Facebook Remarketing Will Not Name the Buyer. Meta’s Own Docs Say So.

Meta will happily show your ad to the 4,000 people who read your pricing page last month. It will not tell you the name of a single one of them.

That is not a privacy setting you forgot to flip. It is the product working as designed. Meta’s own Website Custom Audiences documentation says it in plain language: no personal information is reported to the advertiser about any individual person on a website. You can target the audience. You cannot look inside it.

So sit with this for a second. You identified your highest-intent traffic, handed it to an ad auction, and got back impressions. Where exactly did the sales opportunity go?

Facebook Remarketing Is a Reach Product. Most Teams Use It as a Pipeline Product.

Here is the mismatch nobody names out loud. Facebook remarketing is extraordinarily good at one job: putting a message back in front of someone who already touched you, cheaply. It is structurally incapable of a second job that every B2B sales team desperately needs: telling a human being who to call.

The mechanics make this worse than it sounds. Per the same Meta docs, website custom audiences are built with a retention_days window capped at 180 days on the Marketing API, and membership expires when that window lapses unless the person comes back. Your remarketing audience is a decaying, anonymous, rented list. It forgets. And the person who read your pricing page in March, disappeared, and reappeared in September looks to Meta like a brand new anonymous visitor.

Meanwhile, Gartner’s research on the B2B buying journey found buyers spend only about 17% of their total buying time meeting with potential suppliers at all. The overwhelming majority of the decision happens where you are not in the room. Remarketing on Facebook is one of the few ways to be present during that other 83% — which is exactly why it is so painful that the channel hands the intelligence back to you scrubbed of identity.

This is the same structural problem we wrote about with ads and retargeting strategy: the platform keeps the graph, you keep the invoice.

The Uncomfortable Research: Most of That Retargeting ROAS Was Never Incremental

Before we sell you on anything, here is the finding that cuts hardest against everyone selling retargeting — including us.

Gordon, Moakler and Zettelmeyer analyzed 663 large-scale randomized experiments run at Facebook (published 2022, with access to over 5,000 user-level features). The median true experimental lift for lower-funnel outcomes was 5%. When they tried to recover that number using the non-experimental methods most advertisers and their measurement partners actually use, the estimates came back at 24% to 64% — off by a factor of five or more, in the flattering direction.

That is not an isolated result. The earlier Gordon, Zettelmeyer, Bhargava and Chapsky study in Marketing Science, built on 15 US field experiments spanning 1.6 billion ad impressions, found observational methods frequently mis-estimated true lift by 3x or more even after conditioning on rich demographic and behavioral data.

Translate that into operator language: the dashboard number your agency is showing you for Facebook remarketing is, in expectation, dramatically overstating what the ads actually caused. Retargeting pools are full of people who were going to convert anyway. The platform cheerfully takes credit for them.

Where this gets uncomfortable for us: if you run a proper holdout test and discover your Facebook remarketing is producing near-zero incremental revenue, the correct move is not to buy a scoring system. It is to cut the spend. We would rather you find that out in week two than pay us for eighteen months of plumbing on top of a channel that was never moving the number. The scoring layer we are about to describe is worth building because it makes the data useful regardless of whether the ads are incremental — not because it rescues bad ad economics.

The Hypothesis: What If Remarketing Is an Input, Not an Outcome?

Our thesis is this: the highest-value output of a Facebook remarketing program is not the click. It is the behavioral event that fires on your own property on the way to and from that click — and that event belongs to you, permanently, with a name attached, if you build somewhere for it to land.

Two things make this more buildable in 2026 than people assume.

• The cookie apocalypse did not arrive. Google formally announced it would maintain its current approach to third-party cookie choice in Chrome rather than deprecating them. Planning your entire data strategy around a doomsday that got cancelled is its own kind of expensive.
• Server-side event capture is documented and open. Meta’s Conversions API lets you send events server-side, linked to a dataset ID and processed alongside pixel events. If you are already firing those events to Meta from your server, you can fire them to your own database in the same breath. The hard part was never technical.

The identity gap Meta enforces is a gap in their reporting, not in your building. On your side of the wall, you know who is logged in, who clicked from an email, who filled the form, who came back from an ad with a UTM. Meta refuses to name them. Nothing stops you.

The Architecture We’d Build

This is a proposed workflow — the architecture we would build, described honestly as a build, not as a case study.

ICP / TAM LIST  →  OUTBOUND EMAIL  →  CLICK (carries identity)
                                              ↓
                              SITE VISIT — NOW NAMED, NOT ANONYMOUS
                                              ↓
             PIXEL + CONVERSIONS API fire to Meta AND to your own store
                                              ↓
   META SIDE: build remarketing audience   |   YOUR SIDE: write event to CRM
                                              ↓
                          LEAD SCORE (recency + depth + repetition)
                                              ↓
                        SCORE CROSSES THRESHOLD → HOT LEAD ALERT
                                              ↓
                              HUMAN CALLS A NAMED PERSON

The critical design decision is the identity bridge: outbound email click, form fill, or login is what converts an anonymous session into a named one. That is why TAM mining and outbound email is not a separate program from paid social — it is the thing that makes paid social legible. The Prospect Pump™ feeds identified people into the top; Facebook remarketing keeps them warm in the middle; the Easy Button™ lead scoring system decides who gets a phone call.

Meta’s docs note that website custom audiences update within minutes of the pixel firing. Your CRM should be just as fast. If a scored event takes overnight to reach a rep, you have built a reporting system, not a CRM automation system. Watchtower™ exists for exactly this: watch the signal, not the calendar reminder.

The Economics

Illustrative example — run your own numbers. These are structural, not a client result.

Say 4,000 people hit your site from remarketing and outbound over 90 days. Treated purely as an ad audience, that is a $2,000 media spend producing some number of self-serve conversions you cannot verify without a holdout. Treated as a signal set with identity attached, suppose 12% are identifiable through email clicks and form fills — 480 named people. Suppose 8% of those cross a meaningful score threshold: 38 people worth a call.

A rep dialing a random 4,000-record list at 3% connect-to-meeting is doing archaeology. A rep dialing 38 people who read pricing twice last week is doing sales. Even at a conservative 15% meeting rate, that is 5 or 6 meetings out of the exact same media spend — meetings that the ad platform reported as nothing at all, because the ad platform cannot report a name.

The asymmetry is the point. The ad budget was going to be spent either way. The scoring layer is a one-time build. Every dollar of remarketing spend after that produces two assets instead of one.

The Playbook: Build the First Version Yourself in a Week

You do not need us for version one. Here is the whole thing.

1. Put a UTM-carrying identifier on every outbound email link. Not just utm_source — a contact ID. This is the single highest-leverage line of work in the entire build, and it takes an afternoon.
2. Capture that ID into a cookie or session on landing and write it to your database with the page URL and timestamp. One table: contact_id, event_type, url, timestamp.
3. Stand up the Conversions API alongside your pixel. Same server call, two destinations: Meta for audience building, your own store for scoring. Meta’s docs walk through the dataset ID and event payload.
4. Score three dimensions only. Recency (last 7 days beats last 60), depth (pricing and services pages beat blog posts), repetition (three visits beat one). Resist a 40-variable model. Three works.
5. Set a threshold and an alert. Make.com webhook to Slack or straight into a Pipedrive activity. Name, company, score, top signal, one sentence of why. Nothing else.
6. Build your Facebook remarketing audiences off the scored segments, not off all traffic. Retarget the 87s with a demo offer and the 30s with content and engagement. Same budget, wildly different message-to-intent fit.
7. Run a holdout before you scale. Hold back 20% of the remarketing audience for four weeks. If the treated group does not meaningfully beat the holdout, the research above just predicted your result. Believe it and reallocate.

Step seven is the one most agencies will never suggest to you, which tells you something about who is optimizing for what.

Who Should Not Build This

If your site gets 300 visitors a month, the scoring layer will produce a list of four people and you will be tempted to call all four anyway. Skip it. Go fix top-of-funnel first. If you sell a $40 product with a same-session purchase cycle, the named-lead-to-phone-call model is the wrong shape entirely — your remarketing genuinely is just a conversion channel and you should optimize it as one. And if nobody at your company will actually pick up the phone when the alert fires, no amount of architecture fixes that. The bottleneck is the call, not the code. We have watched that one kill more builds than any technical failure.

Your Prospects Are Already Talking

Facebook remarketing is not the problem. Treating it as the finish line is. The channel is a legitimately efficient way to stay in front of people during the 83% of the buying journey you are not invited to — and it will never, by design, hand you a name.

So build something that listens on your side of the wall. Connect the systems. Track the behavior. Score meaningful engagement. Compress thousands of anonymous events into a short list of named humans worth a phone call. Then put your salespeople where they belong: talking to the people most likely to care.

Want to build it yourself? Steal the framework above — all seven steps. Want us to bolt it onto the stack you are already paying for? See what this costs or work with Jeremy directly for 90 days.