Strategy
Part of Paid media strategy: a governance framework
Paid media strategy benchmarks: how to compare them
Paid media strategy benchmarks are useful only when their definitions, cohorts, maturity windows, quality guardrails, and source limitations are clear.
What to take away
- Paid media strategy benchmarks are definitions, cohorts and maturity windows. Change one and you have a different statistic, not a better one.
- This article gives method, not numbers. Platform reports supply figures; the method decides whether those figures describe your account.
- Internal history is the first source while your definitions hold steady, because you control exclusions, currency and maturity.
- Pair every efficiency metric with a quality guardrail such as cancellation rate or complaint volume, and publish both.
- Recompute the figure from the governed export before you present it, and record the pull date.
Why this gives method, not numbers
Published cost per click, CPM and return on ad spend move with auction pressure, geography, format, season and industry. Quoted without that context, a figure describes somebody else's account rather than a target.
This article covers how to define a benchmark, where to source one, and how to compare two without fooling yourself. For figures, use your own export or a named platform report.
Settle the wider paid media strategy that these numbers are meant to serve before you quote any of them.
Define the benchmark before you source it
Every usable benchmark has four parts: numerator, denominator, cohort and maturity window. Write all four in words a colleague could audit.
Define the benchmark
| Layer | Examples | Context required |
|---|---|---|
| Delivery | Impressions, reach, frequency | Counting method, eligible population, platform, period |
| Attention proxy | Viewability, completed view, interaction | Format, placement, threshold, audibility, fraud controls |
| Response | Click, visit, call, form | Unique rules, quality, geography, destination, bot handling |
| Outcome | Qualified sale, net order, retained customer | Source, delay, cancellations, refunds, capacity |
| Economics | Contribution, marginal return, payback | Full cost, cash timing, incrementality, uncertainty |
| System health | Tracking coverage, match rate, reporting delay | Expected denominator, incident threshold, owner |
The six layers show what each metric can and cannot prove. Mixing them is the most common cause of a comparison that looks valid and is not.
Viewability guidance comes from the Media Rating Council. A viewable impression is a media quality unit, not evidence that anyone noticed or acted. Verify the current document and the implementation behind a vendor claim.
Google describes Reach Planner figures as planning forecasts for reach and frequency, shaped by product methods and the settings you choose. Treat them as inputs, not as delivery promises.
Where benchmark data comes from, and its biases
Google Ads includes a Benchmarks view that compares your campaigns with similar advertisers. The peer set is assembled by the platform and cannot be audited. Platform conversions follow the platform's attribution window, so they will not match your CRM.
Third party studies carry selection bias. Agencies and vendors publish them to attract clients, so weak accounts rarely appear, and methods differ from yours.
Bid settings move reported costs too. Google Ads Target CPA bidding sets bids toward a cost per action the advertiser chooses, so two accounts can show different costs for the same traffic.
Checklist before citing an external source:
Benchmark data biases
- Who paid for the study, and what do they sell?
- What are the cohort, period, currency and attribution window?
- Is their metric defined the way mine is?
- Are closed or dormant accounts included?
Never fill a missing plan with an unlabeled industry average. Buyers ask hard questions about these sources, and common paid media strategy questions covers how to answer them out loud.
Example: one comparable cohort, full arithmetic
The figures below are placeholders that show the arithmetic. They are not market data.
Decision: should we move budget out of Display and into US Search lead generation?
Numerator: total spend in the cohort. Denominator: leads accepted by sales after validation.
Cohort: US, Google Search, non brand, one service line, desktop and mobile, campaigns one to four. Exclusions: job seekers, existing customers and internal traffic.
Maturity window: a lead counts if it arrives within 45 days of the click. The export is pulled 60 days after the month closes.
Arithmetic: USD 12,000 of spend divided by 40 qualified leads gives USD 300 per qualified lead. The platform reported 70 conversions at USD 171 under a 30 day click window. Different denominator, different window, different statistic.
Compare only against the same cohort one quarter earlier, computed under the same rules.
Steps to reproduce a benchmark comparison
Reproduce a benchmark comparison
- Name the decision the number will inform, and who will act on it.
- Write the numerator and denominator so a colleague could check them.
- Fix the cohortchannel, campaign type, geography, device, returning customers, exclusions.
- Set the maturity window and the pull date before you look at results.
- Recompute from the governed export, never from a dashboard screenshot.
- Record any mismatch with the platform figure and its likely cause.
Inclusion rules and windows are the ground covered item by item in the paid media strategy checklist, which is worth running before you publish a comparison.
Benchmark reporting checklist
- Show count, rate, median, percentiles, range
- Separate new and returning customers
- Segment market, offer, channel role, delay
- Keep attributed and experimental results distinct
- Annotate product, policy, tracking, definition changes
- Set decision thresholds before viewing period
The GAO evaluation design guide ties an evaluation question to the evidence it needs, which is the discipline of naming the decision first. NIST guidance on selecting an experimental design starts from the objective and practical constraints, which is why a recorded observation stays an observation.
Guardrails that travel with efficiency numbers
Lower cost per lead can come from more cancellations. Higher reach can raise complaint volume. Faster lead response can reward premature contact.
Choose one quality metric and one risk metric for each efficiency benchmark, and report them together. Cancellation rate, refund rate, complaint volume and contact frequency all work.
Set the guardrail before you attach any incentive, or the incentive will find the cheap route. The repeating patterns are catalogued in paid media strategy mistakes tested against experience.
