OpenAdLibraryOpenAdLibrary
Affiliate & Media Buying

Publisher-Level Bidding: Adjusting Bids by Site, Section and Widget

Performance can vary widely between publishers in one native campaign; your placement report shows how widely. Publisher-level bidding sets different bids per site, section or widget, with a review workflow and worked example.

Editorial illustration: Publisher-Level Bidding: Adjusting Bids by Site, Section and Widget

Publisher-level bidding means setting a different bid, or bid adjustment, for individual publisher sites, sections or widget IDs inside a native campaign, instead of running one flat bid across the whole network. On Taboola, Outbrain, MGID and Revcontent, performance can vary widely between publishers inside the same campaign, even after the algorithm has had time to learn; how widely in your account is something only your placement report can tell you. Adjusting bids where the evidence supports it is one way to stop paying the same price for placements that perform very differently.

Why one flat bid is the wrong default#

When you set a single bid across a campaign, the network applies that bid wherever your targeting matches, from large news sites to small content sites showing the same widget. Some of those placements may cost far more per conversion than your target; others may produce no conversions at all. A flat bid prices them identically. The networks' own tools let you change that per site (Taboola), per publisher section (Outbrain) or per widget or source (MGID and Revcontent).

Automated bidding only optimizes toward the signals it receives. If you bid on clicks, or your conversion tracking does not report back to the network, the platform cannot see which publishers deliver poor leads, and your own conversion data may lag behind spend. Outbrain's Conversion Bid Strategy adjusts CPCs toward tracked conversions, and Taboola's SmartBid adjusts bids automatically, but neither can see business signals you never send, such as lead quality, refunds or lifetime value. Publisher-level adjustments put those signals back into the bid; Outbrain's Section Volume Control exists for exactly that case.

What you're actually changing#

Every major network exposes this lever, just under different names:

Network Where it lives How the adjustment works
Taboola (Realize) By-site report in the campaign summary Percentage boost or de-boost on the campaign CPC per site: -99% to +100% on Fixed Bid, -99% to +200% on SmartBid; publishers can be blocked per campaign or for the whole account
Outbrain (Teads) Amplify reporting, By Publisher and By Section Percentage CPC adjustment per publisher section (not whole publishers); publishers and sections can be excluded temporarily or permanently; Section Volume Control for Conversion Bid Strategy campaigns with at least 300 clicks
MGID Widget Optimization (Selective Bidding) and Sources Optimization CPC multiplier per widget ID or source, minimum 0.1, within regional minimum and maximum CPCs; campaign and account blocklists
Revcontent Widget Targeting tab Absolute CPC bid per widget ID (or a percentage change), including CSV upload; widgets can be blocked for a campaign or the whole account

The mechanics differ. Taboola, Outbrain and MGID apply a percentage or multiplier to your base bid, while Revcontent's widget bids are absolute CPCs (with a percentage tool for bulk changes). Set a 1.4x adjustment on a strong publisher and a 0.6x on a weak one, and the network adjusts the effective bid it enters on each, within its own limits and minimum CPCs. All four also let you exclude a publisher or widget entirely, which is the blunter version of the same idea. On Outbrain, manual section CPC adjustments override Conversion Bid Strategy, and Outbrain recommends removing them from CBS campaigns. For the exact fields and API parameters each network exposes for this, see how Taboola ads work, how Outbrain works, how MGID native ads work and how Revcontent works.

Why it is easy to skip#

Three reasons, honestly:

  • The reporting takes digging. Publisher-level data usually sits a few clicks deep in a network dashboard, exported as a flat CSV with many rows and no obvious next action.
  • It looks like busywork for small accounts. If you're spending $200 a day, the instinct is that a handful of publisher exclusions won't move the needle. But at low spend a single publisher can take a meaningful share of the daily budget before you notice, so check the share each publisher takes in your own report.
  • It requires waiting for a real sample. Zero conversions on a small spend proves little. If a publisher were converting exactly at your target CPA, the chance of seeing no conversions is about 37% after spending one target CPA, about 14% after two and about 5% after three (Poisson arithmetic, ignoring conversion delay). That is why two to three times target CPA per publisher is a reasonable floor before judging a zero-conversion publisher, not a guarantee. Buyers who pull the report on day two see nothing conclusive and give up on the exercise.

That last point is the real cost. Publisher-level bidding is a discipline, not a one-time setup. It needs a standing review on a regular schedule.

A practical workflow#

  1. Let the campaign run to a real sample. Judge each publisher on its own spend, not the campaign's: a campaign that has spent twice its target CPA has too few conversions to compare publishers. Use the per-publisher floor above before acting on a zero-conversion publisher.
  2. Pull the placement or site report and sort by spend, descending. Start with the publishers that take most of your budget, and check what share of spend they account for in your account rather than assuming one.
  3. Bucket into three tiers: profitable (bid up), breakeven (leave alone), losing (bid down or exclude). How far to move each is your call; moving 20-50% at a time and measuring the result is an illustrative starting point, not a benchmark. MGID's own guidance says multiplier changes usually do not shift volume on small publishers and recommends raised rates of 2, 3 or 4 when you bid up. Don't try to fine-tune every row; the top and bottom of the list matter most.
  4. Apply adjustments, not full stops, on your first pass. Excluding a publisher outright loses the option to re-test it later; a reduced bid still lets you learn whether the problem was price or fit. Check your platform's limits first: MGID's lowest multiplier is 0.1, and Taboola's largest de-boost is -99%.
  5. Recheck on a schedule. Publisher performance shifts as new content is added and traffic mix rotates. Weekly is a sensible default for active accounts; adjust it to your conversion volume.
  6. Feed winners into scaling decisions. A publisher that's profitable at a 1.5x bid is also a candidate for a dedicated campaign, which is the horizontal scaling move once you've exhausted vertical gains on the original one.

A worked example#

A hypothetical example (made-up numbers, not a measured campaign): say you're running a $300/day campaign on Taboola with a target CPA of $40. After two weeks you have spent $4,200, and the placement report shows 40 publishers took meaningful spend. The top five publishers by spend produced 60% of conversions at an average CPA of $28. The bottom eight publishers took 25% of spend ($1,050) and produced two conversions in total, an effective CPA of about $525. Everything else sits in the middle, roughly breaking even.

In this example, an obvious move is to bid up the top five and cut the bottom eight to a reduced bid rather than excluding them outright, since one or two might just be under-sampled rather than genuinely bad. As a group they are far behind target, but check each publisher against the per-publisher floor before acting on it alone. Whether the campaign-wide CPA then improves is something to measure over the following weeks, not something this example can predict: shifting spend changes which auctions you win, and the network's own optimization reacts too. The case for treating publisher-level bidding as a standing task rather than a one-time cleanup is that the placement mix inside "the campaign" behaves like many separate micro-markets, and a flat bid prices them all the same.

Setting up a standing review#

A recurring calendar block works better than relying on remembering to check. A simple version: every Monday, pull the prior week's placement report for each active campaign, sort by spend, and spend 15 minutes per campaign adjusting the top and bottom deciles. Log the adjustments you set so you can tell later whether a publisher's improvement came from your bid change or from something else shifting in the account. Publisher IDs, report formats and adjustment tools differ between Taboola, Outbrain and MGID, so this review needs to happen per network, not once across the account.

What good publisher data looks like#

Your own conversion data is the evidence that matters here. Before you have spent anything on a publisher, outside research can tell you which advertisers appeared there and when, but not whether those placements converted: how long other advertisers' ads were observed reflects their budgets, tests and our capture coverage as much as performance (see what ad longevity can and cannot show). OpenAdLibrary's ad detail shows captured ads and, for native ads, the supply chain behind them; use that to choose publishers to watch, not to predict their quality. You can cross-reference captured creative history through the ad intelligence tool alongside your own account's data.

Common mistakes#

  • Bidding down too aggressively. A very low bid can fall below what wins on that publisher, which may stop delivery there and leave you nothing to observe. Some platforms floor it instead: MGID applies its regional minimum CPC, and Taboola applies its minimum CPC to site-level bids.
  • Reacting to one bad day. Native traffic is noisy day to day, especially on smaller publishers. A single day of zero conversions on $40 of spend is not a verdict.
  • Ignoring device split within a publisher. A publisher can be excellent on mobile and dead on desktop (or vice versa); collapsing the two into one bid adjustment hides that. Layering device and geo adjustments on top of publisher-level work, where the platform supports them, closes that gap.
  • Forgetting to re-test excluded publishers after a creative refresh. A publisher that failed with a tired angle may perform differently with a new hook. Outbrain's temporary exclusions, which expire after a set period, make this easier.

FAQ takeaways#

Publisher-level bidding is unglamorous account maintenance, but it is one of the few levers fully within your control, unlike auction dynamics or platform algorithm changes. We have no comparative data on how much it improves CPA, so measure it in your own account: log each change, compare the following weeks with the weeks before, and keep the changes that hold up. Whatever split you're using between testing new concepts and scaling proven ones, publisher-level bidding is how you make the scaling side more efficient rather than just louder.

Frequently asked questions

What is publisher-level bidding?
It's setting a different bid or bid adjustment for individual publisher sites, sections or widget IDs inside a native campaign, instead of one flat bid across the whole network. Taboola applies percentage adjustments per site, Outbrain per publisher section, MGID a multiplier per widget or source, and Revcontent an absolute CPC bid per widget.
How much spend do I need before adjusting publisher bids?
Spending two to three times your target CPA on a publisher is a reasonable floor before acting on zero conversions: if the publisher were converting at your target, you would see no conversions about 14% of the time after two target CPAs and about 5% after three (Poisson arithmetic, ignoring conversion delay). Judging a publisher that does have some conversions needs more data still, because small conversion counts are noisy.
Should I exclude bad publishers or just lower the bid?
Lower the bid first; how far is your call (0.5x to 0.7x of your base is an illustrative starting point, not a benchmark), within your platform's limits. A full exclusion removes the option to re-test that publisher later if your creative or offer changes, unless the platform offers temporary exclusions, as Outbrain now does. Reserve permanent exclusions for publishers with clear fraud or brand-safety issues, or ones that stay far off target after re-testing.
Does publisher-level bidding work the same on every native network?
The concept is the same but the mechanics differ. Taboola adjusts bids by site from its by-site report; Outbrain adjusts by publisher section (not whole publisher), and its manual section bids override Conversion Bid Strategy; MGID multiplies bids per widget or source; Revcontent sets absolute bids per widget. Check each network's current documentation for exact field names and API access.
How often should I review publisher performance?
Weekly is a practical default for many active accounts: daily reviews tend to react to noise, and monthly reviews can let losing publishers spend too long. Adjust the cadence to your conversion volume; we have no data on how often inventory mix shifts on each network.
The OpenAdLibrary Team
Written byThe OpenAdLibrary Team
Ad intelligence & native advertising research

We build OpenAdLibrary, the open ad-transparency platform. Every day our systems capture live native ads across Taboola, Outbrain, MGID, Revcontent, Teads, Yahoo and MSN, identify the real advertiser behind each one, and follow the click to its landing page. These guides distill what we see in that data so you can research the market faster.