Retargeting for SaaS

Retargeting is the line item nobody defends and nobody cuts. It reports a good CPA, it costs a few thousand a month, and when you ask what it’s actually adding the answer is a dashboard rather than an argument. The dashboard is usually wrong, and the reason is structural.

Two fixes change the economics. Segment the audience by intent depth so the message matches where someone actually is, and run a holdout so you know what the spend is adding. Most SaaS teams do neither, which is why most SaaS retargeting reporting is fiction.

Why does one all visitors audience waste most of its budget?

Because your site traffic is mostly people who will never buy, and an undifferentiated pool spends on all of them equally. A blog reader who arrived from a search for “what is SOC 2” and a prospect who abandoned a demo form on Tuesday get the same banner.

Segment by how deep the intent went. Six pools cover almost every SaaS site.

The demo abandoner pool is small and worth more than everything above it combined. On a typical mid market site it might be 120 people a month, and a 40 dollar CPM against 120 people costs almost nothing while reaching exactly the audience that was three fields from a meeting.

What recency and frequency windows should you use?

Short and capped. Fourteen to thirty days on recency, three to five impressions per week on frequency, and both tighter than instinct suggests.

The intuition that a 90 day sales cycle needs a 90 day window is wrong for a specific reason: the pool is small. A 90 day window on a site with 6,000 monthly qualifying visitors means 18,000 people in the pool, and your 2,000 dollar budget spread across 18,000 people buys almost nothing per person. The same budget across a 14 day pool of 3,000 people buys real presence for the ones who visited recently, which is where the intent is.

Frequency caps matter more in B2B than in ecommerce because the pools are so much smaller. Without a cap, Google Display will cheerfully serve one person 60 impressions in a fortnight. Nobody converts on the 40th banner. They do develop an opinion about you.

Set caps at the campaign level and check the frequency report weekly. If average frequency is above six, your audience is too small for the budget, and the correct response is to cut the budget rather than widen the window.

Which channel should carry which segment?

Different platforms buy attention of different quality at very different prices, and the segment should decide.

Channel	Typical CPM	Control	Audience floor	Best segment
Google Display Network	$2 to $8	Low, placement quality varies wildly	~100 users	Broad recency and all visitors
YouTube in stream and Demand Gen	$10 to $30	Moderate, good creative control	~1,000 users	Pricing visitors and unactivated trials
LinkedIn retargeting	$25 to $60	High, firmographic layering available	300 members	Demo abandoners and closed lost
Meta	$6 to $18	Moderate, strong creative formats	~1,000 users	Practitioner led products and self serve trials
Programmatic via StackAdapt or similar	$8 to $25	High with managed placement lists	Varies	Account based display alongside ABM

YouTube is the underused one. A considered B2B purchase benefits from 30 seconds of explanation more than from a 300 by 250 banner, and in stream placements on a warm audience are among the better performing formats we see. Make the video specific: a product screen recording with a real workflow beats a brand film at this stage.

Display’s honest role is cheap recency for the broad pool. It’s not where your demo abandoners should live. If display is more than half your retargeting spend, the segmentation is probably not real.

Google Display versus LinkedIn is a recurring argument, and the comparison page covers the full tradeoff. For retargeting specifically: LinkedIn’s cost per impression is five to ten times higher and it is still the right choice for a 200 person demo abandoner pool at enterprise ACV.

What is the view through conversion trap?

It’s the reason your display report looks good. A view through conversion credits an impression that nobody clicked. In retargeting, every person seeing the ad already knows you, which means a large share of those conversions were going to happen with or without the impression.

Three things to do about it.

• Report click through conversions as the primary number, in a separate column from view throughs. Not blended.
• Shorten the view through window to one day if the platform allows it. The default 30 day window in Google Ads will claim almost everything.
• Never let view through conversions feed your bid strategy on a retargeting campaign, because the algorithm will then optimise for impressions served to people already about to convert.

How do you design a holdout that proves lift?

Exclude a random 10 percent of every retargeting audience, hold it for eight weeks minimum, and compare conversion rates between exposed and excluded.

Ten percent is the right size for most SaaS programmes: large enough to detect a real effect on pools of a few thousand, small enough that the opportunity cost is trivial. If your total retargeting audience is under 1,000 people, a holdout will not reach significance and you should instead run the whole channel on and off in alternating months.

The position here is blunt. Any retargeting spend above roughly 2,000 dollars a month should be validated with a holdout before it’s scaled, because the reported numbers are systematically optimistic and the correction is usually large.

What does this cost to run properly?

More setup than people expect. Six audiences with distinct creative means six creative sets, and a video for the YouTube segments. Budget two to three weeks of design and copy work up front, plus a quarterly refresh, because retargeting creative wears out faster than acquisition creative against a small pool.

Ongoing, expect one person spending a few hours a month on frequency reports, audience size checks and creative rotation. Programmes that get set up and left alone drift within a quarter: pools shrink, frequency climbs, and nobody notices because the CPA still looks fine.

Two failure modes beyond measurement. Audience floors quietly stop campaigns serving, particularly on LinkedIn where a 240 member pool simply does nothing and doesn’t shout about it. And exclusion hygiene breaks: existing customers and current opportunities keep seeing acquisition ads because someone forgot to exclude the logged in user audience, which wastes money and irritates the people paying you. That belongs on the same maintenance list as your negative keyword list.

What to do next

Start by splitting the audience. Build the six pools this week, exclude blog traffic from the main one, and write one distinct message per pool using the ad copy swipe file as a starting point. That change alone usually improves performance before you touch a single bid.

Then set the frequency caps, shorten the view through window, and stand up the 10 percent holdout before you increase spend. If the holdout shows real lift, scale into YouTube and LinkedIn for the deep intent segments and read the ABM playbook for the account based layer. If it doesn’t, move the budget into search and let the advertising strategy playbook and bid math decide where. For the cheaper practitioner channels, the Meta ads guide is worth a look, and common PPC mistakes covers what else tends to be leaking. The channel overview lives at SaaS PPC.