Competitive Intelligence for Media Buyers: A Weekly Research Routine
The buyers who reliably spot what is worth testing do not spy only when they are stuck; they run a 45-minute weekly routine that turns scattered ad-checking into pattern recognition you can act on.

Most media buyers open a spy tool the way they open the fridge at midnight: only when something is wrong. The campaign is flat, the angle is tired, the account manager wants ideas by Friday. So you browse competitor ads for an hour, screenshot a few, and move on. Next time you do it, you have no memory of what you saw last time, so you cannot tell what actually changed.
That gap is the whole difference between ad spying and competitive intelligence. Spying is a snapshot. Intelligence is the time series. The buyers who reliably spot promising angles before everyone else are not better at staring at ads. They have a routine that turns weekly snapshots into pattern recognition. This guide is that routine: a 45-minute weekly cadence for native media buyers that surfaces new entrants, fresh creatives, and geo expansions, plus a simple log structure that makes the patterns obvious over a month.
What a weekly review actually buys you#
Run a fixed weekly review of competitor ads and you stop reacting to individual creatives and start reading the market. New advertisers entering your vertical, creatives that stay in rotation for weeks, and offers appearing in new countries all show up as changes between snapshots. They are leads you can follow up before they are obvious to everyone scrolling the same feeds, though none of them proves that an advertiser is making money.
The reason weekly works for native is rhythm. Many native buyers review tests on a cycle of a few days to a couple of weeks (this varies widely by account and budget, and we have no measured figure), cutting some creatives and scaling others. Check daily and you drown in tests that will be gone by Thursday. Check monthly and an angle may already have been copied by a dozen affiliates. Weekly is a practical compromise that roughly matches the buying it observes.
You are working in a haystack: OpenAdLibrary describes its library as millions of Meta, TikTok, Google and native ads, many of them observed repeatedly. You are never going to look at all of that. The point of a routine is to make a tiny, deliberate slice of it useful every week.
This is the operational layer of ad intelligence, the discipline that compounds rather than the one-off look. If you want the strategic case for why this matters across native, display and social, the pillar guide on how to spy on competitor ads in 2026 frames the whole landscape. This piece is the narrow, repeatable execution.
The three signals worth watching every week#
You cannot watch everything, and trying to is why most monitoring habits collapse after two weeks. Pick three signals. These are the ones that actually change buying decisions.
1. New entrants#
A new advertiser appearing in your vertical is often the most useful signal to log, and one many buyers miss, because they only check the same five competitors they already know. A new name may mean a fresh offer or a player testing your space, and the captures cannot show whether it is working or well funded. Either way you want to know in week one, not when their creatives are everywhere.
Track this by watching the vertical, not just named brands. If you buy in a busy category, unfamiliar names will appear regularly, and a fresh name with a polished funnel is worth a closer look. Filter for your category, sort by first-seen date, and anyone you do not recognise gets a name, a note, and a place on the watchlist (search, filters and saved research are on the OpenAdLibrary Founders plan). The companion guide on building a competitor watchlist for ad monitoring goes deeper on structuring this so it stays maintainable.

That ad above is the kind of observation worth logging: a tax-relief offer with a deadline hook, already observed running for 13 days when we captured it. Thirteen days is a short span. It shows the ad was in rotation that long, not that the math was working.
2. Creative refreshes#
When a competitor swaps creative on an offer that has been running, that is worth noting. The old angle may have fatigued (useful: note what they abandoned), they may have found a new hook worth scaling (more useful: note what they replaced it with), or they may be running a routine test. The pace of refresh is itself an observation. An advertiser cycling through ten new images a week looks like it is testing heavily. One running the same three for a month may be keeping a control, or may simply not be refreshing; the captures cannot tell you which.
Health and senior-finance angles can refresh often, and the headlines show which framings advertisers were running, though not which of them worked. The two below, captured in June 2026, are examples of the curiosity-gap formula.


Both had been seen for only three days when captured. That is consistent with testing, though a capture can only show when an ad was first seen, not when it launched. When you log the headline, the angle and the date, a pattern of which framings are still in rotation after two weeks can start to emerge across the vertical.
3. Geo expansion#
An offer showing up in a new country is worth logging. Advertisers expand for several reasons: a test in a new market, a different affiliate or brand running the same offer, a change in how the capture samples that country, or a campaign that is working. When a creative that was only seen in the US shows up in the UK, Canada or Australia, it suggests the advertiser decided to try those markets. It does not show that the unit economics travel. It is a lead you can evaluate for your own geos, and a reason to look hard at the landing page, because the localisation choices show how seriously they are taking the new market.
| Signal | What you log | Why it matters | Action it triggers |
|---|---|---|---|
| New entrant | Advertiser name, vertical, first-seen date | Possible new spend or a new offer to test | Add to watchlist, profile the funnel |
| Creative refresh | Old vs new creative, refresh date, cadence | Possible fatigue, or a new hook being tried | Compare old and new angles, test your own adaptation |
| Geo expansion | Offer, old geo, new geo, date | A market test or wider rollout; economics unverified | Evaluate the geo for your own buys |
The 45-minute weekly routine#
Block the same slot every week. The consistency is the point. Same time, same filters, same log, because comparison only works when the conditions are held constant. Here is the sequence.
- Scan new entrants (10 min). Open your vertical, sort by first-seen date, and review anything from the last seven days you do not recognise. Name the real advertiser behind each one. Native ads hide the actual brand behind tracker redirects, so a tool that resolves the supply chain to the genuine advertiser saves you the manual detective work.
- Review your watchlist (15 min). Go advertiser by advertiser through your tracked competitors. For each one, ask three questions. Any new creatives since last week? Any creatives that disappeared? Any new geos? Log the deltas, not the full picture. You only care about what changed.
- Pull the long-runners (10 min). Sort your watchlist's ads by how long they have been observed. Anything running for several weeks across multiple publishers and geos is a candidate worth studying, not a confirmed winner. These are your study targets. Longevity and spread are observations you can check yourself, which an impressions estimate cannot offer, but neither shows performance.
- Trace one funnel end to end (10 min). Pick the single most interesting ad of the week and follow it from creative to landing page to offer. Do not click the live ad. That costs the advertiser money and can skew your own retargeting and analytics. Study a capture that already followed the click for you instead. The deep mechanics of this are in the guide to reverse-engineering a competitor's native ad funnel.
That is it. Forty-five minutes, and the only output you carry forward is your log.
The log: where snapshots become patterns#
Every snapshot you take is worthless in isolation and valuable in sequence. The log converts one into the other. Keep it dead simple, a spreadsheet works fine, with one row per observation:
- Date observed, so you can reconstruct the timeline
- Advertiser, the real brand, not the tracker domain
- Network: Taboola, Outbrain (now operating under the Teads brand after its 2025 acquisition of Teads), MGID, Revcontent, Yahoo, MSN
- Signal type: new entrant, refresh, geo expansion
- Creative note: the hook, the angle, the image concept
- Geo, where it ran
- Longevity: first-seen to last-seen, if available
- Link, to the captured ad and its landing page
After four weeks this log starts answering questions a single session never could. Which advertisers refresh creative fastest? Which angles appear across multiple advertisers, which may mean the angle is being copied or shared rather than proven? Which offers appeared in a new geo shortly before you saw copycats? That rhythm-detection is what separates a media buyer who reacts from one who anticipates.
A note on data hygiene: log the real advertiser every time. Native creatives route clicks through redirect chains, so the brand on the ad is not always the entity paying for it. If your log is full of tracker domains instead of real advertisers, your pattern recognition is built on sand. Resolving the supply chain to the genuine advertiser is exactly the gap that purpose-built ad intelligence platform tooling closes.
Reading longevity, not novelty#
A common mistake in this exercise is treating the newest creative as the best one. A newly seen ad has only a short record behind it. An ad that stays in rotation for weeks is worth a closer look, but it is not a proven winner: advertisers keep ads running for reasons other than profit, and the captures cannot show results.
You will read industry posts about "90-day winners" as a gold standard. That is general native lore: no run length has been shown to mark a winner, and it is not something our own data can show. What our data shows is the durable end of what we observed. Our index timestamps first-seen and last-seen on every creative, and in a June 2026 snapshot the longest continuous runs we had captured were about 28 days each. That figure reflects how far back our continuous capture reached at the time; it is not a limit on how long advertisers run ads. SmartAsset has an Outbrain finance ad ("Ask a Pro: How Can I Avoid Paying Taxes on IRA Withdrawals?") at 28 days. Combat Siege has a gaming ad at 28 days. Hidden Hearing and My IQ both have hearing-aid and IQ-quiz creatives at the same mark. When an advertiser holds a creative steady for nearly a month while other creatives rotate, it is worth recording. It may be a control the advertiser is protecting, a long-running test, a brand placement or an ad nobody switched off, and the captures do not show which.

The Nebroo hearing-device ad above had been observed running for 26 days when we last checked in June 2026. In a category where creative often rotates quickly, an ad that stays in rotation for nearly a month is a better study target than whatever launched yesterday. It does not tell you that the offer and the angle work.
Turning the log into action#
Intelligence that does not change a buy is a hobby. Each signal type maps to a specific move:
- A new entrant with a polished funnel means evaluate their offer for your own portfolio, or at minimum understand who is now competing for your placements.
- A creative refresh on a long-running offer means the surviving angle is a candidate. Adapt it (adapt, do not photocopy) and test it against your control; the competitor's results are not yours.
- A geo expansion is a lead that an offer may have room to grow. Whether it travels for them is unknown; whether it travels for you is something only your own test can show.
The cleanest way to keep these moves from getting lost is to wire the routine into a documented workflow rather than relying on memory. The guide to building a competitive ad intelligence workflow shows how the watchlist, the weekly log and the action queue connect into a single loop.
Why native suits a weekly routine#
Native is well suited to weekly intelligence because the public surface is informative. Native ads run on open publisher pages, so a platform that captures the placement can record the creative, the publisher it ran on and, where the redirects can be followed, the path to the landing page. Search and social formats have their own public libraries (Meta's and TikTok's ad libraries and Google's Ads Transparency Center), but each covers only its own platform, which is why native needs separate capture. Capture has gaps: you will not see every placement, and a captured landing page is not the whole funnel.
Regulation covers only part of this. Under Article 39 of the EU's Digital Services Act, designated very large platforms must maintain public repositories of the ads they serve. Taboola, Outbrain/Teads, MGID and Revcontent are not on the Commission's list of designated platforms (updated 31 August 2026), so their ads still need their own capture. A structured weekly routine works with whatever transparency sources exist. The detailed network-by-network mechanics for Taboola, Outbrain and MGID live in the guide on spying on competitor native ads.
Common ways the routine fails#
- Watching too much. Three signals, a focused watchlist, a fixed time box. Scope creep kills the habit by week three.
- Logging snapshots instead of deltas. The value is in what changed. Re-recording the full picture every week buries the signal in noise.
- Confusing novelty with performance. A newly seen creative has not been shown to be better or worse than an old one. A long-running creative is a better study target, but longevity does not show performance.
- Clicking live ads to see the funnel. It costs the advertiser money, can distort their reporting and can corrupt your own retargeting and attribution. Study captures instead.
- Tracking tracker domains. If your log does not name the real advertiser, you cannot connect a creative to a brand to a funnel, and the patterns never resolve.
Putting it together#
Competitive intelligence for media buyers is not a tool you buy or an hour you spend when you are stuck. It is a weekly habit: three signals, a simple log, and one funnel traced per session. Do it for a month and the market starts narrating itself. Who is new, what is staying in rotation, where advertisers are showing up. Skip it and you are back at the fridge at midnight.
OpenAdLibrary's library is built for this cadence. It covers Meta, TikTok, Google Ads and native ads, including native ads from Taboola, Outbrain/Teads, MGID, Revcontent, Yahoo and MSN. For native ads it records first-seen and last-seen dates, the geos where each ad was observed and the redirect trail to the landing page without clicking live ads, and the Founders plan resolves the real advertiser. The free plan shows the 20 newest Meta, TikTok and native ads with no card; search, filters, saved research and the API are on the Founders plan at $29.99 a month. Start free to see the newest ads, then plan your first weekly review.







