Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Performance Marketing › Channel Playbooks

Prospecting vs Remarketing

Prospecting and remarketing compared: what each does for growth, reported efficiency against incremental value, frequency caps, and splitting the budget.
Comparisonintermediate8 min readUpdated

Two jobs, one budget

Every paid social and display budget does two jobs at once. One is finding new people: showing the business to those who have never visited, chosen by interest, behaviour, similarity to existing customers or the platform's own prediction. The other is following up: showing ads to people who have visited, engaged, added to basket or bought before, built from site and app data and customer lists. The platforms describe the second as Remarketing or retargeting, and their help centres document how such audiences are built from tags, app events and uploaded lists.

The two jobs compete for the same money and are usually measured on the same dashboard, which is where the trouble starts. Remarketing audiences are small, warm and cheap to convert; prospecting audiences are large, cold and expensive to convert. On cost per acquisition (CPA) and return on ad spend (ROAS) as the platforms report them, remarketing wins every comparison, and a budget managed on those figures moves steadily towards it. The remarketing pool, though, is filled by prospecting and by every other source of visitors. Starve prospecting and the pool shrinks, frequency on the remaining people climbs, and the reported efficiency that drove the shift begins to fall. The marketing funnel framework names the stages; this page is about not confusing the bottom of the funnel with the whole of it.

Compare table

Prospecting campaigns vs Remarketing campaigns

Row labelProspecting campaignsRemarketing campaigns
Who they reachPeople who have not interacted with the businessPeople who have visited, engaged, added to basket or bought
Audience sizeLarge and renewableSmall and limited by traffic from all sources
Typical audience sourcesInterests, behaviours, Lookalike Audience lists, broad audiences with platform optimisationSite and app visitors, engagers, basket abandoners, customer lists
Reported CPA and ROASHigher cost, lower returnLower cost, higher return
Incremental share of reported resultsUsually high; most of these people would not have arrived otherwiseOften much lower; many would have bought anyway
Role in growthFills the funnel and the remarketing poolConverts and recovers interest already created
Main riskUnder-funding because it looks expensiveOver-funding because it looks efficient
Frequency pressureLow; spend spread across many peopleHigh; spend concentrated on few people
Creative needsIntroduce the brand, earn attention from cold audiencesRemind, reassure, answer objections, offer a reason to finish
How to judge itCost per new customer in your own data; incrementality testsHoldout tests against reported results, under a cap

The pattern in the table is consistent across platforms. Google and Meta build and deliver these audiences differently, and the audience setup belongs to the platform roots: the Meta custom audiences guide and the Google Display targeting options cover the mechanics. The economics of the split do not change with the platform. Keeping Google Ads prospecting on people who have not bought is set out in the guide to excluding existing customers, campaign type by campaign type.

Where prospecting wins

Prospecting is the only part of a performance budget that brings genuinely new people to the business. It wins whenever the question is growth: more customers next quarter than this one, a new market, a new product line, a category where people do not yet search for what the business sells.

It also wins on incrementality. Because the people it reaches have not shown interest, a larger share of the conversions it drives would not have happened without it. A prospecting campaign with a reported cost per acquisition well above the remarketing figure can easily be cheaper per incremental customer, and that is the comparison that matters to the business.

Prospecting has become easier to run as platforms have automated audience selection. Broad audiences with the platform's delivery system choosing who sees the ad, and audiences seeded from the best existing customers, both find new people efficiently when the conversion event and creative are right. The weakness is patience. Prospecting takes time to produce results, those results are spread across a longer path to purchase, and much of its effect shows up later in other channels, such as branded search or direct visits. Judged by last-click figures in the week it ran, prospecting nearly always looks worse than it is.

Where remarketing wins

Remarketing earns its place where people genuinely need a second prompt. Considered purchases with long decision cycles, high prices or several people involved benefit from staying visible while the decision is made. Abandoned baskets and incomplete sign-ups are real interruptions that a timely reminder can recover. Subscription and repeat businesses use customer lists to bring lapsed customers back or to sell the next product, which is a retention job rather than an acquisition one. In search advertising, remarketing lists usually work as a layer on top of keywords rather than as the targeting itself, and the guide to audience segments in Google search campaigns sets out when a list earns a campaign of its own.

Remarketing also wins on creative opportunity. Because the audience already knows the business, ads can answer specific objections, show the exact product viewed, explain delivery and returns, or present a relevant offer. That is a different and often more effective message than anything a cold audience would understand. The same precision becomes an irritation when the ads keep following people who have already bought, and the Meta diagnosis for retargeting ads reaching recent buyers traces which exclusion, event, window or product set let them back in.

What remarketing does not do is grow the business on its own. Its audience is bounded by the traffic everything else generates, and its results concentrate on people who were already close to buying. Its value is real but usually smaller than reported, and it shrinks as spend rises because each extra pound reaches the same people more often.

Reported efficiency vs incremental value

The gap between reported and incremental results is the centre of this comparison. A platform reports a conversion when someone who saw or clicked an ad later converts within the attribution window. It cannot know whether the ad caused the conversion. For a cold prospecting audience, most reported conversions are plausibly caused by the ad. For a remarketing audience of recent basket abandoners, a large share would have returned and bought anyway, and the ad merely happened to be the last thing they saw.

Incrementality testing measures the difference. A holdout test withholds ads from a random part of the audience, or from whole regions, and compares conversions between the exposed and unexposed groups. The difference is what the campaign actually added. Some platforms offer built-in lift studies, which their help centres document; geographic holdouts work across all platforms. The design of these tests belongs to the analytics root, and the incrementality testing guide covers it in full.

The usual finding is uncomfortable. Remarketing's incremental return is often a fraction of its reported return, while prospecting's is much closer to its own. The budget implication is that the two should never be compared on reported figures alone.

Illustrative example — not a real result

A homeware brand tests what its remarketing really adds

Linden & Loom, a homeware retailer, spent 45% of its Meta budget on remarketing, which reported a ROAS of 9.2 against 2.4 for prospecting. A four-week holdout withheld remarketing ads from a random 30% of the site-visitor audience. The exposed group converted at 3.4% and the holdout at 2.7%, so only about a fifth of the conversions remarketing claimed were incremental, giving an incremental ROAS of roughly 1.9. A parallel geographic test on prospecting found an incremental ROAS of 2.1. The team capped remarketing at 15% of spend, moved the difference to prospecting, and within a quarter the remarketing audience had grown by a third, because more new visitors were arriving to be remarketed to.

Splitting budget and setting caps

There is no correct universal split, and any rule that states one as a share of budget is guessing. The split should be set from three things.

Audience size. Remarketing spend should scale with the size of the remarketing pool, not with its reported return. When spend grows faster than the pool, frequency rises and incremental value falls. A reach and frequency calculator shows how quickly a small audience saturates.

Frequency. Cap how often each person in the remarketing audience sees ads in a period, using the frequency controls the platforms document, and watch for fatigue. The frequency and creative fatigue guide covers the signs on Meta, and the same logic applies to display.

Test evidence. Once a holdout test has measured remarketing's incremental return, set its budget at the level where that incremental return still meets the target, and treat everything above as better spent on prospecting. Retest whenever the audience, offer or creative changes materially.

The trigger to move budget is simple: when the remarketing audience stops growing, or its frequency keeps rising, shift spend to prospecting. The cross-channel performance planning guide shows how this layer fits into the plan as a whole, and the channel playbooks hub sets out the roles around it.

Before moving budget between the two jobs, the retargeting campaigns guide to the ROAS calculator shows how to put remarketing on an incremental footing and read it against break-even.

When sizing the remarketing side, the reach and frequency guide for retargeting campaigns shows how a small list saturates and where the budget ceiling sits.

For a shop selling online, the paid social channel page for online sales applies this split with buyer exclusions, a retargeting cap sized to the visitor pool and holdout-adjusted returns.

If remarketing and branded search appear to drive most sales, the diagnosis of last-touch tactics taking the credit covers the model, audience, untagged links and testing causes.

Common mistakes

  • Comparing the two on reported ROAS. Remarketing will always win that comparison, and the budget will drift towards the layer that adds least per pound.
  • Letting automation blend them unseen. Automated campaigns favour likely converters; check how much spend reaches existing visitors and customers.
  • No frequency cap on small audiences. A small remarketing pool with an uncapped budget becomes an annoyance and stops adding sales.
  • Never running a holdout. Without one, nobody knows whether remarketing adds sales or merely claims them.
  • Cutting prospecting when results dip. The remarketing pool and later search volume depend on it; cuts show up weeks afterwards, elsewhere.
  • Same creative for both jobs. Cold audiences need an introduction; warm ones need a reason to finish.
  • Excluding nothing. Prospecting that keeps reaching recent purchasers wastes money and inflates its own results; exclude converters where the platform allows.

Frequently asked questions

Why does remarketing always show the best ROAS?

Because it reaches people who have already visited, added to basket or engaged, and a large share of them were going to buy regardless. The platform credits the purchase to the last ad they saw, so the reported return includes sales the ad did not cause. Only a holdout test, where part of the audience is deliberately not shown ads, reveals how much of that return is incremental.

Should remarketing be switched off entirely?

Rarely. In most tests remarketing adds some sales, particularly for considered purchases, abandoned baskets and long decision cycles. The problem is usually scale rather than existence: too much budget, too much frequency and too much credit. A capped layer judged on incremental results is the usual answer, with the cap set from test evidence.

How do automated campaigns that mix both affect this?

Platforms increasingly blend prospecting and remarketing inside one automated campaign, and their help centres document settings that control or report how much reaches existing customers and site visitors. Left unchecked, automation favours the audiences most likely to convert, which drifts spend towards remarketing. Use the available audience controls and reports to see the split, and test the blended campaign's incrementality as you would a remarketing one.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Part of these learning paths

Sources

  1. Google Ads Help — Google Ads Help
  2. Meta Business Help Centre — Meta

Related