KEY TAKEAWAYS

The Information Scent Chain

Information Foraging Theory (Pirolli & Card, 1999) describes how users navigate information environments like animals following scent trails. Users assess "information scent"—cues that signal proximity to their goal—and follow or abandon paths based on scent strength.

In paid search, the scent chain begins before your ad appears:

The Information Scent Chain

Each transition must confirm or strengthen the scent. Any discontinuity—where the user's expectation isn't confirmed—triggers re-evaluation and potential abandonment.

Prospective Memory: The Intention Already Formed

By the time a user clicks your ad, they have already formed a prospective memory—an intention expecting fulfillment. The search query creates a mental model of what they will find. The ad reinforces or refines that model. The landing page either confirms the expectation (conversion proceeds) or violates it (cognitive strain, bounce).

This is why keywords with different information needs cannot share a landing page effectively. The user arrives with a specific expectation. A page optimized for a different expectation creates immediate friction—even if both keywords are semantically related.

The Scent Confirmation Hierarchy

On landing, users need immediate scent confirmation. In The CRO Research Advantage, I described how the hero copy at low contrast (2.0-2.1) allows pattern recognition without triggering evaluation—the user confirms "I'm in the right place" before engaging analytical circuits.

But the hero can only confirm one scent. If your ad group sends traffic with multiple different expectations, some percentage of users will land and immediately experience scent mismatch. They may stay—but they'll convert at a lower rate because the page wasn't designed for their psychology.

The Homogeneous Conversion Mechanism Principle

Keywords don't just have topics—they have conversion mechanisms. A conversion mechanism is the specific combination of information architecture, objection handling, copy framework, and psychological sequence that moves a user from landing to conversion.

Two keywords might be semantically similar but require different conversion mechanisms:

Both keywords relate to food sensitivity. Both might end in a test kit purchase. But the psychological journey is different. The "symptoms" searcher needs to understand their problem before evaluating solutions. The "best test" searcher has already accepted they need a test—they're in evaluation mode.

Grouping these keywords together forces a choice: optimize the landing page for direct purchasers (abandoning the symptom-seekers) or add symptom education (diluting the direct-purchase conversion path). Neither option is optimal.

The Weighted Average Problem

When a heterogeneous ad group sends traffic to a single landing page, the resulting conversion rate is a weighted average of the segment conversion rates—and it's mathematically inferior to proper segmentation.

The Math

Consider an ad group with two keyword types: direct-purchase keywords that would convert at 15% on a properly optimized page, and information-seeking keywords that convert at 7.5% on that same page (because the page isn't optimized for their needs).

If the traffic mix is 60% direct-purchase / 40% information-seeking:

Weighted Average Conversion Rate

Even though both segments are present, the blended rate (12%) is lower than the optimal rate for the targeted segment (15%). But the real damage is worse than this number suggests—because the 7.5% segment could have converted at a higher rate on a page designed for their psychology.

The Full Calculation

Let's work through a concrete example with $300 daily budget and $2 average CPC. (In practice, CPCs vary by keyword type—symptom keywords often have lower CPCs than direct-purchase keywords due to lower commercial intent. The $2 average is used for illustration; the principle holds regardless of CPC differences.)

Scenario A: Combined Ad Group

Scenario B: Segmented Ad Groups

Segmented Total: 22.5 conversions at $13.33 CPA

A Note on Inherent Conversion Ceilings

The example above assumes both segments can achieve 15% when properly optimized. In practice, symptom-seeking keywords—earlier in the purchase funnel—often have lower inherent conversion ceilings than direct-purchase keywords, even with perfectly optimized pages. A realistic scenario might show 15% for direct-purchase and 10% for symptom education.

This doesn't invalidate the principle—it reinforces it. Even with a lower ceiling:

• Symptom keywords at 10% on an optimized page still beat 7.5% on a mismatched page
• Segmentation reveals the true ceiling, enabling accurate budget allocation decisions
• You can now evaluate whether the symptom segment is worth its cost at its true conversion rate

The 15%/15% example is illustrative. The principle is that each segment reaches its optimal rate when properly segmented—whatever that optimal rate may be. Heterogeneous grouping guarantees sub-optimal performance for at least one segment.

The Delta

Same $300 budget. Same traffic sources. But proper segmentation produces 25% more conversions (22.5 vs 18) at a 20% lower CPA ($13.33 vs $16.67).

This is not optimization—it's arithmetic. The blended conversion rate in a heterogeneous group is mathematically inferior to segmented rates where each landing page is optimized for its conversion mechanism.

The Triple Hit: How Heterogeneity Compounds Damage

Poor ad group structure doesn't just reduce landing page conversion. It creates cascading damage at three stages—plus a hidden fourth effect on auction mechanics.

The Smart Bidding Amplification Effect

In The Bid Ceiling Model, I established the formula:

The Bid Ceiling Formula

Your conversion rate is a fixed input—your historical performance over the lookback window. Google's Smart Bidding derives the effective CPC it can bid on your behalf.

Here's where ad group heterogeneity amplifies: a lower blended conversion rate directly reduces your implied max CPC.

Same business economics. Same $25 Target CPA. But the homogeneous structure produces a 25% higher implied max CPC ($3.75 vs $3.00). In competitive auctions, that's the difference between winning impressions and losing them.

The cascade: heterogeneous grouping → lower conversion rate → lower implied max CPC → fewer won auctions → less volume at your target CPA. Ad group structure directly impacts auction competitiveness.

Strategic Optimization Sequence

Optimization should proceed in a specific order: top-down psychological grouping first, then bottom-up data refinement.

Phase 1: Top-Down Psychological Grouping

Start with the behavioral science question: what conversion mechanism does this keyword require?

For each keyword, identify:

• Information need: What does the user need to learn or confirm before converting?
• Objections: What concerns must be addressed? Trust? Price? Legitimacy?
• Psychological stage: Are they problem-aware, solution-aware, or product-aware?
• Conversion device: What page architecture moves them to action? Comparison table? Educational flow? Direct offer?

Keywords that share these attributes belong in the same ad group—regardless of semantic similarity. Keywords that differ on any major attribute should be separated.

Phase 2: Bottom-Up Data Refinement

Once psychological grouping is established, refine based on performance data:

• Geographic variation: Do certain regions convert differently? Consider separate campaigns.
• Device performance: Does mobile require different treatment?
• Time-of-day patterns: Do conversion rates vary by daypart?
• N-gram analysis: Are there search term patterns suggesting intent leakage?

Use negative keywords to prevent intent leakage between groups. If n-gram analysis shows that "symptoms" queries are matching to your direct-purchase group, add "symptoms" as a negative and route that traffic to the appropriate group.

The Pool Dynamics Constraint

Segmentation has diminishing returns. As you restrict your audience more precisely, the available impression pool shrinks—and this has mechanical consequences for CPC.

The Shrinking Pool Problem

Google's Smart Bidding tries to spend your budget at your target. If you restrict your audience too much—say, excluding all ages except 64+, or limiting to a single geographic region—the pool of eligible impressions shrinks dramatically.

With a smaller pool:

• Competition intensifies for available impressions
• Implied CPCs rise mechanically as you fight over fewer opportunities
• You may not be able to spend your full budget at your target CPA
• Delivery becomes irregular and unpredictable

Finding the Balance

The optimal segmentation balances:

The Segmentation Trade-off

Over-segmentation occurs when the conversion rate improvement from homogeneity is smaller than the CPC increase from pool shrinkage. The math:

If segmenting improves CR from 12% to 15% (25% gain) but shrinks your pool such that CPCs rise from $2.00 to $2.80 (40% increase), you've hurt your economics—even though conversion rate improved.

The math proves this. Since CPA = CPC ÷ CR:

• Before segmentation: $2.00 ÷ 12% = $16.67 CPA
• After over-segmentation: $2.80 ÷ 15% = $18.67 CPA

CPA increased by 12% despite conversion rate improving 25%. The CPC increase outpaced the CR gain. Segmentation only improves economics when: (CPC increase ratio) < (CR increase ratio).

Monitor impression share and CPC trends when segmenting. If impression share collapses or CPCs spike, you may have over-restricted. Consider:

• Recombining marginal segments
• Using on-page devices (sitelinks, sidebar navigation) to handle intent variation rather than separate ad groups
• Accepting some heterogeneity in low-volume segments

Handling Intent Leakage in Homogeneous Groups

Even a properly structured ad group will have some intent variation. Google's close variant matching means users searching for slightly different variations may trigger your ads.

Pre-Click Devices

Sitelinks can capture intent variation at the ad level. If your primary ad targets direct purchasers but you're seeing symptom-related search terms, add a sitelink for "Food Sensitivity Symptoms." Users with that intent can self-select into the appropriate path without requiring a separate ad group.

Negative keywords should be deployed aggressively. Run search term reports regularly. When you see patterns (n-gram clusters) that indicate mis-matched intent, add them as exact-match negatives and route that traffic to the appropriate group.

Post-Click Devices

On the landing page, assume some users will arrive with adjacent intents. Design escape valves:

• Secondary navigation: A sidebar or footer link to symptom information catches users who need education before purchase.
• Content sliders: A splide carousel can present multiple conversion paths without diluting the primary message.
• Conditional content: Quiz-style interactions can route users to different page sections based on their expressed needs.

These devices don't replace proper segmentation—they handle residual variance that segmentation can't eliminate.

The Bottom Line

Ad group structure isn't an organizational convenience—it's a behavioral sciences decision that determines whether your landing pages can execute their conversion mechanisms.

The Framework

The Cascade

The relationship between these papers forms a complete system:

The Bid Ceiling Model establishes that conversion rate is the primary input you control. Your CR determines your implied max CPC, which determines your competitive position in the auction.

The CRO Research Advantage establishes that behavioral science—information scent, psychological narrative arcs, trust sequences—drives landing page conversion. Research reveals what mechanisms your market has converged on.

The Behavioral Sciences of Ad Group Structure establishes that your ad group structure determines whether your landing pages can execute their behavioral science. Heterogeneous groups force pages to serve multiple mechanisms, producing weighted average conversion rates that underperform segmented approaches.

The conclusion: ad group structure isn't a traffic management decision—it's a CRO decision that directly impacts your auction competitiveness.