B2B retargeting strategy: Audiences, frequency caps and the holdout test

A B2B retargeting strategy shows paid ads to people who already visited your site or sit in your customer relationship management (CRM) system, split into intent segments, with exclusions, a manual frequency cap and a holdout test. Google needs at least 100 active users in the last 30 days before Display ads serve. In our illustrative holdout, only 36 of 180 reported conversions were incremental, one fifth.

What a B2B retargeting strategy is and the figures that drive it

B2B retargeting means showing paid ads to people who already touched your company: visitors to your website, or contacts you upload from your CRM system. ZoomInfo describes it as using tracking pixels, first-party data, or IP-based identification to re-engage visitors who left without converting. Google Ads has folded the term into its own wording: “remarketing” is now referred to as “your data” in its audience reports.

The B2B version runs on a longer clock. ZoomInfo notes that buying cycles are longer, often spanning months rather than days. A plan has to keep a small group of accounts warm for a long time without overspending on them.

The four inputs

Every retargeting plan comes down to four numbers. Get them on one page before you set a budget.

Audience size has a floor. On Google, the Display Network must have a minimum of 100 active visitors or users within the last 30 days for ads to serve. Older uploads follow a stricter rule: any list uploaded before February 1, 2024 must still meet the previous requirement of minimum of 1,000 matched users. You also specify a membership duration, which controls how long a visitor stays in the pool. A new list can take 48 to 72 hours for it to fully populate, so build segments before launch day.

A worked example

These figures are illustrative, not benchmarks. Plug in your own.

• Uploaded contacts: 5,000
• Frequency cap: 3 impressions per person per week, over 4 weeks
• Impressions: 2,000 × 3 × 4 = 24,000
• Assumed CPM: $20
• Spend: 24,000 ÷ 1,000 × $20 = $480

The formula is matched audience × frequency × weeks × CPM ÷ 1,000. If the spend you want is higher than this result, the pool is too small to absorb it. Adding budget then raises frequency on the same people, not reach. For account-level programs, size the pool against your account-based marketing list first. Then plan a holdout so you can measure incrementality rather than counting conversions the ads did not cause.

Which audiences should B2B retarget

Start with four segments. Three are built from site visits with the Google tag. The fourth comes from your CRM system as an uploaded list. Each segment needs a rule, a membership duration and one offer that matches what the visitor was doing.

The rule format follows Google's own example. An airline built a "Flight deals" segment with a membership duration of 30 days and a rule to collect all visitors to a webpage that has a URL that ends with "flight-deals.html." Swap in your own URL patterns.

Worked example for durations

Say your median time from first visit to opportunity is 90 days. Product and content segments get 90 days so a prospect stays reachable through the cycle. Pricing visitors get 30 days, because that visit signals a short window. Check your own cycle against sales cycle length benchmarks before you copy these.

Return visits extend membership. Every time someone visits a page on your site containing an event snippet, the visitor's timestamp is updated, so an engaged reader stays in the segment.

Why the exclusions matter

If people visit multiple pages on your website that contain event snippets, they can be added to multiple customer segments. Ads from all those campaigns become eligible, and the ad with the highest ad rank will then compete for placement in the auction. Without exclusions, a pricing visitor can see a blog ad instead of the demo offer. The exclusion rules in the table keep each person in one segment.

Customer lists

Customer Match lets you use customer information to target ads to your customers. You upload a CSV or connect a data source. If you normalize the file yourself, format phone numbers according to the E.164 standard. Lists Google builds for you behave differently: data segments that are automatically created, like conversion-based customer lists, can’t be edited.

Housekeeping

Leave segments open while you test. If you set a segment to "Closed", and the membership duration is too short, the segment may become too small to use for targeting. Unused segments also expire: if you don't use the segment in your campaign or ad group targeting for 540 days, Google Ads will automatically close it. For account lists built from buying signals, see intent data.

Who to exclude from retargeting

In our reading, exclusions are the cheapest change in a retargeting program because they remove spend instead of adding it. Google frames the purpose the same way: dropping segments that do not fit can "help reduce your costs and increase the likelihood that your ad will attract clicks and conversions", according to Google Ads Help.

How exclusions work in Google Ads

Four rules from Google Ads Help shape the setup:

• Keyword and audience exclusions apply only to first-party data.
• You can exclude segments from Search, Display, Demand Gen, Standard Shopping, Video and Performance Max campaigns.
• Audience exclusions are not available during campaign creation. You add them to an existing campaign.
• Exclusions may not apply to people who have opted out of ads personalization.

The second-to-last rule changes the launch checklist. Exclusions cannot go in at setup, so build the campaign, save it, then return and add them. Google recommends watching performance with your audience segments before you add exclusions. In our reading, that advice fits optimization exclusions. It does not fit the core B2B list, which is about who should never be in the pool at all. We add those exclusions as soon as the campaign exists.

The opt-out rule matters too. Some customers will still see your ads, so keep retargeting creative safe for a customer to read.

The standard B2B exclusion list

Converters and open opportunities overlap. Keep both. Form fills do not catch deals that sales sourced, and CRM exports lag behind the website.

A worked example

Refresh the CRM-based lists on the same schedule as your marketing and sales SLA handoffs, so a deal that opens this week leaves the retargeting pool this week.

What frequency cap to use for retargeting

Google Ads caps frequency differently on Display and Video, so the same number means different things in each. Set the cap with the rules of the campaign type in front of you.

The sources for each row: Display caps can apply to the campaign, ad group, or ad, and Google states that only impressions that were viewable count on Display. A Display cap can only be set after you create the campaign, from the Frequency management section. On Display, third-party cookies are the default, and first-party cookies are used to approximate impressions when those are not available.

Video caps can only be set on the campaign level and apply to signed in users, or to devices. Google has a separate flow to set a frequency cap in a new Video campaign. The cap value must be a whole number. When a user reaches it, the stop only applies to in-stream and bumper ads in the auction.

The Display step people miss

Because a Display cap cannot be set at creation, a new retargeting campaign runs uncapped until someone goes back into its settings. Put that step in your launch checklist, next to the exclusions covered above. Google also offers to let Google Ads optimize how often your ads show, which it marks as recommended. We prefer a manual cap on retargeting, because a fixed exposure level is easier to test against a holdout.

A starting cap and a rule for changing it

Our rule for changing the cap:

1. Hold the starting cap for at least two full weeks so each user sees the full weekly allowance more than once.
2. Raise it by one only if the frequency report shows most users at the cap and the exposed group converts ahead of the holdout.
3. Lower it by one if users sit at the cap and the gap to the holdout does not move.

The cap is a lever you tune against incrementality, not against click-through rate.

How big the audience needs to be and what it costs per month

Retargeting budget is arithmetic you can check line by line. The formula is: monthly cost = people reached × frequency a month × CPM ÷ 1,000, where CPM is cost per thousand impressions. Every input is yours to swap, and the cost moves in direct proportion to each one.

Monthly cost for 12,000 reachable people rises from $1,080 at 6 impressions and a $15 CPM to $7,200 at 15 impressions and a $40 CPM.

Worked example: monthly spend to retarget 20,000 visitors (12,000 reachable) at a $25 CPM

Item	Value
6 impressions a person	1800 US dollars a month
10 impressions a person	3000 US dollars a month
15 impressions a person	4500 US dollars a month

Source: Take the Market calculation: people reached x frequency x CPM / 1,000 (illustrative inputs)

Start from reachable people, not visitors

Your analytics visitor count is the wrong starting number. Google counts "active users" differently: that figure is different from total users reported in platforms like Google Analytics or Firebase, because Google Ads focuses on users who have interacted with Google services and ads recently and meet specific criteria. To size the list before you spend, Google suggests you check the 'Advertising' segment in GA4 (Google Analytics 4) for a closer number.

In our reading, customer lists shrink the same way. Only the records Google can match to signed-in users become reachable, so an uploaded list of contacts is a ceiling, not an audience.

There is also a floor. The Google Display Network must have a minimum of 100 active visitors or users within the last 30 days before ads serve. For customer lists, that 100-user threshold is valid only for lists uploaded or refreshed after February 1, 2024. A narrow segment, such as pricing-page visitors from target accounts, can fall below it once matching is applied.

The worked example

12,000 × 10 × $25 ÷ 1,000 = $3,000 a month.

The table runs the same 12,000 people across three frequencies and three CPMs.

The bottom-right cell is almost seven times the top-left. Frequency and CPM multiply, so a cap set loosely on a costly placement adds up quickly.

Template for your own numbers

1. Pull the advertising-eligible user count for the segment, not raw sessions.
2. Multiply by your own match or reachable share. Use the platform's reported segment size once the list has populated.
3. Check that the result clears the 100-user minimum for each network you plan to use.
4. Multiply by the frequency cap you chose in the previous section.
5. Multiply by your observed CPM and divide by 1,000.

Then set the total against pipeline. In our reading, retargeting spend belongs in the same allocation as your other channels, so run it through the demand gen budget allocator and judge it on CAC payback, not on impressions bought.

Why retargeting reports claim conversions that would have happened anyway

Retargeting reports look strong for a structural reason. The ads go to people who already visited, already read a pricing page, already know the brand. Many of them were coming back anyway. The platform has no way to tell a return visit the ad caused from one it only preceded, so in our reading the reported number is an upper bound on what the ads did, not an estimate of it.

How Google assigns the credit

Two rules in Google Ads shape what a Display retargeting report shows.

First, for Display Network and video ads, "the last impression leading to conversion gets credit for the view-through conversion." A visitor who saw an ad and later converted counts toward that ad, even if they would have converted without it.

Second, these view-through conversions are usually reported separately. Outside Performance Max for store goals, App and Demand Gen campaigns, they are "for reporting only" and sit in the "View-through conversions" and "All conversions" columns, not the main "Conversions" column. A team that reads "All conversions" for Display is reading a larger number than one that reads "Conversions."

The window setting matters too. Google says a longer view-through conversion window usually increases the view-through conversions it records. In our reading, a 30-day window on a warm audience with a B2B sales cycle will credit plenty of buyers who were already mid-evaluation.

What the report counts versus what the ads caused

A worked example

The figures below are illustrative, not benchmarks.

1. A retargeting audience of 10,000 people is split. 9,000 can see ads, 1,000 are held out.
2. The exposed group converts at 2.0%: 180 conversions. The platform reports all 180.
3. The holdout converts at 1.6% with no ads at all.
4. Applied to the exposed group, that baseline is 9,000 × 1.6% = 144 conversions that would have happened anyway.
5. Incremental conversions: 180 − 144 = 36. That is one fifth of what the report shows.

If the campaign cost $7,200, the reported cost per conversion is $40. The cost per incremental conversion is $200. In our reading, budget decisions made on the first number will overfund retargeting.

Platforms offer this comparison themselves. Meta's conversion lift testing, as Brainlabs describes it, splits the target audience into a group that sees ads and a group that does not, then measures the difference in behavior. That difference is the incrementality figure. The difference between the two is the same one behind marketing-sourced versus influenced reporting: preceding a deal is not the same as causing it.

How to measure whether retargeting works with a holdout test

The only clean way to see what retargeting adds is to withhold it from part of the audience. Meta's conversion lift studies do this with a test group that sees ads and a control group that does not, as covered earlier. The same logic works on any platform where you can exclude a list. For the wider method, see our page on incrementality testing for paid media.

The holdout template

Copy this into your test plan before launch and fill in every row.

Do not use platform-reported conversions as the primary metric. Google Ads, for example, automatically excludes view-through conversions from people who also interacted with other ads, so its columns follow the platform's own attribution rules, not your holdout split.

The lift formula, with a worked example

Incremental lift = (test conversion rate - control conversion rate) / control conversion rate

Say the test group holds 8,000 people and produces 120 conversions, a rate of 1.5%.

• Lift: (1.5% - 1.0%) / 1.0% = 50%
• Expected conversions in the test group without ads: 1.0% × 8,000 = 80
• Incremental conversions: 120 - 80 = 40

Divide spend by 40, not by 120, to get cost per incremental conversion. That figure belongs in your demand gen dashboard next to the platform number.

Check for knock-on effects on search

Brainlabs ran 17 incrementality studies with Meta and also tracked traffic landing from search engines, paid and organic. Across the studies, the exposed group had 19% more search visits than the population excluded from seeing Meta Ads. Of those incremental visits, 71% came through organic search and 29% through paid search ads. In the studies that split brand from non-brand, 31% of the incremental search traffic came from branded terms.

Those visits were not being represented in any reporting as being driven by paid social. Those studies covered paid social, not retargeting specifically. Still, in our reading, a retargeting holdout that only counts direct conversions can understate the effect if exposed buyers come back through search. Add the secondary metric to the template, compare search visits per person across the two groups, and report it as a separate line. Use the same formula for search lift that you use for conversions.

The retargeting playbook step by step, with owners and a timeline

This is our sequence for taking retargeting from a tag check to a holdout readout in roughly 12 weeks. The order follows the platform. Display caps go on after the campaign exists, and audience exclusions do too: Google says exclusions aren't available during campaign creation, only on an existing campaign.

One note for whoever builds segments. In Google Ads, "remarketing" is now referred to as "your data", so the lists sit under that label.

Two checks people miss

Exclusions have a data limit. Google states that keyword and audience exclusions are only applicable for first-party data, so the customer and open-opportunity lists have to come from your own CRM system or site tags, not a bought list.

Video caps can be shared. If the videos in a campaign are used in other Video campaigns, those other impressions and views count toward that campaign's cap. If one video runs in a prospecting campaign and a retargeting campaign, we think the retargeting cap will be reached sooner than the retargeting report suggests. Use separate creative or plan the cap for both.

Metrics to track, by phase

• Weeks 1 to 3: list size per segment, match rate on exclusion lists.
• Weeks 4 to 6: reach, average frequency per user, viewable impressions, spend per segment.
• Weeks 6 to 12: pipeline created, opportunity rate and pipeline velocity in test versus control. Platform-reported conversions stay in the log but do not decide the readout.

The week 12 decision is the only one that changes budget. Scale, hold or cut retargeting on the incrementality gap between groups, not on attributed conversions.

What this page does not cover

This page covers audience setup, exclusions, frequency caps, budget arithmetic and holdout measurement, mainly in Google Ads with reference to Meta lift testing. It does not cover ad creative, other ad networks in detail, or benchmark CPMs and conversion rates. It is not legal advice on tracking, consent or data use rules, and it is not financial advice.

Related on Take the Market: Google Ads for B2B SaaS: campaign structure, match types, negatives and offline conversion import; LinkedIn ads benchmarks: CPC, CPL and CTR by objective, from published data; Self-reported attribution: how to design the field, code the answers and reconcile them with software attribution.