Retargeting that doesn't feel like being followed around the internet

Retargeting is one of the highest-return channels in most paid media mixes, and it's also the channel most likely to actively annoy the exact people it's trying to convert. The line between "helpful reminder" and "why is this brand everywhere I go" is mostly about frequency and creative variety, both of which are easy to get wrong by default.

We inherit more retargeting accounts with a creepiness problem than a performance problem, and the two are more connected than they first appear: a campaign that feels oppressive to the audience it's chasing is usually also burning budget on impressions well past the point of any marginal benefit, which means fixing the creepiness and fixing the waste are frequently the same fix.

Uncapped retargeting can show the same ad to the same person a dozen times a day across different sites, which produces exactly the stalker-ish feeling that gives retargeting its bad reputation. We cap impression frequency per user per day, typically two to three, and cap total campaign duration for a single visitor, since someone who didn't convert after two weeks of retargeting is unlikely to convert from a third week of the identical ad.

We set these caps explicitly in the platform rather than relying on the algorithm's own default pacing, since default pacing optimizes for delivering the full budget, not for what a reasonable human on the receiving end would consider a reasonable frequency. Left uncapped, the platform's own incentives and the audience's tolerance point in opposite directions.

Showing the exact same static ad repeatedly is what makes retargeting feel oppressive even at a reasonable frequency. Rotating through three or four genuinely different creative angles, different product benefits, a review, a limited-time offer, prevents the sense of being chased by the same image and lets us see which angle actually converts best for a given audience.

We refresh this rotation on a fixed schedule, roughly every two to three weeks for an active campaign, rather than waiting for a specific creative to visibly fatigue before replacing it. By the time fatigue shows up clearly in the performance data, the audience has usually already been seeing the same ad long enough to have formed an opinion about the brand that a fresher creative can't immediately undo.

Someone who added a product to cart and abandoned checkout is a very different audience from someone who read one blog post and left. Retargeting both groups with the same generic "come back" ad wastes budget on the low-intent group and under-serves the high-intent group with a message that isn't specific enough to close the sale.

We build at least three distinct intent tiers for any retargeting campaign with enough volume to support them: high-intent (cart or checkout abandonment), medium-intent (product page views without cart activity), and low-intent (general site browsing), each with its own message, its own frequency cap, and often its own budget allocation, since treating all three the same wastes the specificity that makes retargeting worth doing in the first place.

Continuing to retarget someone who already purchased, with an ad for the exact product they just bought, is a specific and avoidable failure mode we still see on client accounts before we take them over. A same-day exclusion list, updated automatically from purchase events, is a basic setup that a surprising number of accounts still lack.

This mistake is also one of the fastest ways to visibly damage brand trust, since a customer seeing an ad for something they just bought reads it as evidence the company doesn't actually track its own customers well, a much worse impression than simply being retargeted a bit too often would create on its own.

We watch for a specific signal: brand search volume and direct traffic should hold steady or rise during an active retargeting campaign, not dip. A dip suggests the campaign is generating enough irritation that some users are actively avoiding the brand rather than being nudged toward it, which is the clearest sign frequency needs to come down regardless of what the click-through rate says.

We also occasionally ask a small sample of the client's own customers directly whether they've noticed the retargeting and how it felt, a qualitative check that platform metrics alone can't provide, since a campaign can look perfectly healthy on every quantitative measure while still leaving a genuinely bad impression on the specific people it reached most often.

A 30-day retargeting window makes sense for a moderately considered purchase and is far too short for something like enterprise software with a months-long sales cycle, and far too long for a genuinely impulse-driven low-cost item where anyone still uncapped after a week has likely moved on entirely. We set the retargeting window based on the client's actual historical time-to-purchase data, not a platform default, since the default rarely matches the real decision timeline for any specific business.

Getting this window wrong in either direction wastes budget: too short and you stop reaching people who were still genuinely deciding, too long and you keep spending on an audience that's already moved on, with the added cost of that extended reach being exactly the kind of unwanted persistence that makes retargeting feel invasive in the first place.

We revisit these windows any time a client's actual sales cycle shifts, since a business that's moved upmarket toward larger, more considered purchases needs a longer window than the one that made sense for its original, faster-moving customer base, and a stale window is an easy thing to leave unexamined for months once it's initially set.

None of this requires sophisticated tooling to get right, just an honest look at the client's own CRM data for how long a typical customer actually takes from first visit to purchase, which is usually sitting there unused in favor of a platform default nobody's ever questioned.