OpenAdLibraryOpenAdLibrary
Native Ad Data Studies

What Products Are Advertisers Scaling? How to Read the Signals in Captured Ads

Spend charts only update after the race is over, so to see what advertisers are scaling right now you read live native ad capture for longevity, spread, and variant signals. Here is the method, with real numbers from 589,000+ captured creatives.

Dashboard view of live native ads grouped by product category, with longevity and placement-spread signals highlighting which products advertisers are scaling

Spend dashboards look backward, and third-party spend figures are modeled rather than measured. If you buy or analyze media, a more useful question is what advertisers are visibly putting into market: which ads keep appearing, where, in how many versions and in which countries. Captured ads can answer that observational question. They cannot tell you which campaigns are profitable.

This is a methodology piece, not a leaderboard of made-up dollar figures. Nobody outside an advertiser's account sees their invoices, their bids, or their true impression counts. Any tool that claims to is guessing. What you can observe is the ad as captured: when it was first seen, how long it kept appearing, how many placements and networks carried it, how many variants appeared, and which countries it was seen in. Read together, these signals can suggest which offers are getting sustained, broad distribution. They do not show spend, bids, impressions, clicks, conversions, profit or ROAS. Here's how to read them, what the patterns mean, and how to run the same analysis yourself.

The examples below come from OpenAdLibrary captures dated June 2026. The aggregate totals and category counts in the earlier version of this article are withdrawn because no reproducible extract is available for review. Capture coverage is incomplete and uneven by network and country, so the examples illustrate the method rather than measure the market.

What "scaling" actually looks like in the wild#

When an advertiser scales a product, it usually runs more than one ad: it pushes the offer across more placements, makes creative variants and may enter new countries. That leaves a footprint you can look for in captured native ads: how long each creative kept appearing (longevity), how many placements and networks carried it (spread), how many near-duplicate variants appeared (iteration), and whether new countries appeared over time (geo expansion). Sustained, broad, multiplying ads are consistent with scaling, but they do not prove it, because a brand campaign, a fixed test budget, slow optimisation or gaps in the capture can produce the same pattern. A short, narrow, single-variant footprint may be a test, a small budget, a targeting choice or something the capture missed.

The word has a precise meaning in this trade, and the precision matters here. See Scaling (Media Buying) for the definition. A buyer who has found a profitable offer wants to spend more on it without breaking the unit economics. That effort can leave a visible footprint, which is why some scaling activity is partly observable from outside the account. It remains an inference, not a measurement.

Take this Taboola creative from "Fresh Start Information": a tax-relief offer styled as a news bulletin, seen in our capture across 13 days in June 2026. That is an observation of run length, not evidence that the offer was profitable or approved. Thirteen days is also consistent with a testing budget, a slow optimisation cycle or a campaign nobody had paused yet, and we have not verified the ad's claim.

Taboola finance ad: IRS forgives millions by June 30 tax deadline
Caption: A Taboola finance ad, headline 'IRS Forgives Millions By June 30th Tax Deadline,' seen in OpenAdLibrary's capture across 13 days, June 2026 (historical capture; the ad's claim is not verified).

A test usually leaves a small footprint and a scale-up a large one, but how visible an ad is in a capture also depends on where and when the capture looked. The pattern to look for is the same offer with fresh creative, more placements and more countries over several weeks. Treat it as a lead to investigate, not a conclusion about what is working.

The four signals, ranked by how hard they are to fake#

Not every signal is equal. Some are measured straight off the capture. One is modeled, and therefore soft. Here is how to weight them.

Signal What it measures How you observe it Limits
Longevity How long a creative kept appearing in the capture First-seen vs. last-seen dates for the same creative Bounded by capture frequency and the observation window; says nothing about profit
Spread Breadth across placements and networks Distinct placements and networks where the offer was captured Covers only the placements and countries the capture visits
Iteration How many variants appear Near-duplicate creatives for one offer or advertiser Depends on how variants are matched; variants can be generated automatically
Geo expansion New markets appearing New countries recorded for an existing offer Depends on regional coverage; a new country in the capture may not be a new market for the advertiser
Estimated spend Modeled budget Third-party impression and CPM assumptions Inferred, never measured

The point of the table is the bottom row. Estimated spend is a headline number on some ad intelligence tools, and it is the least trustworthy thing on the page. It is a calculation stacked on assumptions about impressions and CPMs the tool never actually sees. Use it to sort if you must. Never quote it. The top four rows are read from captured ads rather than modeled, but they inherit every gap in the capture, so report the observation window and coverage alongside them.

This is also why capturing the real creative matters. If your source only logs that an ad existed, you can count longevity but you cannot cluster variants, because you cannot see whether two ads are the same offer wearing different headlines. Capturing the actual image and copy at full quality is what makes the iteration signal exist at all.

Reading the pattern: tested vs. scaled#

A capture contains many short-lived ads and fewer long-running, widely placed ones. Telling a likely test from a likely scale-up is the skill, and it stays a judgment rather than a measurement.

A likely test looks like one or two creatives, a handful of placements and a single country, seen for a few days and then gone from the capture. That fits a buyer checking whether an offer converts, but it also fits a small budget or a gap in the capture. A likely scale-up is the opposite: the same offer across dozens of placements and often several networks, multiple variants, run time measured in weeks, and new countries appearing over time.

You can see the start of a footprint in our June 2026 capture. This dating ad from "ThisRomance" had just been captured for the first time, with no earlier sightings. A first sighting is not the launch date, and one sighting cannot tell you whether the ad was a test or the start of a larger campaign.

Taboola dating ad: meet people who know what they want
Caption: A Taboola dating ad on its first day in OpenAdLibrary's capture, June 2026. It had no earlier sightings in the capture.

Now picture the opposite footprint. This is a hypothetical example, not a captured ad: a blood-sugar supplement seen for nine weeks across 40+ placements on both Taboola and Outbrain, rotating six headline variants over one pre-lander, and adding CA and AU after starting US-only. That pattern is consistent with scaling, but it still would not show spend or profit. (When this article was written in June 2026, our continuous-observation window topped out at around 28 days per creative, so "longest running" meant within that window, not that an ad existed for only 28 days.)

Health, finance, insurance and ecommerce offers were common in our June 2026 captures. A common explanation is that these categories can carry high payouts or lifetime value, but creative counts do not show margins, profit or which category is being scaled, and the category totals in the earlier version of this article are withdrawn because no reproducible extract is available. Our vertical overview, Top Native Ad Verticals in 2026: Nutra, Finance, Crypto & Sweeps by the Numbers, describes these categories in more detail.

The product categories that reward this analysis most#

It helps to track two levels separately: the broad category, which tends to change slowly, and the specific product or angle inside it, which can change within weeks. We have not published a measurement of either rate, so treat these as working assumptions to check against your own captures.

Health, supplements and wellness offers appeared often in our June 2026 Taboola captures. Some use recurring billing, which can make sustained spend affordable, but we have not measured that, and the per-network count in the earlier version is withdrawn. Watch which condition or angle rotates through the creative. This "Nebroo" hearing-device ad was seen in the capture across 26 days, near the top of the observation window at the time. Long observation does not show that its claims were substantiated, that it was approved, or that it was profitable.

Taboola health ad: Americans are ditching hearing aids for this new device
Caption: A Taboola health ad, 'Americans Are Ditching Hearing Aids for This New Device,' seen in OpenAdLibrary's capture across 26 days, June 2026 (historical capture; the ad's claim is not verified).

Personal finance, insurance and lead-generation offers were also common in our June 2026 captures on both Taboola and Outbrain, often with many headline variants. Lead payouts can be high in these categories, which is one plausible reason for long, broad campaigns, but captures do not show payouts or margins. The per-network counts in the earlier version are withdrawn because they cannot be reproduced from a published extract.

Ecommerce and gadget offers can turn over quickly: a product may run heavily for a season and then disappear. The "Consumer World" AC-unit ad below ("Does This $138 AC Run On Almost No Power?") had been in our capture for 5 days when this example was taken in June 2026. At that point its footprint could not tell a test from the start of a larger push.

Taboola ecommerce gadget ad testing a cheap air conditioner
Caption: A Taboola gadget ad, 'Does This $138 AC Run On Almost No Power?', after 5 days in OpenAdLibrary's capture, June 2026 (historical capture; the ad's claim is not verified).

Sweepstakes, survey and reward offers ask little of the user up front, and in our June 2026 captures some of them appeared on many placements. The IQ-quiz offers from "My IQ" are an example: several variants reached the 28-day observation ceiling at the same time. That shows sustained, broad distribution in the capture, not conversion rates or profit.

Software, app and subscription offers are another category to watch, since subscription revenue can justify a longer payback period. The software count in the earlier version is withdrawn because no reproducible extract is available, and creative counts would not show how these offers perform.

For named advertisers observed in these categories, see the network breakdowns: Who Advertises on Taboola? Top Native Advertisers by Vertical and the cross-network view in Top Native Advertisers on Taboola, Outbrain & MGID (Network Share Breakdown). Read them alongside this method to build a named-competitor watchlist, keeping in mind that their counts describe captures on given dates, not market share.

What longevity can and cannot tell you#

Longevity is the easiest signal to read, which makes it easy to over-read. On CPC pricing an advertiser pays when its ad is clicked, so a creative that keeps appearing for eight weeks may be profitable. It may also be a brand campaign judged on reach, a fixed test budget, a slow optimisation cycle, an ad nobody has paused, or an artifact of when and where the capture looked. Long-running ads are also a survivor sample: you see the survivors, not the many similar ads that stopped. Run length tells you that an ad kept appearing. It does not tell you spend, click-through rate, conversion rate, ROAS, profit or policy approval.

Studying the longest-running native ads is still useful, as long as you treat them as survivors in a capture rather than proven winners. In June 2026 the top of our 28-day window included SmartAsset running an IRA-tax angle on Outbrain, Hidden Hearing promoting hearing aids on the Microsoft Audience Network, and many "My IQ" quiz variants: finance, health and reward-quiz offers. When a new product starts appearing among long-running, widely placed ads, that is a good moment to start watching it, not proof that it is winning.

Longevity also guards against hype. A product can spike across many placements for a week because a buyer put budget into a test, then disappear. Spread without longevity is weak evidence. Spread plus longevity is a stronger lead, though still not proof of profit. The solar-battery ad below is an example of a long footprint: a home-energy lead-generation offer seen in our capture across 27 days in June 2026.

Taboola home and garden ad about solar home batteries
Caption: A Taboola home-energy ad, 'Solar home batteries: Electricians agree about 1 thing,' seen in OpenAdLibrary's capture across 27 days, June 2026 (historical capture).

The click trace tells you what the ad won't#

The creative tells you the angle. The destination tells you more about the business. Two ads with different headlines can lead to the same pre-lander and the same checkout, which often means one advertiser or one offer sits behind both, although affiliates promoting the same offer can share a checkout. Tracing each ad's redirect trail to the landing page, without clicking the live ad (so you don't add a click to the auction or cost the advertiser money), helps group those variants and gives a better view of an offer's footprint. The landing-page count in the earlier version is withdrawn because no reproducible extract is available.

The destination also tells you the mechanism: a quiz pre-lander, an advertorial, a straight product page, a free-plus-shipping offer. Pattern those across a category and you learn how its offers are being sold, which connects to the work in The Most Common Native Ad Angles (Analyzed From Real Creatives). For ecommerce specifically, watch for catalog-driven creative, Dynamic Product Ads (DPA), where one feed scales hundreds of SKU variants automatically. The variant explosion you see may be one system, not one hundred decisions.

How to run this study yourself#

You don't need our data to do this. You need a source that captures live ads and the discipline to read signals instead of headline numbers. Here is the repeatable procedure.

Define the window. Pick a date range; thirty days is a reasonable default, and state it when you report results. Segment by category and network. Filter to a vertical and a network so you compare like with like. The footprint for supplements on Taboola differs from gadgets on MGID. Sort by run length, then placement count. Surface the creatives seen longest and on the most placements. That is a candidate shortlist, not a list of winners. Cluster variants by advertiser and destination. Group creatives that share an advertiser identity or land on the same page. A high variant count under one advertiser is an iteration signal. Check for geo expansion. For your shortlist, look at whether new countries appeared over the window, and check that the capture covered those countries before the date they first appear. New countries are consistent with expansion but do not prove the offer is profitable. Read the destination. Trace the redirect trail to the landing page and note the mechanism. This helps separate one coordinated campaign from a coincidental cluster. Refresh on a cadence. Re-run monthly or quarterly and compare snapshots: the category mix is the slow trend line and the specific products are the fast signal. Record what you could not observe, such as spend, clicks, conversions, profit and approval status.

One note on coverage and honesty: any capture-based method only sees what it captures. Regional coverage varies, and an offer running in a market you under-sample will look smaller than it is. The EU's Digital Services Act (Article 39) requires providers of very large online platforms and very large online search engines that show ads to keep public ad repositories, which can be a cross-check for those services. None of the native networks covered here (Taboola, Outbrain/Teads, MGID or Revcontent) was on the European Commission's list of designated services as of its 31 August 2026 update, so their ads are not in those repositories and capture remains the main way to observe them. State your coverage limits when you publish. A data study that pretends to omniscience is less credible, not more.

For the wider context on how native spend, formats, and networks are shifting, including the post-merger landscape where Outbrain and Teads now run as a single combined company, see our pillar, The State of Native Advertising 2026 (Data From Live Ad Capture). This study is the tactical, refreshable companion to that overview. The pillar gives the broad picture. This article gives you a method for spotting which offers are gaining distribution in your own dated captures.

From signal to action#

Spotting a likely scale-up is step one. What you do next matters more: study the angle, test your own version with your own substantiated claims, and measure the result in your own account. Capturing the real creative helps because you are studying the actual asset that ran, not a thumbnail or a description of it. It does not tell you that the asset won the auction or made money; only your own test can answer that for your offer.

OpenAdLibrary supports this workflow. It captures native ads across Taboola, Outbrain, MGID, Revcontent and other networks and follows each native ad's redirect trail to its landing page without clicking the live ad. The free plan ($0, no card) shows the 20 newest Meta, TikTok and native ads, with ad detail, the supply chain behind native ads, and redirect trails and captured landing pages. The search, filters and saved research this method depends on, plus funnel tracing and real-advertiser resolution for native ads, require the Founders plan ($29.99/month).

Frequently asked questions

How can you tell what products advertisers are scaling without spend data?
You can't know for sure from outside the account. What you can do is read four observable signals in captured ads: longevity (how long a creative kept appearing), spread (how many placements and networks carried it), iteration (how many near-duplicate variants appeared) and geo expansion (whether new countries appeared). An offer seen for weeks, on many placements, with fresh variants is consistent with scaling; one that disappears within days may have been a test, a small budget or a capture gap. None of these signals measures spend, conversions or profit.
Are spend estimates from native ad spy tools accurate?
Not as measurements. Spend figures from third-party tools are modeled, because no third party sees a competitor's invoices or bids; treat them as rough estimates at best. Longevity, spread and variant counts are observed directly in captures, though they carry their own coverage limits and still say nothing about spend.
How often do trending native ad products change?
We have not published a measurement of turnover rates. A reasonable working assumption, which you should check against your own captures, is that the specific angle or product inside a category changes faster than broad categories such as health and supplements, finance, insurance, ecommerce gadgets and sweepstakes. That is why this kind of study should be re-run on a cadence and compared snapshot to snapshot.
What is the difference between a product being tested and a product being scaled?
A likely test is short-lived: one or two creatives, a handful of placements and a single country, seen for a few days. That fits a buyer checking whether an offer converts, but also a small budget or a gap in the capture. A likely scale-up is the opposite: sustained run time, the same offer across many placements and networks, several variants and new countries over time. Neither footprint proves intent or results; it tells you what to watch.
Can I run this analysis myself in OpenAdLibrary?
Yes, with the Founders plan ($29.99/month), which includes search, every filter and saved research across Meta, TikTok and native ads, plus funnel tracing and real-advertiser resolution for native ads. You can filter captured native ads by network, category, country and days running and sort by placements, then read the longevity and spread signals for yourself. The free plan ($0, no card) shows the 20 newest Meta, TikTok and native ads with ad detail, the native supply chain, redirect trails and captured landing pages; it does not include search or filters.
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.