Sprints & Retros for Taboola and Outbrain Media Buying (2026)

You launch a campaign on Taboola or Outbrain, it turns profitable at a $50 CPA, and then you stare at the dashboard wondering what to do with the rest of your day. You see other media buyers writing, commenting, posting all day, and it feels like everyone has work you don't. That's not a sign you've made it. It's a sign your optimization process has a hole in it.

A profitable campaign is the floor, not the ceiling. A $50 CPA that's "okay, a bit profitable but nothing crazy" is exactly the campaign that can be pushed to $35, then $33, if you treat optimization like an engineering discipline instead of a morning coffee ritual.

Marcel Sattler, founder of native-advertising.net, has deployed over $100M across Taboola, Outbrain, Newsbreak, MGID, Yahoo Native, Mediago, and RevContent since 2015, and the system he uses to grind CPA down week after week is borrowed straight from software development: Sprints, Retrospectives, and a change log. Here's exactly how Marcel runs it.

Why a profitable native campaign is not a finished campaign

Native advertising is slow. Pressing F5 on a Taboola campaign every two minutes does nothing except waste your day. Traffic sources like Taboola and Outbrain need time to gather data before any number means anything, so the obsessive refreshing you see from new buyers is pure anxiety, not optimization.

That slowness is also why so many buyers quit too early. The campaign goes green, they lay back, sip a tequila on the beach, and call it done. But the median CPA across a whole campaign hides the truth: inside that $50 average sits one ad pulling a $35 CPA and another bleeding at $70. The average is a blend, not a verdict.

The goal is to systematically find the $35 ad, kill the $70 ad, and then beat the $35. If you run $500 a day, optimization isn't a full eight-hour job, and that's fine. But it is a few hours every week, every week, indefinitely. Higher budgets push it toward full-time. Either way, "profitable" is the starting baseline, not the exit.

Set a baseline before you change anything

You cannot improve a number you haven't written down. Before touching a campaign on Taboola, Outbrain, MGID, or RevContent, set a baseline: the KPIs that actually matter to you.

Depending on your vertical, the baseline KPI is one of these:

• CTR
• Cost per click
• Cost per acquisition (CPA)
• Return on ad spend (ROAS)

Pick the one that drives your business. For a DTC or ecommerce play, CPA or ROAS usually is the number. For a lead-gen account, cost per lead. Say the campaign-level CPA is $50 right now. That's your campaign baseline, the number every Sprint has to beat.

Then go one level deeper. Inside that single campaign you're running multiple creative directions, and they don't perform the same.

Find your winning creative direction first

A native ad is a headline plus an image. Before you iterate, you need to know which direction inside the campaign is actually carrying it.

Take the image angle. In a typical campaign you might be running three distinct directions:

• A vector graphic or illustration, not a real photo
• The product on its own
• The product shown together with a human

Analyze the three. Often the "product plus a human" ad is the one doing the work, posting a $35 CPA while the campaign median sits at $50. That gap exists because $50 is the blended average across every ad, including the $70 laggards. The $35 belongs to one specific headline-and-image combination.

That winning combination is now your real baseline. Not the $50 campaign average, the $35 winner. This is the picture you build the next Sprint around. You can run the same exercise on headlines, and in practice Marcel optimizes the full ad, the headline and image together, since that's the unit that actually buys the click.

Run weekly Sprints with lookalike creatives

Here's the loop. Native is slow, so the Sprint is one week long. Week one establishes the baseline; week two attacks it.

In week one, your best ad is the woman-plus-product image at a $35 CPA. For week two, you stay in the same winning campaign and build lookalikes of that winner. If the winner was a woman, the lookalikes are:

1. A man
2. A family
3. Other close variations on the same proven angle

Let it run a week, then read the results. Say the man comes in at $38 and the family at $33. The family beat the previous $35 baseline. Now the family is the new baseline, and the next Sprint iterates on it: a family with a dog, a family with two kids, a family with two cats. Same proven structure, fresh variations.

This is the engine that keeps you out of ad fatigue. Ad burn happens when you run one creative until the audience is sick of it and performance rots. By continuously rotating lookalikes of the winner, the creative stays fresh while the performance stays strong. A few notes on running this cleanly on Taboola and Outbrain:

• Don't always stay in the same campaign. Whether you keep iterating in one campaign or split out depends on the traffic source. In the $35-to-$33 example it made sense to stay put; on other sources it won't.
• One variable per Sprint. If you change the image, hold the headline. You can't read a Retro when two things moved at once.
• One week per Sprint, because native data needs that long to be trustworthy.

If you run Taboola or Outbrain at scale, this is the operating rhythm we use on client accounts at our Taboola agency and Outbrain agency desks.

The Retro: read the data, don't guess

The Retrospective is the half of the loop that most buyers skip. At the end of each one-week Sprint, you sit down and read what the data says, not what you hoped would happen.

The Retro answers one question: which variation won? The man at $38, or the family at $33? The data decides, and your opinion about which creative "feels" stronger is irrelevant. The family won, so the family becomes the baseline for the next Sprint.

This is the discipline that separates a buyer who drifts from one to who compounds. Every week you either beat the baseline or you learn that a direction is a dead end. There's no week where nothing is being tested, and there's no week where a decision gets made on a gut feeling instead of a CPA number.

Keep a change log so you can roll back

Borrow one more tool from software: the change log. Write down every change you make, every Sprint, the way developers note what shipped in each release.

The change log earns its keep when a Sprint goes the wrong way. Say you swap in the family lookalikes and the CPA climbs back toward $50 instead of dropping. Without a record, you're guessing about what you altered. With a change log, you roll back to the exact state the campaign was in before, the $35 winner, and you're protected.

If you're serious about buying native traffic, this is non-negotiable, whether you're running affiliate offers, ecommerce, or lead-gen. Every change is reversible, every winner is preserved, and a bad Sprint costs you one week, not your whole campaign. The combination of a logged baseline, weekly Sprints, and honest Retros is what turns a $50 campaign into a $33 one without ever blowing it up.

Watch the full breakdown

Is your account a fit for the same play?

This framework scales. At $500 a day it's a few hours a week; at higher budgets it becomes a full-time operating system, and that's where most solo buyers run out of hours. The Sprints, Retros, and change-log discipline are exactly how we keep cutting CPA across Taboola, Outbrain, MGID, RevContent, and the rest, week after week, on accounts we manage.

If you're running native traffic and your campaigns are profitable but stuck, that's the best time to talk, because there's a $35 ad hiding inside your $50 average right now. Book a strategy call, look through our case studies to see the kind of results this process produces, or browse the full library of breakdowns in our resources.