in the appendix.

Where an honest forecast breaks

Published benchmarks fail in four places: sample size, seasonality, inventory mix and attribution model.

A median from a small sample moves sharply when one large advertiser joins or leaves. A common working minimum is 200 conversions per segment.

Seasonality does the same work. November and February medians describe different auction pressure. Compare like with like over 12 months, not month to month.

Inventory mix matters too. An open-market display benchmark says little about a private marketplace deal. Check that half the benchmark's impressions match your deal type.

Attribution is the fourth. Model choice moves credit between channels. Keep one model fixed for at least two quarters before comparing. The programmatic advertising attribution methods guide sets out how to keep a model stable.

For example, the ISBA and PwC study (2020) found 15% of spend unaccounted for. If your own data shows 9% (illustrative), your forecast uses 9% and cites the 15% prior.

How to fold a benchmark into a forecast

Write down which source you used and when you last checked it. Forecasts that carry a dated citation survive scrutiny better than ones that carry a rounded number.

The IAB UK and PwC Digital Adspend study is published annually, giving a natural checkpoint for cost assumptions. Monthly delivery metrics need a monthly refresh from your own platform reporting.

Common questions

How often should benchmark figures be refreshed?

Match the cadence to the process. Delivery metrics need a monthly look; cost assumptions can wait for a quarterly review.

What makes a benchmark source credible?

It names the sample, date and definition. If any is missing, treat the figure as a hypothesis, not a fact.

Can we compare our CPM with a published median?

Only when the definition matches. Check whether the published median excludes invalid traffic and covers the same markets, formats and deal types as your activity.