How to fix: Broken audience flow between funnel stages

TL;DR

When a Google Ads account treats every visitor as one undifferentiated remarketing pool — same bid, same creative, same campaign — Smart Bidding cannot tell a cold prospect from a demo-requester, so CPA averages out and the highest-intent users get under-bid into oblivion. The fix is a progressive funnel with three audience layers (TOFU, MOFU, BOFU), each in its own campaign or ad group, each with its own bid strategy and creative. Accounts that move from single-pool remarketing to a layered structure typically see bottom-funnel conversion rate lift two to four times.

Why it matters

A single broad audience pool collapses three very different intent states into one bid. The cart-abandoner who just left checkout and the cold visitor who read one blog post both get the same Smart Bidding signal — "anyone in this list" — so the algorithm prices them at the average. The average is wrong for both ends. The cart-abandoner is worth several times the average bid; the cold reader is worth a fraction of it. Smart Bidding cannot fix this, because the signal does not encode the gap [1].

Layered audiences fix the signal. A top-of-funnel layer (similar-to-customers, in-market segments, custom segments built from competitor URLs) feeds prospecting demand. A mid-funnel layer (website visitors 30 to 90 days, video viewers past 50%, blog and feature-page readers) catches users in the consideration window. A bottom-funnel layer (cart abandoners, demo-request starters, Customer Match of free-trial signups) closes the loop on the highest-intent users [2]. Each layer carries its own CPA, its own bid strategy, and its own creative — the algorithm learns three patterns instead of one blurred average.

The financial impact compounds. With a single pool, every dollar bid on prospecting also reaches retargets, and every dollar bid on retargets gets diluted across cold traffic. With a layered structure, BOFU spend goes to BOFU eyes only, and TOFU spend buys the cheaper cold inventory the model needs to keep filling the pipeline. Whitead's audit baseline shows the conversion-rate lift on the bottom layer typically lands between 2x and 4x within 30 days — large enough that the rule carries severity: medium even though delivery never halts.

How to fix

1. Map the funnel into three layers. Decide what counts as TOFU, MOFU, and BOFU for your business. A SaaS example: TOFU = visitors who landed on a blog post or competitor-comparison page; MOFU = visitors who hit a pricing or feature page in the last 30 days or watched 50% of a product video; BOFU = users who started a demo request, abandoned a checkout, or signed up for a free trial (Customer Match list). An e-commerce example: TOFU = similar-to-customers and in-market for category; MOFU = product-page viewers 30 days; BOFU = cart abandoners and past purchasers (for repeat-buy campaigns).

2. Build each segment in Audience Manager. Open Tools and Settings, then Shared Library, then Audience Manager. Create:

   • Customer Match lists for first-party data (free-trial users, past purchasers, demo-requesters) — upload hashed email or phone via the Customer Match interface or via the Google Ads API.
   • Custom Segments for keyword and URL-based intent (competitor URLs, in-market keywords). This replaces the deprecated Similar Audiences mechanism.
   • Website Visitor lists for behavioural segmentation by URL path (pricing, features, blog).

3. Split campaigns or ad groups by layer. Do not mix TOFU and BOFU under one campaign. Budget pacing, bid strategy, and creative all differ across layers — co-mingling them forces compromises that defeat the segmentation. The simplest structure is one campaign per layer; tighter structures use one campaign with one ad group per layer if budget is small.

4. Apply different bid strategies per layer. Smart Bidding signals differ by layer:

   • BOFU: Maximize Conversions or tCPA with a tight target. Intent is highest, so the model can afford an aggressive bid.
   • MOFU: tCPA with a looser target. Volume is higher than BOFU but intent is lower.
   • TOFU: Maximize Conversion Value if value tracking is reliable, otherwise Manual CPC or Maximize Conversions on a capped budget — until conversion volume on the layer clears the 30 conversions in 30 days learning gate, Smart Bidding cannot stabilise on tCPA.

5. Match creative to intent. TOFU runs problem-aware messaging and soft offers ("Free buying guide", "Calculator", "Common mistakes"). MOFU runs comparison and proof creative ("Compare to X", "Customer story", "Case study"). BOFU runs hard CTA and urgency ("Finish your demo", "Complete your order", "20% off ends Friday"). Reusing the same creative across all three layers wastes the segmentation.

6. Add cross-stage exclusions. When a user crosses from TOFU into MOFU, they should stop seeing the prospecting creative. Apply audience exclusions at campaign level: TOFU campaign excludes MOFU and BOFU lists; MOFU campaign excludes BOFU lists. This prevents a single user from absorbing budget from all three layers simultaneously.

7. Re-audit after 14 to 30 days. Open the Audiences report and compare CPA, CTR, and conversion rate across the three campaigns. If they look identical, the layers are not actually separated — usually the audience definitions overlap too much or exclusions are missing. Bottom-funnel conversion rate should be visibly higher than TOFU within 30 days.

Common mistakes

• Treating "remarketing" as one switch. Operators often add a single "All Website Visitors" list and call it remarketing. That list is the start, not the end — without splitting by behaviour and intent, the bidding signal stays flat.
• Building layers but using the same bid strategy across all of them. If TOFU and BOFU both run on tCPA with the same target, the model converges on the average and the segmentation contributes nothing.
• Skipping Customer Match because the list is small. Customer Match lists with as few as 1,000 members still inform Smart Bidding signals and seed lookalike-style expansion. The 540-day membership cap [2] is the real constraint, not list size.
• Forgetting cross-stage exclusions. Without exclusions, the same user appears in TOFU, MOFU, and BOFU simultaneously, and each campaign bids on them in parallel — wasting spend and confusing reporting.
• Locking TOFU on tCPA before the layer clears 30 conversions in 30 days. Smart Bidding starves on low volume; below the learning gate it does worse than Manual CPC. See Smart Bidding minimum conversions for the threshold mechanics.

FAQ

Do I need three full campaigns, or can I use ad groups within one campaign? Three campaigns is cleaner — budgets, bid strategies, and reporting separate naturally. Three ad groups under one campaign works if the account is small (< $5k/month) and you accept that the campaign-level budget will get pulled toward whichever layer converts most, which defeats the segmentation logic for TOFU.

What replaced Similar Audiences? Google sunset Similar Audiences in mid-2023. The replacement for TOFU prospecting is Custom Segments seeded with competitor URLs and in-market keywords, plus first-party Customer Match as a strong audience signal for Performance Max and Search. See Customer Match for list-freshness mechanics.

How small can my BOFU list be before it stops working? A few hundred members is enough to drive Smart Bidding signal — the list does not need to be statistically large because Smart Bidding extrapolates from the pattern. Below 100 members the signal is noisy; refresh the list with new conversions weekly to keep it useful.

Does this apply to Performance Max? PMax uses audience signals, not audience targeting — see PMax audience signals weak. The principle still applies (feed PMax your highest-value Customer Match list as a signal) but the structural fix is different from Search remarketing.

How does this interact with budget allocation by funnel stage? Directly — the layered audience structure is the precondition for funnel-aware budget allocation. Without separated layers there is no stage to allocate to. See Budget allocation per stage for the budget-side mechanics.

Sources

1. Google Ads Help — About audience targeting. Audience segment types (Affinity, Custom, In-market, Your data, Detailed demographics, Life events) and which campaign types support each.
2. Google Ads Help — About Customer Match. Customer Match list mechanics, 540-day membership duration cap, list-refresh requirements, EU consent rules.