Display Advertising Benchmarks 2026: Costs and Evidence

The Display Benchmark Landscape

Separate market statistics from campaign benchmarks. A forecast of advertising spend describes a modelled market total. An observed network report describes activity within that provider's measurement scope. An advertiser sample describes the accounts included in the study. These can all be useful, but they answer different questions and require different acceptance tests.

The sources reviewed for this correction do not establish the previous claims about current programmatic share, network market shares, native-ad spending or television display growth. Those numbers are withdrawn. A missing source trail is an evidence gap, not proof that the opposite is true. It is a reason to keep the figures out of a budget justification until their definitions and provenance can be checked.

A credible benchmark record should preserve the observation dates, publisher, geographic market, account population, buying route, ad format, outcome definition and aggregation method. Also record what was excluded. A median account CTR and a ratio calculated from all pooled clicks and impressions can both be legitimate statistics while producing different answers.

Keep forecasts visibly labelled with the forecast publication date and target period. Once the target period arrives, the projection does not become an observation automatically. Compare it with measured results from the same market definition before describing growth. Do not build a percentage-change calculation from a forecast at one endpoint and an observed total at the other without explaining that mixed basis.

For campaign planning, the first comparison should usually be a consistently measured internal baseline. Record material changes in creative, targeting, placements and tracking beside it. That allows the team to distinguish a change in performance from a change in reporting. A broad external statistic can then provide context without silently replacing the account's actual economics.

CPM Benchmarks and Industry Comparisons

CPM expresses cost per thousand impressions. Before comparing it, define the cost numerator: media only, media plus platform and data fees, or the fully loaded campaign cost including creative. Then define the impression denominator and any filtration. A rate billed on viewable impressions cannot be compared directly with a rate based on all delivered impressions.

Hypothetical example: buy A spends $1,000 for 200,000 impressions, while buy B spends $1,200 for 150,000 impressions. Their delivered-impression CPMs are $5 and $8 respectively. If the same measurement approach identifies 80,000 viewable impressions for A and 120,000 for B, total spend per thousand measured-viewable impressions is $12.50 for A and $10 for B.

These inputs are invented. The second comparison reverses the ranking because its denominator changed. It is an analytical ratio using total spend, not a claim about a platform's billed viewable-CPM metric. It also does not establish which buy creates more sales. Viewability provides delivery context; commercial evaluation still needs an outcome relevant to the advertiser.

If the campaign incurs a further $200 in fees on buy A, its total cost becomes $1,200 and its delivered-impression CPM becomes $6. This fictional extension shows why a media-only proposal and an all-in proposal can look misleadingly different. Request the same cost boundary from each supplier before negotiating against a benchmark.

When an industry label helps

Industry comparisons are useful only when the underlying campaigns are comparable enough for the question. Prospecting for a complex service and retargeting an existing shopping cart may sit in the same vertical while addressing different audiences and outcomes. Inspect the sample and objective before deciding that an account is expensive.

The old industry CPM series is removed because the reviewed evidence does not support its current values. Build an initial range from a scoped test or a supplier forecast whose assumptions are visible. Treat that forecast as a planning estimate until delivery occurs. Revise the plan against qualified outcomes rather than declaring the cheapest inventory the most efficient.

CTR: Historical Evidence and Format

The 0.46% display CTR figure is historical. WordStream's Google Ads Benchmarks for YOUR Industry reports it in a dataset covering August 2017–January 2018. The study describes 14,197 U.S.-based client accounts across Search and Display; that is the overall study sample, not a stated display-only account count. Its methodology says that the reported averages are medians.

This publisher evidence was available through an indexed primary-source extract during the October 2026 correction; direct page retrieval was unavailable. The collection window is explicit, while the exact revision publication date was not independently established. The figure is retained only to clarify the historical provenance of a widely reused number, not to set a current target.

CTR is clicks divided by impressions for the chosen scope. It is not conversion rate, attention or a measure of incremental demand. An apparently strong CTR warrants investigation when landing-page engagement or qualified outcomes are weak. Conversely, a low CTR cannot establish that a brand campaign had no effect; that requires evidence appropriate to its objective.

Hypothetical example: a campaign records 300 clicks from 100,000 impressions, producing 0.3% CTR. If another records 100 clicks from 100,000 impressions, its CTR is 0.1%. Combining their counts produces 0.2%. Because the impression totals match here, this also equals the simple average of the rates; that equivalence will not generally hold when volumes differ.

For an account comparison, preserve raw clicks and impressions so a reviewer can recalculate the chosen rate. Do not average campaign percentages without considering their denominators. Keep the aggregation level explicit: a median of account rates, an impression-weighted rate and a median of individual campaign rates are not interchangeable.

The earlier current-year rankings for native, rich media, video and banner sizes are withdrawn. Test format alongside the message and placement rather than importing a universal uplift. A larger clickable area or a different audience mix can change click behavior without improving customer quality. Evaluate what happens after the click before scaling a format.

Viewability and Attention Metrics

The Media Rating Council's Viewable Ad Impression Measurement Guidelines, desktop version 2.0 dated August 18, 2015, define a standard display opportunity using at least 50% of pixels in an in-focus browser tab for at least one continuous second after rendering. For desktop ads covering at least 242,500 pixels, the large-ad exception uses 30% for the same duration.

Those are qualifying thresholds, not claims about the percentage of campaign impressions that will qualify. The desktop document also permits specified, defensible strong-interaction rules with separate reporting. It does not make mobile or television inventory automatically viewable. The format, environment and measurement method remain part of the definition.

Google's Active View documentation, checked October 4, 2026, distinguishes measurable impressions from viewable impressions and defines viewable rate using the measurable denominator. It separately reports viewable impression distribution against total impressions. It also says that viewability cannot guarantee someone was looking at the screen. The page has no stated publication date for these definitions.

Coverage changes the interpretation

Hypothetical example: 100,000 delivered impressions include 80,000 measurable impressions, of which 60,000 qualify as viewable. Measurability is 80%; viewability among measurable impressions is 75%; measured-viewable distribution across all delivered impressions is 60%. All inputs are fictional, and the rates answer different questions.

The remaining 20,000 unmeasurable impressions have unknown viewability in this example. They should not be silently classified as nonviewable. Nor should a good result within the measured subset be assumed to describe the unmeasured subset. Record coverage beside the rate and investigate systematic differences in the inventory that could and could not be measured.

Attention is a further question. A qualifying display opportunity does not prove that the message was noticed, understood or remembered. If a vendor supplies an attention measure, request its definition, collection method, population and validation. Do not turn a pixel-duration threshold into an attention-time benchmark. Likewise, a completed video and a viewable banner are different events, even if both appear under a broad display budget.

Conversion Rate and CPA Benchmarks

The old current-year conversion-rate and CPA series is withdrawn. A conversion can mean a form submission, qualified enquiry, paid order or another action. The same campaign can therefore produce several valid rates, provided each names its event and denominator. A reported conversion also depends on attribution settings and does not automatically represent a new customer.

Hypothetical example: a campaign spends $2,000 and records 1,000 clicks, 20 enquiries and 5 qualified enquiries. Enquiries per click are 2%; qualified enquiries per click are 0.5%. Cost per enquiry is $100, while cost per qualified enquiry is $400. These invented inputs show why an attractive front-end CPA can conceal poor lead quality.

If the team changes its definition of qualified enquiry, the new result needs a comparable baseline before it can be described as an improvement. Keep duplicate, spam and ineligible responses distinguishable in the underlying records. Report accepted outcomes using the business's actual criteria, not whichever conversion setting creates the most reassuring chart.

Attribution is not incrementality

A post-view conversion records an attribution relationship under specified rules; it is not proof that seeing the ad caused the sale. Retargeting deserves particular care because the audience already showed interest. A lower attributed CPA may reflect that prior intent, so comparisons with prospecting should not be presented as causal lift without a suitable experiment.

When the decision requires incremental impact, plan a controlled test with a clear eligible population and outcome. Confirm that the comparison groups can be interpreted before launch. When a credible control is unavailable, report attributed results and their limitations explicitly. A transparent observational result is more useful than an unsupported claim of causal improvement.

Include the time needed for outcomes to mature. Comparing a newly launched campaign against one whose leads have already completed the sales process can create a misleading ranking. Maintain both an operational view for current delivery and a mature-cohort view for economic decisions, with the relevant windows stated. See our analytics services for the measurement work that connects advertising to business records.

Comparing Networks and Buying Routes

A network, a demand-side platform, a private marketplace and a creative format are not parallel categories. A buying platform can access different inventory sources; a private agreement describes a transaction arrangement; native describes presentation. Mixing these labels into a league table creates overlap before performance is even considered.

A planning framework, not measured market shares, prices or performance rankings.

Compare proposals using the same objective, geography, dates, creative assumptions and cost boundary. Request what will be observable after delivery and what remains modelled. A more expensive proposal may include useful controls or services, but those terms do not prove better outcomes. The test should connect the claimed benefit to a measurable result.

Keep brand suitability, fraud filtration, viewability and performance distinct in reporting. A placement can pass one requirement while failing another. If a supplier reports a filtered rate, establish which impressions were removed and whether the denominator matches the invoice. Do not infer a universal fraud rate or quality premium from the existence of a verification service.

The earlier provider market-share and CPM table is removed. The practical replacement is a comparison sheet whose definitions can survive a supplier change. Our YouTube advertising reference covers adjacent video planning, but a video result should retain its own format and metric scope when used alongside display.

Ad Size and Format Performance

A creative specification establishes eligibility, not expected performance. The sources reviewed here do not support the previous size-by-size CTR rankings, impression shares or rich-media multipliers. Those series are removed. Start with the actual placements available in the chosen buy and verify the required assets before assigning production effort.

Test the message, legibility and destination as well as the dimensions. A smaller unit may need a simpler proposition; a larger one may permit more explanation without guaranteeing that the audience reads it. Avoid treating extra space as permission to cram in copy. The relevant result is whether the creative communicates the intended idea and supports the campaign objective.

Keep format tests interpretable. If a new unit also receives a new audience, different publishers and additional spend, the observed change cannot be attributed to size alone. Record those differences and describe the test as a package comparison. A more narrowly controlled design is needed to isolate the effect of a particular creative feature.

Report enough volume to make the result intelligible without inventing a universal minimum sample rule. Show the clicks and qualified outcomes behind a rate, and explain whether the result is stable enough for the intended decision. A small number of outcomes can move a percentage substantially; a confident-looking decimal does not remove that uncertainty.

Use the large-desktop-ad viewability exception only in its stated measurement context. It changes the qualification threshold, not the value of the inventory or the likelihood of conversion. Compare like measurement definitions before claiming that a bigger unit has improved viewability. Preserve creative production costs alongside media costs when deciding what to scale.

Mobile, Desktop and CTV Display

The previous device-level CPM, attention-time, viewability and growth figures are withdrawn. The desktop MRC document cited above explicitly has a desktop-browser scope; it is not a blanket certification of mobile or connected television measurement. For each environment, identify the applicable standard, implementation and reporting limits.

Separate browser display, in-app display, in-stream video and non-video television placements before comparing results. A report called CTV can involve a different interaction and exposure unit from a web banner. Specify whether reach refers to devices, accounts, households or a modelled estimate of people. Do not treat those labels as synonyms.

On a smaller screen, the destination experience is part of the test. A click followed by a difficult form may point to a conversion-flow issue rather than poor audience quality. On a television screen, a later action may occur on another device, making the attribution rules central to interpretation. These are planning considerations, not evidence of a universal performance advantage for a device.

If a partner supplies cross-device conversions, request the matching approach, eligibility and exclusions. Keep observed actions distinguishable from modelled attribution. A household-level association does not identify which individual saw a particular ad, and an association alone does not establish causation. Match the strength of the recommendation to what the method can actually support.

Allocate a test budget according to the business question and measurement feasibility. The earlier fixed budget split and frequency-cap recommendations had no adequate evidentiary basis and are removed. Establish a frequency policy for the campaign's audience and objective, monitor delivery concentration and downstream outcomes, then adjust with documented reasons. An account-specific learning process is more defensible than an unsourced universal optimum.