A media plan asks for six banner dimensions. The creative budget covers three. Choosing the three entries from a generic “most popular banner sizes” list feels efficient, but it can produce polished files that are ineligible for the placements you actually bought.
The practical answer is: make the sizes that unlock the largest amount of eligible inventory in this campaign, not the sizes that rank highest on somebody else’s list. Start with the insertion order, publisher spec, network forecast, or recent size-level delivery report. Rank each exact dimension by the incremental opportunities it opens. Then stop when the remaining inventory can be removed, reallocated, covered responsively, or consciously left unserved.
When no usable plan-level data exists, use a three-family fallback rather than pretending there is a universal top three:
- One rectangle, usually 300 × 250.
- One desktop horizontal, usually 728 × 90.
- The exact mobile horizontal named by the buy, commonly 320 × 50 or 300 × 50.
That fallback is an editorial starting point, not a traffic ranking. Google supports all of those dimensions for uploaded display ads, and its examples of common computer and mobile sizes include the same broad families. But Google also supports many other dimensions, warns that some are region-specific, and recommends responsive display ads for spaces where custom uploads do not fit (Google Ads uploaded display specifications; Google Ads common-size examples). Your fourth file might therefore be 300 × 600, 320 × 100, 970 × 250, 336 × 280, or something regional—not “the next size” in a universal chart.
This guide turns that principle into a production decision. The three campaign examples below are original fictional teaching cases, not market benchmarks. Their opportunity counts are deliberately clean, mutually exclusive buckets so the coverage math can be audited. The delivery filenames are sample production plans, not completed ad downloads. Real forecasts can overlap when a single slot accepts multiple sizes; those overlaps must be deduplicated before you add them.
Decide whether fixed banners are the right first deliverable
Before debating 300 × 250 versus 336 × 280, ask whether the campaign requires fixed uploaded banners at all.
Google says responsive display ads adjust their size, appearance, and format to fit almost any available ad space. Uploaded ads preserve more creative control, but they do not adapt in the same way (About responsive display ads). For a Google Display campaign with limited production capacity, that creates three possible answers:
- Responsive only: Build the required landscape, square, logo, and copy assets when broad automated coverage matters more than exact composition.
- Responsive plus a small fixed set: Use responsive ads for general coverage, then make fixed banners for placements where typography, animation, product sequencing, legal layout, or brand presentation must be controlled.
- Fixed only: Use exact uploads when the insertion order, publisher, ad server, or creative concept demands them.
Do not spend a week resizing a fixed concept across fifteen dimensions just because the platform lists fifteen supported dimensions. “Supported” means the platform can accept a file. It does not prove that your plan contains meaningful inventory for that file.
The same distinction applies beyond Google. IAB Tech Lab’s current New Ad Portfolio is organized around aspect ratios and size ranges so flexible units can scale across screens, replacing older display guidance centered on fixed pixels (IAB New Ad Portfolio). That is direction for an interoperable ecosystem, not permission to submit one fluid file everywhere. A buyer and publisher still need an implementation that accepts it. If the trafficking sheet says 970 × 250, deliver 970 × 250 unless the operator confirms another format can serve.
Rank sizes by inventory unlocked
A useful size-priority sheet needs one row per distinct inventory bucket, not one row per dimension copied from a blog post. Capture:
- network, publisher, deal, or line item;
- device and placement;
- exact accepted dimensions;
- forecast opportunities, contracted impressions, or recent delivered impressions;
- whether multiple sizes can fill the same request;
- whether substitution, slot expansion, or responsive rendering is enabled;
- production effort and delivery deadline;
- whether the placement is guaranteed, optional, or negotiable.
Then calculate:
coverage = unique eligible opportunities opened by selected sizes ÷ all forecast eligible opportunities
For each additional size, calculate only the opportunities that are not already reachable through a selected size. That is its incremental coverage. If a 300 × 250 and a 336 × 280 can both fill the same 100,000 requests, selecting both does not create 200,000 opportunities. It creates at most 100,000 before other constraints.
This is why the raw list of supported sizes is not enough. Google Ad Manager, for example, publishes a supply-side “top-performing” list that includes 728 × 90, 336 × 280, 300 × 250, 300 × 50, and 160 × 600. The page is useful evidence that these are established publisher formats. It is not a dated buyer-side share-of-impressions table for your geography, audience, deal type, or campaign (Google Ad Manager supported ad sizes).
Use the closest evidence available in this order:
- Contracted placements and exact publisher specifications. A guaranteed line item you cannot serve is a direct delivery problem.
- A current forecast broken out by eligible size or placement. Ask the media buyer or platform operator for unique, deduplicated opportunity counts.
- Recent delivery by actual creative size for a comparable buy. Google’s programmatic-direct reporting guidance, for example, recommends including “Creative size (delivered)” for accurate reporting by size (Google Ad Manager creative reporting).
- A network-supported-size list plus device and placement assumptions. Use this only as a fallback, and document the uncertainty.
Coverage is not performance. A size that opens 30% of eligible impressions does not guarantee 30% of conversions, nor does an available request equal a unique person. The metric answers a narrower production question: what share of the planned opportunities can receive at least one eligible creative?
Download the interactive banner-size priority planner to change capacity and opportunity counts, or use the filled scenario matrix in a spreadsheet.
Scenario 1: Three files for mixed-device open-web prospecting
A fictional B2B software advertiser, Northstar Analytics, has a four-week programmatic forecast split into six mutually exclusive size buckets. The team can produce three polished HTML5 executions before launch.
| Priority | Exact size | Eligible opportunities | Running coverage | Decision at three-file capacity |
|---|---|---|---|---|
| 1 | 300 × 250 | 340,000 | 34% | Build |
| 2 | 728 × 90 | 220,000 | 56% | Build |
| 3 | 320 × 50 | 180,000 | 74% | Build |
| 4 | 300 × 600 | 110,000 | 85% | Next if capacity expands |
| 5 | 160 × 600 | 80,000 | 93% | Hold |
| 6 | 970 × 250 | 70,000 | 100% | Hold |
The first three files open 740,000 of 1,000,000 forecast opportunities, so selected coverage is 74% and missing inventory is 260,000 opportunities, or 26%.
The decision is not “300 × 250 always wins.” It is “300 × 250 opens the largest unique bucket in this forecast.” The same reasoning selects 728 × 90 and 320 × 50. If a fourth file becomes affordable, 300 × 600 adds 110,000 opportunities and raises coverage from 74% to 85%. Producing 970 × 250 fourth would add only 70,000 in this specific plan.
Delivery list
northstar-prospecting-core-300x250-v1.zipnorthstar-prospecting-core-728x90-v1.zipnorthstar-prospecting-core-320x50-v1.zip- one trafficking sheet mapping all three files to the same approved landing URL, campaign, concept, and version;
- requested static backups, if the media plan requires them;
- a production note recording that 300 × 600, 160 × 600, and 970 × 250 remain unbuilt and represent 26% of forecast opportunities.
This is also a useful creative warning. The 300 × 250 can carry a short proof point, product view, and call to action. The 728 × 90 needs a more horizontal information order. The 320 × 50 may support only a compact promise, brand cue, and button. Treating them as three compositions built from one message hierarchy is faster and safer than scaling one finished canvas until the copy becomes unreadable.
Scenario 2: A mobile-heavy retargeting plan
KiteCart, a fictional retail app, is retargeting recent product viewers. Its plan is strongly mobile, but the original creative request came back as “the usual display sizes.” A size-level forecast changes the answer.
| Priority | Exact size | Eligible opportunities | Running coverage | Decision at three-file capacity |
|---|---|---|---|---|
| 1 | 320 × 50 | 410,000 | 41% | Build |
| 2 | 300 × 250 | 270,000 | 68% | Build |
| 3 | 320 × 100 | 210,000 | 89% | Build |
| 4 | 728 × 90 | 70,000 | 96% | Hold |
| 5 | 300 × 600 | 40,000 | 100% | Hold |
The selected set opens 890,000 of 1,000,000 opportunities: 89% coverage, with 11% missing.
An unexamined desktop-leaning set of 300 × 250, 728 × 90, and 300 × 600 would open only 380,000 opportunities in this example, or 38%. The team would spend the same number of production slots and leave most of the forecast ineligible. Device mix is therefore not a demographic footnote; it can change the first three deliverables.
Google’s uploaded-display specification accepts both 320 × 50 and 320 × 100, but that does not make them interchangeable (Uploaded display ads specifications). The taller unit has twice the vertical pixel count and can support a different hierarchy. Build the 320 × 100 as its own layout, not as a 320 × 50 with extra empty background.
Delivery list
kitecart-retargeting-viewed-product-320x50-v1.zipkitecart-retargeting-viewed-product-300x250-v1.zipkitecart-retargeting-viewed-product-320x100-v1.zip- a copy matrix defining what may be removed as the layout gets shorter;
- the product-feed and fallback-image requirements supplied by the trafficking team, if dynamic rendering is part of the actual buy;
- a gap note showing that 728 × 90 and 300 × 600 account for the remaining 110,000 opportunities.
The copy matrix matters because “same campaign” does not require identical density. A weak mobile banner might try to preserve a headline, subhead, discount, date, product name, legal line, and button. A better 320 × 50 keeps the one claim necessary to recognize the offer, one product or brand cue, and one action. The 300 × 250 can restore product detail and qualification without inventing new promises.
Scenario 3: A direct publisher launch with exact reserved formats
Harbor Week, a fictional streaming release, has a direct package with three reserved placements. The production team can finish two dimensions before the material deadline.
| Priority | Exact size | Contracted opportunities | Running coverage | Decision at two-file capacity |
|---|---|---|---|---|
| 1 | 970 × 250 | 480,000 | 48% | Build |
| 2 | 320 × 100 | 320,000 | 80% | Build |
| 3 | 300 × 600 | 200,000 | 100% | Blocked until resolved |
The first two files cover 800,000 of 1,000,000 contracted opportunities: 80% coverage, with 20% blocked. The missing 300 × 600 is not an abstract reach loss. Under the example’s exact-size, no-substitution assumption, a 200,000-impression line item has no eligible creative.
A generic 300 × 250, 728 × 90, and 320 × 50 package would cover 0% of this fictional contract because none of those dimensions appears in the order. That is why direct specifications outrank a general popularity list. Real publisher portfolios can be exact: BBC Studios lists a 970 × 250 billboard for its standard desktop and landscape large-tablet placement outside the UK, while Reach Solutions publicly lists a portfolio including 970 × 250, 728 × 90, 300 × 600, 300 × 250, 320 × 50, and 320 × 100 (BBC Studios billboard specification; Reach Solutions display ad specifications). Those examples do not prove what every publisher sells. They prove why the current publisher’s own sheet is the controlling document.
Delivery list
harborweek-launch-takeover-970x250-v1.zipharborweek-launch-mobile-tall-320x100-v1.zip- a written exception for the unbuilt 300 × 600 line item;
- one of three resolutions before booking is treated as complete: fund the third file, reallocate that line item into an accepted delivered size, or remove the 200,000 opportunities from the commitment;
- final files delivered by the publisher’s material deadline, not merely by the campaign start date.
The disciplined answer is not to hide the gap in a handoff email. Put the missing dimension, affected line item, opportunity count, owner, and resolution date in the delivery sheet. A two-file creative scope and a three-size media commitment are inconsistent until somebody changes one of them.
Similar dimensions are not automatic substitutes
It is tempting to assume that 300 × 250 will cover 336 × 280, that 300 × 50 will cover 320 × 50, or that a publisher will scale 728 × 90 into 970 × 90. Sometimes an ad server can place a similar-sized creative in a slot or expand a slot. Google Ad Manager documents both behaviors and lets publishers control them (similar-sized ads and slot expansion).
That capability belongs to the serving setup, not to the designer’s imagination. Record substitution only when the network or publisher confirms it for the relevant line item. Also ask what happens to the creative: centered with blank space, scaled, cropped, expanded, or rendered responsively. Eligibility is not the same as acceptable presentation.
Regional inventory can make the same point. Google lists additional regional dimensions such as 750 × 100, 750 × 200, and 750 × 300 in Poland; 930 × 180 in Denmark; and 580 × 400 in Norway (regional sizes in Google Ads specifications). A campaign concentrated in one of those markets may rationally prioritize a dimension that never appears in a US-oriented checklist.
Build a size system, not a pile of resizes
Once the priority list is approved, organize production around layout decisions rather than export buttons.
Define the message floor
Write the smallest unit first. Identify the minimum combination that still communicates a truthful proposition: usually a brand or product cue, one useful message, and one action. Legal text that must appear is part of the floor, not optional decoration. Larger units can add evidence, context, imagery, or product detail, but the core claim should not mutate accidentally across dimensions.
Create composition families
At minimum, plan separately for rectangle, desktop-wide, and mobile-wide units. Add a tall family for 160 × 600 or 300 × 600, and a premium-wide family for 970 × 250, only when selected inventory justifies them. Shared tokens—type scale, spacing, button treatment, logo rules, motion timing—can keep the campaign coherent without forcing every size into the same geometry.
Name files for trafficking, not for the designer’s desktop
Include campaign, concept, exact size, and version in every filename. final-banner-new-2.zip creates avoidable mistakes; kitecart-retargeting-viewed-product-320x100-v1.zip exposes the information an operator needs. Keep the landing URL, click-tracking setup, format, backup requirement, and owner in the delivery sheet instead of encoding every detail into the filename.
Test at delivered dimensions
Review every banner at 100% pixel size, not only enlarged on a design canvas. Confirm that essential copy is readable, the visual focal point survives, the button remains identifiable, and animation or interaction rules match the receiving platform. Then validate the exact file against the current network or publisher specification. A correct 300 × 250 does not prove the 320 × 50 is correct just because both came from the same source file.
Know when to stop making sizes
The next dimension is worth producing when it does at least one of four things:
- unlocks a meaningful, deduplicated block of eligible opportunities;
- satisfies a guaranteed or strategically important placement;
- prevents a known device, market, or publisher gap;
- gives necessary creative control where a responsive format is not sufficient.
Stop when the remaining inventory is immaterial to the plan, can be reallocated without harming the buy, can be covered by an approved responsive format, or costs more to unlock than the team is willing to spend. Set that threshold with the buyer before production—not after the missing-size warning arrives.
The opportunity cost is not only money. If the team has capacity for six outputs, six dimensions of one tired concept may be less useful than three high-coverage dimensions across two genuinely different concepts. Size coverage and creative variation solve different problems. First meet the agreed inventory floor; then decide whether the next unit of effort should expand eligibility or improve the message being tested.
The final approval should fit in one sentence:
We are producing 300 × 250, 728 × 90, and 320 × 50 because they open 74% of this forecast’s unique eligible opportunities; the unbuilt sizes represent 26%, and the media owner has accepted that gap.
Replace the dimensions and percentage with your own evidence. If nobody can write that sentence, the size list is not approved—it is merely familiar.
Sources
- Uploaded display ads specifications — Google Ads Help; accessed September 9, 2026.
- About responsive display ads — Google Ads Help; accessed September 9, 2026.
- About common sizes for responsive display ads — Google Ads Help; accessed September 9, 2026.
- Supported ad sizes — Google Ad Manager Help; accessed September 9, 2026.
- Creative reporting — Google Ad Manager Help; accessed September 9, 2026.
- New Ad Portfolio: Advertising Creative Guidelines — IAB Tech Lab; last updated February 7, 2024; accessed September 9, 2026.
- Billboard ad specification — BBC Studios; specification updated February 18, 2025; accessed September 9, 2026.
- 970 × 250 Billboard and display ad portfolio — Reach Solutions; accessed September 9, 2026.
