July 27, 2026

What Should a Retargeting Ad Say? (Everything the Cold Ad Could Not)

Your retargeting ad should say what the cold ad could not. Not an introduction, an answer to the one objection that made a warm viewer leave.

A retargeting ad should say something the cold ad could not. The person seeing it already knows who you are and what you sell. They stopped anyway. So the job is not to introduce you. It is to name the one objection that made them leave and answer it, in that ad, with proof.

Most retargeting creative is the cold ad with the budget moved.

That is the actual problem in most accounts. Not the audience, not the window, not the bid. Your retargeting underperforms because you are showing a stranger's ad to someone who is not a stranger, and the mismatch reads as either noise or desperation.

What makes a warm viewer different from a cold one?

They have already answered the three questions your cold creative exists to answer, and they left anyway. Who is this, why should I care, what is this thing. All three are settled. So the question they are sitting with is a fourth one, and it is harder to write to: will this actually work for someone in my situation?

That is a different genre of writing. Cold creative earns attention. Warm creative earns belief, and belief is not won by saying the same thing with more emphasis. That distinction is why the layer exists at all, and why retargeting fails when it is treated as a second attempt at the same pitch.

The parent guide lays out the three principles this post builds on: proof over hook, objection over introduction, specificity over breadth. What follows is the sentence-level version, because those principles are easy to agree with and hard to execute at the moment you are staring at a blank ad.

Here is the operating rule for everything below. The first line of your cold ad is the first line your warm ad has to delete.

What should a retargeting ad actually say?

Whatever you were holding back because it was too much detail for a stranger. Four rewrites, cold on top, and what changed underneath.

1. From category to artifact

Cold: "There is a better way to handle client onboarding."

Warm: "The onboarding doc is the whole system. Here is mine, filled in, including the two fields everyone leaves blank."

The cold line asserts that a better way exists. A warm viewer already accepted that. What they have not accepted is that yours is real. Showing the artifact answers a question that asserting the category cannot.

2. From feature to named objection

Cold: "Works for teams of any size."

Warm: "You are one person and you assumed this was built for teams. It was not, and here is the part that changes when you run it solo."

The cold line is written to exclude nobody, because you do not know who is reading. In warm, you do. The pages they visited told you. Breadth is what you write when you are guessing, and once you have behaviour, guessing is a choice.

3. From invitation to the thing itself

Cold: "Learn how to build your first agent."

Warm: "Here is the agent, running, on a task I did not clean up first."

An invitation to learn is a hook, and hooks are for people who have not stopped yet. This person stopped once already. The open question is whether you can actually do the thing, and the unedited version answers it in a way the polished version never will.

4. From instruction to account

Cold: "You need a capture layer before you spend on ads."

Warm: "I ran cold traffic for weeks without one. What that cost me was not the ad spend."

Instruction delivered to someone who already agrees with you is nagging. The same idea told as your own account gives them something new to evaluate instead of something to agree with a second time.

Every one of those rewrites does the same operation. It deletes the part the viewer already has and spends the recovered space on the part they do not. Nothing is added. Something is removed and something more specific fills the hole.

The rewrite deletes what the viewer already hasColdLearn how tobuild your first agent.They already have thisdeleteWarmHere is the agent,running, on a taskI did not clean upfirst.Invitation → the thingitself. Nothing added.

Warm ad brief

Prompt
One ad per brief. If you cannot fill line 3 in a single sentence, you are not ready to write this ad yet.

1. Audience: the exact behaviour that put this person in the pool
2. Already known: what they saw or read before they left (never repeat this in the ad)
3. The one objection: stated in their words, not yours
4. The proof: the specific thing that answers line 3, not a reassurance about it
5. The next step: smaller than buying

Then check:

- Does the first line survive if you delete everything a cold viewer would need? If not, rewrite it.
- Is the proof in line 4 a thing you can show, or a claim you are making? Show beats claim.
- Is there a second objection hiding in the ad? Cut it. It belongs to a different ad.
- If line 3 is not price, is there a discount anywhere in this ad? Remove it.

Why doesn't the discount work?

It works often enough to feel like it works. What it actually does is convert the people who were going to buy anyway, at less, and teach everyone else that patience gets rewarded.

Go one level past that. A discount aimed at a warm audience carries a message you did not intend to send: it says the reason you did not buy was price. If price was not the objection, you have just announced that you were not listening. The person whose real hesitation was "I do not think I have time to set this up" now knows you never understood the problem.

Then there is the audience you already sold. Your warm pool contains full-price buyers, and they are the most attentive people you have. They see the promotion. What they learn is that the correct play was to wait, and they will remember that the next time you ask for money.

And it prices your next launch. A discount is not a temporary event to the people who saw it. It becomes the reference point. The next thing you release opens against your own last discount rather than against your list price, and every launch after that inherits the same anchor.

The narrow case where a discount is correct: price is the actual, named objection for that specific segment, and you can point to where you learned that. That is rare enough that it should feel like an exception when you write it.

What do you show someone who abandoned a checkout?

Show them the answer to the question the checkout page did not answer, and send them straight back. This is the highest-intent audience you have, and the counterintuitive part is that they need the least selling of anyone in the pool.

They already decided yes. They saw the price and started anyway, which is why price is almost never what stopped them. Something small did, and it is usually one of five things.

What happens after I pay. They cannot picture the next ten minutes and it made them hesitate.

Whether it fits what I already run. A compatibility question the page never addressed.

Whether I can get out. Cancellation, refunds, commitment length.

Whether anyone is actually there. No visible sign a human exists on the other side.

Nothing at all. A field that would not accept their input, a payment method that failed, a phone call that ended the session. A real share of abandonment is mechanical, and those people do not need persuading, they need a working link.

What actually stops a checkoutPriceWhat everyone assumes.They started anyway.What actually stops itAfter I paywhat happens nextCompatibilitydoes it fit what I runCan I get outrefunds, cancellingAnyone thereis a human presentNothingit just broke

So the ad is short, it does not sell, and it opens with the answer rather than the reminder. Not "you left something behind." Something closer to: "You stopped at the payment step. If you were wondering what actually happens in the first ten minutes after you pay, here it is." Then the link.

One question per ad. If you know two of the five are common for you, that is two ads, not one longer one.

Do not discount this segment. Paying someone to do the thing they were already doing is the most expensive way to close a sale you had.

How many ads does a warm audience need?

More than cold does, and that is the opposite of what most people assume. A big cold audience spreads impressions across a lot of different people. A small warm pool concentrates them, so the same faces carry the entire load, and the ad wears out on them at a speed cold never experiences.

Which means the relationship runs backwards from intuition. The smaller your warm audience, the more creative variety it needs, not less. A tight, high-intent segment is exactly the one that will see your ad most often.

So build a set rather than an ad. Three or four pieces, each answering a different objection, rotating. The set is also more useful diagnostically, because when one of them pulls responses the others do not, you have learned which objection is actually live.

You can tell fatigue is setting in from behaviour rather than a number. Reported frequency climbing quickly. Response drifting down while delivery stays flat. And the clearest signal, which is free: the comments change character. Questions turn into people telling you they have seen this already.

The fix is not more budget. It is more creative, and the cheapest source of it is content you have already published. Every post, video, or thread where you answered a real objection is an ad that already exists in draft form. That is the practical link between an organic engine and a paid one.

All of this is downstream of audience size, which is why building the pool from assets you already own comes before any of this creative work. A pool too thin to rotate against will fatigue no matter how well the ads are written.

How do you know which warm message is working?

Slowly, and that is the honest answer. Warm tests take longer than cold tests for a structural reason: the audience is smaller, so the signal takes longer to separate from noise. Patience here is a requirement, not a virtue.

The measurement problem itself is not new ground. The pillar covers why campaign-level retargeting numbers flatter themselves, and the companion piece on what cold acquisition actually costs works through the attribution mechanics in detail. Both apply here unchanged.

The specific point for creative testing is narrower. Test one variable, and make it the objection rather than the wording. Two ads that answer different objections tell you something. Two ads that answer the same objection with different phrasing will take longer than you have to separate, on a pool this size.

The fastest readable signal is qualitative. Replies, comments, and messages arrive before the numbers do, and they tell you which objection you named correctly. That is not rigorous. It is available on day two, and the rigorous version is not.

Which brings the whole thing back to one idea. A warm ad is not a better version of a cold ad. It is a different document, written to a person you already have information about, and the information is only useful if you let it delete something.

If you want the full picture of how the content engine, the capture layer, and the creative fit together as one system rather than three disconnected projects, I walk through the whole thing in a free masterclass. No pitch to sit through, just the mechanics.

Frequently Asked Questions

What should a retargeting ad say?
Something the cold ad could not say. A warm viewer already knows who you are and what you sell, and they left anyway. So the ad's job is not to introduce you. It is to name the one objection that stopped them and answer it directly, with the proof you were holding back because it was too much detail for a stranger.
Should retargeting ads offer a discount?
Only when price is genuinely the objection that stopped that specific audience. Otherwise a discount answers a question nobody asked, and it tells your best-informed audience that waiting is the right move. It also converts the people who were already going to buy, at less, and it sets the reference price your next launch has to argue against.
How often should you change retargeting creative?
More often than cold, because a warm pool is smaller and the same people carry the whole impression load. Rather than chasing a number, watch behaviour: reported frequency climbing quickly, response drifting down while delivery stays flat, and comments turning from questions into people telling you they have seen it already. Build a rotating set of ads rather than one ad.
What do you show someone who abandoned a checkout?
The answer to the specific question the checkout page did not answer, and almost never a discount. They already saw the price and started anyway, so price is rarely what stopped them. It is usually something small: what happens after they pay, whether it works with what they already use, whether they can cancel, or a mechanical failure on their end.

Related Articles

I'm building this in public. Come build with me.

The Sprint: Focus Pilot, live weekly mentorship, and a community of operators who ship with AI.

marketingpaid adsretargetingfunnel architecturecontent marketing
Matt Ganzak

Matt Ganzak

Founder, The Sprint & ScaleUp Media

25+ years building software companies. Multiple SaaS exits. Bestselling author of The Million Dollar Plan. Writes about running AI agents for real operational work.