Operations
Part of Reading programmatic advertising with a sceptical eye
Programmatic advertising benchmarks: planning ranges and how to read them
Programmatic advertising benchmarks as planning ranges for CPM, win rates, viewability, invalid traffic and fees, and how to read them against your platform data.
What to take away
- Open-exchange display CPM in the US typically sits between $1 and $4; CTV and premium video run ten times higher.
- Viewability on measurable display impressions usually falls between 60 and 75 percent; general invalid traffic between 1 and 5 percent.
- Win rates run from low single digits to low teens on open exchange, and far higher inside private deals.
- Fees and technology take commonly absorb 30 to 50 percent of a working media budget.
- These are planning ranges, not published point estimates. Read the definition before you compare anything.
The 2027 benchmark table
A usable benchmark travels with its population, period, and metric definition. The table pairs each figure with the units it is measured on and the trap that usually follows. Reading programmatic advertising with a sceptical eye is worth doing before you accept any external number.
| Metric | Typical planning range (US) | Measured on | Main trap |
|---|---|---|---|
| Display CPM, open exchange | $1 to $4 | Served impressions | Blended CPM hiding format mix |
| Video and CTV CPM | $10 to $45, CTV at the top | Served impressions | Comparing CTV against display |
| Win rate, open exchange | Low single digits to low teens | Eligible bid requests | Counting only the auctions you won |
| Win rate, private deals | 40 to 90 percent | Deal requests | Thin deal volume |
| Viewability, display | 60 to 75 percent | Measurable impressions | Reported against served units |
| Viewability, video | 70 to 85 percent | Measurable impressions | Autoplay and muted placements |
| General invalid traffic | 1 to 5 percent | Measured impressions | Vendor-set thresholds |
| Sophisticated invalid traffic | Under 1 percent | Measured impressions | Detection coverage gaps |
| Fees and tech take | 30 to 50 percent of media budget | Contracted spend | Applied to net instead of gross |
[figure 1]
Benchmark Layers and Boundaries
Layer
- Supply
- Authorized seller share, fees
- Auction
- Bid rate, win rate
- Delivery
- Served, viewable, audible
- Quality
- Invalid traffic, disputes
- Outcome
- Qualified leads, retention
- Economics
- Full cost, payback
Measures
- Supply
- Same markets, policies
- Auction
- Same deal type, format
- Delivery
- Same vendor, standard, device
- Quality
- Same detection coverage
- Outcome
- Same customer, offer, design
- Economics
- Same cost boundary, value rules
Comparison boundary
- Supply
- Auction
- Delivery
- Quality
- Outcome
- Economics
Viewability still follows the Media Rating Council convention most buyers reference: half the pixels for one continuous second for display, two seconds for video. Ask for it against measurable impressions, not served ones. A rate reported on served units flatters itself.
What the 2027 view changes
Supply has moved. More working spend now flows through private deals, curated marketplaces, retail media networks, and streaming platforms. Each carries its own CPM band and win-rate behaviour. A blended account CPM is a weak signal now.
Measurement denominators tightened. Buyers increasingly request viewability and invalid traffic reported against measurable impressions, with the denominator written down. Where served units remain the base, the figure is inflated by design.
Bidding is more automated. Strategies such as Target CPA and Maximize Conversions set bids without a manual ceiling, so win rate reflects the algorithm's target rather than your price (Wikipedia). Win rate is a diagnostic, not a goal.
Where these numbers come from
No single registry publishes an official programmatic benchmark. Practical figures are assembled from DSP and SSP reporting, platform benchmark reports, MRC-accredited viewability and fraud vendors, trade association surveys, and agency benchmark decks. Quality varies more than the numbers do.
Discipline travels well here. The GAO evaluation design guide ties evaluation questions to evidence needs and design choices. The NIST experimental design guidance begins design selection with the objective and practical constraints. Neither replaces your own export.
For every figure, record the publication date, sample, geography, metric definition, and funder. Then recalculate from your own platform data. paid media attribution rules move the outcome column further than any CPM shift, so settle the credit rule before comparing anything downstream.
Example: reading one line against the table
Reading one line against the table
- Pull served, measurable, viewable, and invalid-traffic counts for the same flight and date range.
- Compare the CPM against its format band, never against a blended account average.
- Read win rate beside loss reasons. A low win rate with a high floor is a pricing question, not a creative one.
- Add DSP fees, tech fees, and any managed-service fee, then express the total as a share of gross media.
- Judge the line on cost per qualified lead, not cost per click.
Worked this way, a display line at the low end of its band is unremarkable until the outcome column disagrees. A video line priced at a display CPM is a supply-path question. The display advertising benchmarks page carries the format detail this table compresses.
Checklist: fields to record before comparing
Fields to record before comparing
- Live field name and report type, with the extraction date.
- Denominator typeserved, measurable, or viewable units.
- Invalid traffic treatment, including the vendor definition and threshold.
- Currency, geography, and whether fees sit inside or outside the media number.
[figure 2]
Metric Definition Checklist
- Record live field name and report type
- Note attribution settings and extraction date
- Segment by audience, market, format, device
- Annotate supply, bid, creative, fee changes
- Allow qualification and returns to mature
- Separate platform attribution from causal evidence
- Set investigation, pause, and scale thresholds
A dated programmatic advertising checklist stops two teams from recalculating the same figure against different fields. Set a review date and name the change that would trigger an earlier one.
