How to Choose a Remarketing Agency in 2026: A Guide to ROI and Incrementality

Choosing a remarketing agency should not bezzgin with the highest return-on-ad-spend figure in a pitch deck. It should begin with a harder question: how much of that return did the advertising actually create?

Remarketing campaigns reach people who already know your business. Some have viewed a product, visited a pricing page, started checkout, or purchased before. That makes these audiences valuable, but it also makes performance easy to overstate. A person who buys after seeing a remarketing ad may have returned and purchased without seeing the ad.

The right agency must therefore do more than build audiences and run ads. It should be able to work with first-party data, maintain reliable measurement across browsers and platforms, control audience saturation, and distinguish attributed conversions from incremental ones.

This guide explains how to evaluate those capabilities and choose a partner whose results stand out beyond the advertising dashboard.

Key takeaways

• Platform-reported ROAS measures attributed revenue, not necessarily revenue caused by advertising.
• Privacy restrictions and fragmented browser behavior make first-party data and durable measurement infrastructure increasingly important.
• Audience segmentation, creative sequencing, exclusion rules, and frequency management are central to efficient remarketing.
• Holdout tests and other experiments provide stronger evidence of incremental impact than last-click or view-through attribution alone.
• Agency benchmarks should be treated as context, not promises. Your margin, sales cycle, audience size, and break-even economics matter more than a universal target.

Remarketing in 2026: a fragmented measurement environment

Remarketing has not disappeared, but the environment around it has changed.

Safari has restricted cross-site tracking through Intelligent Tracking Prevention, while Firefox uses Total Cookie Protection to isolate cookies by site. Apple’s App Tracking Transparency framework also requires apps to obtain permission before tracking users across other companies’ apps and websites.

Chrome, however, has not followed a simple path toward eliminating third-party cookies. Google moved away from its earlier phase-out plan and retained an approach based on user choice. In October 2025, it also announced changes to the Privacy Sandbox roadmap, including the retirement of several technologies. Google’s latest Privacy Sandbox update reflects that more complicated reality.

The result is not one uniformly “cookieless” internet. It is a fragmented system in which measurement and audience availability vary across browsers, devices, platforms, consent statuses, and technical implementations.

That fragmentation increases the value of data your business collects directly, such as CRM records, customer lists, purchase history, on-site events, and lead-stage information. It also makes reliable conversion tracking more important. Tools such as Google Enhanced Conversions, Google Customer Match, and the Meta Conversions API can help advertisers use consented first-party data and improve measurement. They do not eliminate signal loss or automatically prove that advertising caused a conversion.

A capable agency should understand both sides of this equation: how to preserve useful signals and how to avoid overstating what those signals prove.

Attribution is not incrementality

This distinction is the foundation of credible remarketing measurement.

Attribution assigns credit for a conversion. Incrementality asks whether the conversion happened because of the advertising.

Suppose a remarketing campaign reaches 10,000 previous site visitors. The advertising platform reports 200 conversions from that audience. That does not mean the campaign created all 200. Some people may have returned because they remembered the brand, received an email, saw an organic result, or had already intended to purchase.

A holdout test provides a stronger comparison. One eligible group sees the advertising, while a statistically comparable control group does not. If 8% of the exposed group converts and 7.5% of the control group converts, the measured incremental lift is 0.5 percentage points. That difference is more useful for budget decisions than claiming credit for the entire 8% conversion rate.

Not every business has enough audience volume to run a conclusive holdout test at all times. A good agency should be honest about that constraint. Depending on scale and channel, it may use geo experiments, conversion-lift studies, matched-market tests, or carefully defined suppression tests instead. The important question is whether the agency has a credible method for estimating what would have happened without the ads.

What effective remarketing looks like

1. Segmentation based on intent

Treating every website visitor as one audience wastes budget and produces generic messaging. Strong programs segment people using signals such as:

• the pages or products viewed;
• depth and recency of engagement;
• cart or form abandonment;
• customer and lead status;
• purchase history;
• sales-cycle stage; and
• relevant exclusions, including recent converters.

The logic should match the business. An e-commerce shopper who abandoned a specific product requires a different message from a past customer eligible for replenishment. A B2B prospect who visited a pricing page should not necessarily receive the same creative as someone who read one introductory blog post.

2. Creative matched to the audience

Remarketing creative should address why a person has not yet acted. Depending on the segment, that may mean reinforcing value, resolving an objection, presenting social proof, showing a previously viewed product, or introducing a relevant case study.

Dynamic ads can be useful for large product catalogues, but relevance is not the same as persuasion. Showing someone the same product repeatedly will not fix unclear positioning, weak proof, an unsuitable offer, or a poor landing experience.

3. Frequency and saturation controls

Remarketing audiences are finite. As the spend rises, campaigns can reach the same people too often, increasing costs without producing comparable gains.

Platforms provide frequency-management options, including frequency caps in Google Ads. The correct limit depends on the channel, audience size, creative variety, funnel stage, and length of the buying cycle. Agencies should explain how they monitor reach, frequency, creative fatigue, marginal CPA, and audience overlap rather than applying a single cap to every account.

4. Coordination with the wider media strategy

Remarketing cannot grow indefinitely without new demand entering the funnel. It should work alongside prospecting, search, organic acquisition, email, and other customer-journey activities.

An effective agency will show how remarketing audiences are created, which channels replenish them, and when additional spend begins to produce diminishing returns. It should not recommend a fixed percentage of total media budget without first examining audience volume and marginal performance.

A six-part framework for evaluating an agency

Use the following framework to compare agencies consistently. Score each area from 1 to 5, but treat the score as a decision aid rather than a universal pass-or-fail rule. A weakness in measurement or data governance may matter more than a similar weakness elsewhere.

1. Platform expertise

The agency should demonstrate practical experience with the platforms relevant to your business. That might include Google Ads remarketing and Customer Match, Meta Custom Audiences and Conversions API, or specialist platforms such as Criteo, AdRoll, and RTB House.

Certifications can indicate familiarity with a platform, but they are not proof of strategic ability. Ask for account examples, audience logic, test designs, and explanations of when the agency would choose one platform over another.

2. Data, tracking, and privacy capability

Ask how the agency would assess your existing tracking, first-party data, consent management, CRM integrations, and offline conversion flow. It should understand applicable privacy requirements and work with legal or compliance teams without presenting itself as a substitute for legal advice.

Look for experience implementing relevant tools rather than vague claims of being “cookieless-ready.” The agency should also explain the limitations of each implementation and how it validates data quality.

3. Strategic approach

Request a clear methodology for:

• audience creation and exclusions;
• segmentation by intent and recency;
• creative testing and sequencing;
• landing-page alignment;
• frequency management; and
• budget movement between prospecting and remarketing.

A strong answer connects these decisions to your customer journey. A weak one relies on generic audience windows and platform defaults.

4. Attribution and incrementality

Ask the agency to separate platform-attributed results from experimentally measured lift. It should be able to explain which attribution windows it uses, how it treats view-through conversions, and when a holdout or lift test is feasible.

If the agency has run experiments, ask to see the hypothesis, control design, sample size, duration, confidence interval, and business decision that followed. A screenshot showing high ROAS is not an incrementality study.

5. Reporting and communication

Reporting should connect media activity to commercial outcomes. Depending on your business, that may include contribution margin, qualified pipeline, closed revenue, repeat purchases, or customer lifetime value alongside platform metrics.

Ask whether you will receive raw data access, who will manage the account, how often strategy reviews occur, and how the agency communicates tracking failures or performance risks. A real-time dashboard is useful, but access alone does not guarantee interpretation or accountability.

6. Relevant operating experience

Experience in your industry can shorten the learning curve because purchase cycles, margins, regulations, and audience behavior differ substantially. However, a familiar logo or vertical-specific case study should not replace sound test design.

Ask what the agency learned from comparable accounts, what did not work, and which parts of that experience genuinely transfer to your business.

Questions to ask before hiring a remarketing agency

Use these questions during the selection process:

1. How do you distinguish attributed conversions from incremental conversions?
2. Can you show us a holdout, lift, geo, or matched-market test you have designed?
3. How do you decide whether our audience is large enough for a meaningful experiment?
4. Which attribution windows do you use, and how do you report view-through conversions?
5. How will you audit our current tracking and first-party data?
6. Which server-side or enhanced-conversion implementations are relevant to us, and what will they not solve?
7. How do you segment audiences beyond all-site visitors?
8. How do you manage exclusions, audience overlap, frequency, and creative fatigue?
9. How do you decide the appropriate balance between prospecting and remarketing?
10. Which business metrics will appear in our reports?
11. Who will manage the account day-to-day?
12. Can we access and export our raw performance data?
13. Who owns the advertising accounts, pixels, audience definitions, creative files, and historical data if the relationship ends?

Specific answers matter more than polished terminology. The agency should be able to explain its choices in the context of its economics and operating constraints.

How to assess agency case studies

A useful case study should identify:

• the original business problem;
• the audience and channels involved;
• The intervention made;
• the measurement method;
• the period tested;
• the commercial result; and
• Any important limitations?

ROAS, CPA, lead volume, and conversion rate can demonstrate improvement, but they do not independently prove incrementality. If a case study claims that remarketing caused growth, look for an experiment or another credible counterfactual.

Black Propeller’s published work provides examples of broader paid-media outcomes. Its Valet Living case study reports increased conversion volume and qualified leads without additional media spend. Its client results also show improvements in lead and CPA for Santanna Energy Services. These examples can help a prospective client assess campaign-management experience, but they should not be treated as remarketing incrementality studies unless the underlying test design establishes that connection.

That is the standard buyers should apply to every agency, including the one publishing the guide.

Understanding pricing and contract terms

Remarketing agencies commonly charge through a percentage of ad spend, a flat retainer, a project fee, or a performance-linked model.

Each structure creates different incentives. A percentage fee grows with media spend. A retainer provides predictability but requires a defined scope. Performance pricing may appear aligned with results, but its value depends entirely on how the contract defines attribution and success.

Before signing, review:

• minimum spend and contract duration;
• the scope of creative, tracking, and landing-page work;
• termination and notice terms;
• attribution definitions;
• reporting and response commitments;
• ownership of accounts, pixels, audiences, and creative; and
• access to historical and raw data.

The safest arrangement is one in which the advertiser controls its core accounts and can retain its data and campaign history after the engagement ends.

Frequently asked questions

What should I look for in a remarketing agency in 2026?

Look for strong audience strategy, first-party data capability, reliable tracking, thoughtful creative testing, transparent reporting, and a credible approach to incrementality. The agency should explain not only how it will improve performance, but how it will determine whether that improvement was caused by its work.

What is a good ROAS for remarketing?

There is no universal target. A sustainable ROAS depends on gross margin, fulfilment costs, customer lifetime value, return rates, sales-cycle length, and the share of conversions that are genuinely incremental. Calculate your break-even economics first, then evaluate performance against them.

Has remarketing become cookieless?

No. Browser and platform restrictions differ, and Chrome continues to support user choice around third-party cookies. The more accurate description is a fragmented, privacy-constrained environment. First-party data and durable measurement can reduce gaps, but they do not restore every missing signal.

How much budget should go to remarketing?

The appropriate amount depends on the size and quality of your eligible audiences and the marginal return available as spend increases. Begin with audience volume, buying-cycle length, frequency, and incremental economics rather than a fixed industry percentage.

Which agency pricing model is best?

No model is inherently best. Choose the structure whose incentives, scope, and attribution rules are clearest. A seemingly performance-based agreement can still be poorly aligned if the agency receives credit for conversions that would have occurred anyway.

Choose an agency that can defend the result

Remarketing can be efficient because it reaches people who have already demonstrated interest. That same advantage makes its results particularly vulnerable to inflated attribution.

The right agency should be able to activate first-party data, build meaningful segments, manage creative and frequency, and connect media activity to real business outcomes. More importantly, it should be willing to test its own contribution.

Black Propeller approaches paid search, paid social, performance creative, and measurement as connected parts of the media system. Prospective clients can explore its paid social advertising services and display advertising capabilities, or request an account assessment.

Bring your current tracking setup, audience structure, attribution model, and unit economics to the conversation. The objective should not be to produce the most impressive dashboard. It should be to identify which activity is creating additional value and where the next unit of budget can be allocated to work the hardest.