OpenAdLibraryOpenAdLibrary
Native Ad Data Studies

Share of Observed Native Creatives: Calculation and Limits

You can't see a competitor's native spend, but you can calculate each advertiser's share of captured creative records in a defined sample, document weighting and coverage, and avoid confusing presence with spend.

Bar chart ranking native advertisers in a vertical by active creative count and run-days as a share-of-voice proxy

Spend data for native advertising is locked up. No network publishes what Taboola, Outbrain, MGID, or Revcontent advertisers pour into a vertical, and there is no impression-share API the way paid search has one. So when a media buyer asks "who actually owns the finance space on native right now?", the honest old answer was a shrug.

You can calculate each advertiser's share of captured creative records within a defined sample. This is a sample-presence measure, not a validated estimate of spend, impressions or market-wide share of voice.

To anchor the scale we are talking about: as of June 2026, OpenAdLibrary has captured 589,036 distinct creatives from 25,933 advertisers across 42 networks, with 5.4 million ad observations behind them. That is the raw material a vertical SOV pass draws from.

What native ad share of voice measures#

Define the measure explicitly: one advertiser's share of captured creative records, optionally weighted by distinct observed dates, within a stated category and collection window. Do not call a first-to-last span continuous runtime or treat weighting as proof of sustained investment.

Classic share of voice has two lineages. The original, from TV and print, is spend-based: your category ad spend divided by total category spend. The digital era added impression-based SOV, where platforms hand you impression share directly. Native gives you neither. Spend is hidden and there is no impression report. What you can see is the surface area an advertiser occupies across publisher widgets, captured by a native ad spy tool. That surface area is the raw material for the proxy.

Repeated observations can help identify examples to inspect. They do not establish whether an advertiser made money or continued spending between sightings. The quoted finance ad is evidence of a captured message, not verification of its tax claim.

Taboola finance native ad about IRS tax forgiveness
Caption: A live Taboola finance ad, headline 'IRS Forgives Millions By June 30th Tax Deadline', captured by OpenAdLibrary, June 2026.

The cited June 2026 classification reports finance, insurance and health counts. Those figures describe the captured sample and do not establish category budgets or market-wide competition.

The two signals, and why both matter#

Active-creative count#

Count distinct captured creatives using a documented deduplication rule and window. Differences can reflect variants, collection coverage or campaign activity; they do not directly measure testing intensity or scale.

Count alone is gameable, though. Some advertisers spray dozens of throwaway variants that die within days. Raw count rewards that churn and overstates their real footprint. You can see the testing instinct in ads like this one, a fresh consumer-product angle that may or may not survive its first week.

Taboola native ad testing a budget air conditioner angle
Caption: A 5-day-old Taboola creative from Consumer World, headline 'Does This $138 AC Run On Almost No Power?', captured by OpenAdLibrary, June 2026.

Run-days (longevity-weighted presence)#

If daily observation records are available, count distinct observed dates for each creative. Keep this separate from elapsed first-to-last span, which may include days with no observation. Neither measure establishes performance.

A small number of repeatedly observed creatives can receive more weight than many single observations. That is a consequence of the chosen formula, not evidence of profitable spend.

State the observation window and censoring. A roughly 28-day maximum in the cited snapshot limits what can be measured; it does not make the longest-observed examples proven winners. No 90-day result is established by that snapshot.

A defensible SOV formula#

Combine the two into a single, transparent score. The simplest version that survives scrutiny:

Share of weighted sample presence (%) = advertiser's sum of distinct observed dates divided by the same sum for all in-scope advertisers, multiplied by 100. Define treatment of missing dates, duplicated records and zero denominators. If only first/last timestamps exist, report an observation-span score separately rather than calling it observed days.

Optionally cap any single creative's run-day contribution so one ancient evergreen ad does not swamp an otherwise dynamic advertiser. Report both the raw creative count and the run-day-weighted score side by side. The gap between them is itself diagnostic.

Here is how the same five (illustrative, not measured) advertisers can re-rank depending on which signal you trust:

Illustrative weighted totals: A 1,820; B 1,240; C 1,050; D 980; E 760. The total is 5,850, giving rounded shares of 31%, 21%, 18%, 17% and 13%. These invented values demonstrate arithmetic only; they do not represent campaign efficiency, profitability or a 28-day cohort.

Count and weighted-presence rankings answer different questions. Neither identifies the true market leader without a validated relationship to the intended outcome.

Defining the vertical (this is where most analyses break)#

SOV is only as good as the denominator. Get the vertical boundary wrong and every percentage is fiction. Three ways to scope a vertical, in increasing order of rigor:

  1. By category or IAB classification. Fast, coarse. Good for "finance" or "health" at a high level, weak when verticals overlap (is a supplement "health" or "nutra"?). For perspective on the coarse-classification problem: across our index, Taboola alone shows 6,048 health creatives and 5,558 finance creatives, while Outbrain leans finance-first (2,640 finance, 2,016 health). The same advertiser can land in different buckets depending on the network and the classifier, which is exactly why category-only scoping is the weakest method.
  2. By keyword and creative angle. Filter on headline language and the creative angle in play. Tighter, but creative copy is messy. Two ads can sell the same offer with wildly different hooks.
  3. By landing-page destination. The strongest signal. Two ads with different copy that resolve to the same offer page (or the same pre-lander) belong to the same competitive cluster.

Use captured destination evidence to help group related offers, and review ambiguous aliases manually. A destination or tracker does not necessarily identify the payer or legal advertiser. Do not click live ads to investigate.

Taboola health native ad about a new hearing device
Caption: A Taboola health ad from Nebroo, headline 'Americans Are Ditching Hearing Aids for This New Device', 26 observed days, captured by OpenAdLibrary, June 2026.

How to run this yourself in OpenAdLibrary#

The whole point of a proxy is that it has to be reproducible. Here is the workflow end to end:

  1. Scope the vertical. Filter the ad index by category, keyword, or, best, landing-page pattern. Decide which networks count (Taboola plus Outbrain only? Add MGID, Revcontent, Teads, MSN, Yahoo?). Document the choice. It is part of the methodology.
  2. Set the observation window. SOV is a snapshot. "Active in the last 30 days" is a sensible default for a fast-moving vertical; widen it for slow seasonal ones.
  3. Group by advertiser. Use the resolved advertiser identity, then fold known aliases together by hand. This de-duplication step is the difference between a real ranking and noise.
  4. Pull the two signals. Active-creative count and total run-days per advertiser.
  5. Compute and chart. Apply the formula, optionally cap run-days per creative, and rank. Plot count against run-day SOV so the divergences pop.
  6. Sanity-check the leaders. Open the top advertisers' creatives. Are they really in your vertical? Are the long-runners genuinely live or stale captures? Trust, then verify.

Publish the records and calculation needed to reproduce the chart, subject to access and privacy limits. Missing traces, unavailable assets or incomplete coverage must remain visible rather than being described as fully auditable evidence.

What the SOV proxy does and does not tell you#

Keep the limits in view so you do not over-claim:

Limits: the measure describes captured records, not spend or impressions; representativeness is unproven; geography, device, source and capture frequency affect coverage; and last seen is not proof of current delivery. State these limits alongside every comparison.

Taboola insurance native ad targeted at Australian users
Caption: A geo-targeted Taboola insurance ad from Real, headline 'Australians looking for life insurance should read this', captured by OpenAdLibrary, June 2026.

These limitations can change the ranking. They are not a statistical confidence interval; estimating one would require an appropriate sampling model.

Why this matters for your next move#

Use the ranking to select competitors and examples for further inspection. It does not identify profitable creatives, available placements or an underserved market.

Want to build your own vertical SOV chart from live, citable ad capture, with real advertisers and real creatives traced to the landing page? Start free: see the 20 newest ads, no card required, then go deeper from $29.99/mo.


Methodology note: the figures in the re-ranking table are illustrative, constructed to demonstrate the SOV proxy, not measured findings. Index totals and vertical counts are real (OpenAdLibrary index, June 2026). Run the workflow on your own vertical to generate real numbers, and always report your network scope, geo and device, observation window, and capture-coverage caveats alongside any SOV chart.

Sources: Brandwatch, Share of Voice formula and examples, Search Engine Land, Share of Voice guide, Taboola, Native advertising platforms 2026.

Frequently asked questions

What is native ad share of voice?
This method measures an advertiser's share of captured creative records within a defined sample. It is not a validated estimate of total spend, impressions or market-wide share of voice.
Can you measure share of voice without ad spend data?
You can calculate sample-presence shares without spend data, but must label them accurately. No correlation with spend or sustained commitment is established here.
Why use run-days instead of just creative count?
Weighting by distinct observed dates changes the question from record count to repeated sample presence. It does not establish profit. Do not substitute elapsed first-to-last span for actually observed dates without labeling the difference.
How is native ad SOV different from search or social SOV?
Native SOV is reconstructed from third-party ad capture because there is no public impression-share API, whereas search SOV uses platform-reported impression share and social SOV often uses mentions or paid reach. For native you have to infer presence from which advertisers appear, how many creatives they run, and how long those creatives persist across publisher widgets.
How do I run a share-of-voice analysis in OpenAdLibrary?
Define the scope, export available records, review grouping and deduplication, and calculate shares using a disclosed formula. Report missing data and ambiguous advertiser attribution. The result describes that sample only.
OpenAdLibrary Research
Written byOpenAdLibrary Research
Data studies & market analysis

The data desk behind OpenAdLibrary. We turn the platform's corpus of captured native ads, advertisers and landing pages into original studies on what is actually running in the wild, methodology and sample sizes stated on every report.