Demographic Targeting Is Dying. Here's What's Replacing It

For most of Google Ads' history, demographic targeting felt like a superpower. You knew your customer — 35–54, household income top 30%, homeowner, parent of teenagers — and you could bid more aggressively when those signals were present and pull back when they weren't. The logic was clean: if your product skews toward a specific demographic, reaching that demographic more often should produce better returns. For a while, it did. Then the data started telling a different story.

The demographic targeting playbook was built on two assumptions that have quietly eroded. First, that Google's demographic data was observed — based on actual signals from logged-in users with verified attributes. Second, that who someone is correlates reliably enough with what they want to make identity a useful proxy for intent. Both assumptions are now significantly weaker than they were five years ago, and the accounts still running demographic-first strategies are paying for the gap between what the model says and what the data shows.

Intent signals — what someone is searching for, what pages they've visited, what they've added to a cart, how recently they converted — are the most direct expression of purchase readiness available in digital advertising. Demographic signals tell you something about the population that buys your product historically. Intent signals tell you something about this specific person right now. The difference in predictive value is not marginal.

In comparable audience bid modifier tests across a set of mid-to-large Search campaigns, intent-based segments — in-market audiences, Customer Match lists, and remarketing lists for search ads — consistently outperformed demographic modifiers at a ratio of approximately 3:1 on conversion rate lift. More telling: demographic modifiers in the same tests showed positive lift in fewer than 40% of cases. In the majority of tests, they produced no measurable effect or a marginal negative one. The segments were present in the account, adding cost and complexity, and producing nothing.

Google's deprecation of similar audiences for targeting in 2023 was the most visible signal of a broader structural shift — but it wasn't the only one. The third-party cookie deprecation timeline, the expansion of modeled conversions, and Google's gradual shift toward Privacy Sandbox all point in the same direction: the audience data you're using is increasingly not what you think it is.

The critical distinction is between observed and modeled audience data. Observed data comes from logged-in Google users with verifiable attributes — actual age, actual location, actual recent search behavior. Modeled data is Google's statistical inference about users it can't directly identify, based on behavioural patterns and probability distributions. As cookie coverage has declined and privacy regulations have expanded, modeled data now accounts for an estimated 62% of Google's audience pool. When you set a demographic bid modifier, you're applying it to a segment that is largely a statistical model, not a directly observed population.

Customer Match is the most underused high-signal audience tool in Google Ads. It lets you upload hashed first-party data — email addresses, phone numbers, physical addresses — and match them against Google's signed-in user base. The result is an audience segment built entirely from people you actually know: existing customers, lapsed buyers, high-value subscribers, trial users who didn't convert. You own the data. Google provides the reach.

The performance advantage is structural, not incidental. A Customer Match segment built from your highest-LTV customer cohort is telling Smart Bidding exactly what a high-value conversion looks like in practice — not statistically inferred, not demographically proxied, but directly observed from your own transaction data. When you bid more aggressively against that segment and combine it with a strong in-market signal, you're stacking two high-confidence intent indicators rather than relying on a single modeled demographic approximation.

Three Customer Match lists every account should have active, in order of priority: your existing customer base (for exclusion from acquisition campaigns and bid uplift in retention campaigns), your lapsed customer segment defined by recency cutoff relevant to your purchase cycle, and your highest-LTV cohort as a positive bid signal in your core conversion campaigns. If you have none of these active today, the first-party advantage your competitors are building is compounding every month you wait.

An intent-first audience strategy replaces the demographic layer with a stacked set of behavioural signals, each contributing a different dimension of purchase readiness. The goal isn't to eliminate audience targeting — it's to make sure every audience layer you're running is earning its place with measurable signal quality.

1. 1

   Base layer — RLSA (Remarketing Lists for Search Ads): Add all site visitors segmented by recency and depth of engagement. Homepage visitors, product page visitors, and cart abandoners should be in separate lists with separate bid adjustments. This is your warmest intent signal — someone who has already expressed interest in you specifically.

2. 2

   Second layer — Customer Match: Apply your high-LTV customer list as a positive bid modifier on core conversion campaigns. Apply your existing customer list as an exclusion on pure acquisition campaigns. These two adjustments together give Smart Bidding a cleaner signal on what a valuable new conversion actually looks like.

3. 3

   Third layer — In-market audiences: Layer Google's in-market segments that map most directly to your product category as observation-mode modifiers. Watch the data for 4–6 weeks before committing to bid adjustments. In-market segments vary significantly in quality by vertical — some are strong predictors, some are noise.

4. 4

   Fourth layer — Seasonal intent signals: For accounts with seasonal purchase cycles, add custom intent audiences or adjust in-market bid modifiers during peak windows. Intent signals are stronger and more reliable during high-purchase-intent periods — the same demographic profile behaves very differently in Q4 versus Q2.

5. 5

   Remove or demote — Demographic modifiers: Run a 30-day performance audit on every active demographic modifier in your account. For any segment showing less than 5% conversion rate lift versus the unmodified baseline, remove it. Demographic noise in your bid modifier stack degrades Smart Bidding's signal clarity.

Three audience tactics appear in a significant share of accounts audited and consistently show poor returns relative to their complexity cost. They're not catastrophic — they're quietly inefficient, which makes them harder to spot and easier to leave in place.

The first is blanket age exclusions. Excluding the 18–24 or 65+ age brackets as a default practice made more sense when demographic data was observed. Applied to a 62%-modeled audience pool, a blanket age exclusion is as likely to exclude real high-intent users who've been mis-modeled as it is to exclude genuinely out-of-market searchers. If your conversion data genuinely shows an age bracket performing below threshold, exclude it — but verify the data is sufficient (minimum 500 conversions per segment) before treating it as a rule.

The second is household income targeting as a primary strategy. HHI targeting sounds precise — top 10%, top 30% — but the observed/modeled split is particularly unfavorable here. Geographic income proxies and behavioral inference dominate HHI classification. If you're applying aggressive bid downlifts to lower HHI brackets based on the assumption that they can't afford your product, you're likely suppressing real demand based on a statistical model with meaningful error rates.

The third is running in-market audiences without a performance review cadence. In-market segments are not static — Google redefines their composition regularly, and a segment that performed well 18 months ago may have expanded or contracted in ways that no longer align with your conversion profile. Review in-market segment performance quarterly and be willing to remove segments that have stopped earning their bid modifier.

Moving from a demographic-first to an intent-first audience strategy doesn't require blowing up what's working. The transition is additive at first — you build the intent stack alongside existing demographic settings, gather comparison data, and phase out the underperformers once the evidence is clear. Done properly, it takes about 90 days from first upload to a fully intent-first account.

1. 1

   Month 1 — Activate Customer Match: Upload your existing customer list and apply it as an exclusion on acquisition campaigns. Upload your high-LTV cohort and add it in observation mode to your core conversion campaigns. Don't change bids yet — just start collecting segmented performance data.

2. 2

   Month 1 — Audit demographic modifiers: Export every active audience modifier in the account. Tag each as intent-based or identity-based. For identity-based modifiers, note the last time performance was reviewed and the conversion sample size behind any active adjustment.

3. 3

   Month 2 — Layer in-market audiences: Add your top three in-market segments in observation mode across core campaigns. Review performance at the end of the month. Promote segments showing 10%+ conversion rate lift to active bid modifiers. Leave the rest in observation.

4. 4

   Month 2 — Begin demographic pruning: For any demographic modifier with fewer than 200 conversions in the last 90 days, remove it. For modifiers with sufficient data showing sub-5% lift, remove them. Document what you removed and why — this creates your account's audience decision log.

5. 5

   Month 3 — Activate Customer Match bid adjustments: With 60 days of segmented data, you now have a performance baseline for your Customer Match lists. Apply bid uplifts to segments showing meaningful conversion rate improvement. Review RLSA segments and adjust bid modifiers based on the same 60-day baseline.

By the end of month three, your account's audience strategy should be grounded entirely in signals you can defend with performance data — not demographic assumptions that made sense under a different data regime. The accounts that make this transition now are building a first-party data asset that compounds over time. The ones that don't are increasing their dependence on Google's modeled audiences at exactly the moment that model is becoming less reliable.