Native Ads Campaign Structure: How Many Campaigns and Creatives to Run

The reliable default for native ads: one campaign per offer, geo, and device combination; five to ten creatives per campaign at launch; cut to the two or three earners once each variant has meaningful spend; and refresh before fatigue sets in. Across OpenAdLibrary's index — 725,000+ creatives from 29,000+ advertisers across 49 networks (June 2026) — the average advertiser runs about 25 creatives, and the advertisers who stay live for weeks structure theirs as deliberate angle clusters rather than one-off uploads.

Why structure matters: the campaign is the learning container#

On Taboola, Outbrain, MGID, and the rest, the campaign is the unit that bidding algorithms learn on, budgets attach to, and reports aggregate by. Put UK desktop and Brazil Android into one campaign and every number you see is an average of two unrelated markets — and any automated bidding is forced to optimize a fiction. Native networks give you fewer targeting dials than social platforms do, which means structure is your targeting. It's the discipline the native media buying guide treats as fundamental, and it's cheaper to get right on day one than to migrate later.

What forces a new campaign (and what doesn't)#

The rule of thumb: split when the economics of the audiences differ; don't split just to organize your dashboard. Every split divides your conversion data, and over-split accounts end up with fifteen campaigns none of which has enough signal to learn from.

How many creatives per campaign#

Launch with five to ten, built angles-first: three or four genuinely different angles, each in one or two executions, beats ten variations of a single concept. The batch exists to answer "which story works?" before "which photo of that story works?" — the sequencing covered in depth in the angle research workflow.

• Judge in two passes. CTR first, because it sets your effective click price; then conversion rate on the CTR survivors, because clicks that don't convert are just expensive applause.
• Keep two or three alive. Concentrating delivery on proven creatives is what the platform's rotation wants to do anyway; your job is making sure it happens after a fair test, not before.
• Replace in pairs. As creative fatigue erodes a winner's CTR, introduce replacements two at a time so the campaign never depends on a single aging image.

What creative counts look like in the index#

The index average — roughly 25 creatives per advertiser — hides the pattern that matters: advertisers who persist cluster their creatives around a small number of concepts, while churners upload one or two ads and vanish. Live examples from the index (June 2026):

• Perpetual Ad Tech runs three simultaneous angles for one service on MGID — a cost anchor, an agency-fee comparison, and a time-loss frame — a structured angle test, not a pile of uploads.
• Loop of Now keeps four or more live headline framings of a single "house cleaning rates in New Zealand" concept running on MediaGo at once: one concept, systematic variation.
• Flight Centre runs the same Teads ski-holiday creative at two different package prices simultaneously — structure used for price discovery.
• Tri-Lift ports one validated angle construction across Taboola and MGID against different skin concerns — a cluster that travels as a unit.

The interpretation: creative count follows conviction. A new concept gets a small cluster; a validated one gets variations, price tests, and ports to new geos and networks.

A reference structure to copy#

A concrete starting shape for a DTC offer targeting the US and UK, mobile-heavy:

• Campaign 1 — US / mobile / automated bidding. Six creatives: three angles, two executions each. Conversion postback on purchase.
• Campaign 2 — US / desktop / manual CPC. The same six creatives; desktop auctions price differently, so let bids differ.
• Campaign 3 — UK / mobile / automated bidding. Seeded with US winners after a localization pass — currencies, idiom, proof points.
• Campaign 4 — US / mobile / whitelist / manual CPC, higher bids. Publishers graduated from Campaign 1 once they've proven conversion volume — the whitelist pattern.

Name campaigns mechanically — network, geo, device, offer, version — so reports sort themselves. On budget: set each campaign's daily budget high enough that every live creative can collect a meaningful number of clicks per day; if it can't, run fewer creatives rather than starving all of them. The budgeting math lives in how much do native ads cost.

Split spend deliberately between discovery and exploitation, too. A common allocation once a whitelist campaign exists: the majority of budget on the whitelist, where economics are proven, and a steady minority on discovery, which exists to find the next publishers worth graduating. Buyers who let discovery starve stop finding new inventory; buyers who never graduate winners keep paying test-tier attention to placements that earned better.

Scaling without breaking the structure#

Two directions, different risks. Vertical scaling raises budgets and bids inside proven campaigns; horizontal scaling replicates the structure into new geos, devices, publishers, and networks — the trade-offs are mapped in horizontal vs vertical scaling. Three structural rules keep scale from destroying what worked:

• Duplicate winners into whitelist campaigns instead of inflating discovery campaigns; discovery stays cheap and broad, proven inventory gets aggressive bids.
• Port clusters, not creatives. What won as an angle cluster should travel as one — the Tri-Lift pattern above. A single creative transplanted alone tells you nothing when it fails.
• Enter new geos deliberately. Shortlist targets by competition and cost using the approach in scaling to new geos, and give each its own campaign from day one rather than bolting it onto an existing one.

Structure mistakes that cost real money#

• The mega-campaign. All geos, all devices, one budget: every metric becomes an average of unrelated markets.
• Over-splitting. Fifteen campaigns with three conversions each teach the bidding algorithm nothing. Consolidate small same-tier geos until volume justifies separation.
• Twenty creatives on a thin budget. Each gets too few clicks to judge; the whole batch reads as noise.
• Restructuring weekly. Every rebuild resets learning. Restructure when the data demands it, not when the dashboard bores you.
• Judging on CPA too early. Prune on CTR first; conversion verdicts need conversion volume.
• Treating structure as permanent. The right structure at launch and the right structure at scale are different structures — revisit quarterly as conversion data thickens and consolidate or split accordingly.

Read competitors' structure before building yours#

An advertiser's public creative set reveals its structure. Search a competitor in a native ad spy tool, group their live ads by concept, and note how many angles they sustain, how many variants each angle carries, and which geos and devices the captures show — the full method is in how to spy on competitor native ads. You're not copying ads; you're copying the shape of a structure the market has already rewarded, then filling it with your own creative.