Google Ads Audiences Guide: Observation vs Targeting & Advanced Assignments

There is a setting inside Google Ads that can change the whole shape of a campaign.

It looks small.

It sounds technical.

But it can decide whether your campaign reaches the full market or only a narrow slice of it.

That setting is the choice between Targeting and Observation.

It used to be easier to explain as "Target & Bid" versus "Bid Only."

The names have changed.

The risk has not.

If you select Targeting by mistake, you tell Google:

"Only show my ads to people who search for my keywords AND are in this audience."

If you select Observation, you tell Google:

"Show my ads based on the normal campaign targeting, but also report how these selected audiences perform."

That difference matters.

One setting restricts reach.

The other gives you data.

One setting can be useful for a deliberate sniper campaign.

The other is usually safer for standard Search campaigns.

For most Search campaigns, Observation should be the default.

Not because audiences do not matter.

They matter a lot.

But because Search already has intent.

If someone searches for "emergency plumber near me", you usually do not want to block them just because they are not inside a selected audience list.

They have shown intent.

That matters.

For 95% of Search campaigns, "Targeting" can be dangerous for volume unless it is chosen deliberately.

Yet, we still see it selected in audit after audit.

Sometimes it was set by mistake.

Sometimes it was copied from an old campaign.

Sometimes a manager misunderstood the setting.

Sometimes an imported campaign carried it across.

Sometimes nobody checked.

The result is often the same.

Impressions drop.

Clicks drop.

Conversions drop.

The business thinks demand has disappeared.

But demand did not disappear.

The campaign was restricted.

In this "Mega-Authority" guide, we will move beyond the basics. We will cover:

1. The Audience Layering Onion strategy.
2. GA4 Predictive Audiences (Purchase Probability).
3. Customer Match v2.0 (Hashing & Match Rates).
4. The "Sniper" Strategy using Combined Segments.

The aim is simple.

Use audiences to understand people better.

Do not accidentally hide your ads from the people already searching for you.

---

Part 1: The Logic of Restrictions (Targeting vs. Observation)

Understanding the logic is critical.

This is not just a platform setting.

It is a reach decision.

Targeting (AND Logic)

• Condition: Keywords + Audience.
• Result: Ad shows ONLY if both are true.
• Use Case:
  • RLSA (Remarketing Lists for Search Ads): Target existing visitors with generic keywords like "software" which would be too expensive otherwise.
  • Sniper Campaigns: Showing broad or expensive keywords only to a very specific audience.

Targeting is restrictive.

That can be good.

But only when it is intentional.

For example, bidding on the broad keyword software is usually too vague.

It could mean almost anything.

But bidding on software only for people who visited your enterprise pricing page in the last 30 days is different.

Now the audience adds context.

The keyword is broad.

The audience creates qualification.

That is a valid use case.

Another example is remarketing search.

Someone who has already visited your website and now searches a broad category term may be worth more than a cold user.

In that case, Targeting can make sense.

But if you apply Targeting to a normal Search campaign without thinking, you may block a large amount of valid demand.

That is the danger.

Observation (OR Logic + Reporting Overlay)

• Condition: Keywords.
• Result: Ad shows if the keyword and campaign targeting match. Audience data is recorded when users also belong to selected audiences.
• Use Case: Every standard Search Campaign where you want insight without restricting reach.

Observation is usually safer because it does not narrow your reach.

It lets you see how selected audiences perform.

You can learn whether:

1. Cart abandoners convert better.
2. Past visitors have lower CPA.
3. Pricing page visitors are more valuable.
4. In-market segments convert better than average.
5. Certain demographics perform poorly.
6. Customer Match lists show stronger ROAS.
7. YouTube viewers later search and convert.
8. High-value audiences deserve separate campaigns.

This is why Observation is powerful.

It does not stop you from buying intent.

It helps you understand the people behind the intent.

(Reference: https://support.google.com/google-ads/answer/7365594)

The human point is important.

A search query is not just text.

It comes from a person.

That person may be new to your brand.

They may have visited yesterday.

They may be comparing competitors.

They may be an existing customer.

They may be a high-value prospect.

They may be outside your target market.

The keyword tells you what they want right now.

The audience helps you understand who they might be.

Good Google Ads management uses both.

---

Part 2: The Financial Impact of Audience Data

Why bother with audiences if you are running Search Ads?

Because not all clicks are equal.

Two people can search the same phrase and have very different value.

A user searching for "CRM software" is worth $X.

A user searching for "CRM software" who also visited your pricing page yesterday may be worth much more.

They are not just searching.

They are returning.

They have context.

They may already know your brand.

They may be closer to a decision.

If you do not use audience data, you may treat these users the same in your reporting.

That can hide value.

It can also hide waste.

A cold prospect may need education.

A returning visitor may need proof.

A past customer may need an upsell.

A cart abandoner may need reassurance.

A pricing page visitor may need a demo.

Audience data helps you see these differences.

The Bid Adjustment Formula

In manual bidding or campaigns where bid adjustments apply, you can use a simple formula:

Bid Adj % = (Average Account CPA / Audience CPA - 1) * 100

Example:

• Average CPA: $100
• "Cart Abandoners" Audience CPA: $50

Adj = (100 / 50) - 1 = 1 (or +100%)

In that example, the audience is converting at half the average CPA.

A higher bid may be justified.

But use this carefully.

Do not make large bid adjustments on tiny data.

Do not bid up an audience after one conversion.

Do not assume last month will repeat forever.

Wait for enough data.

Look at conversion volume.

Look at revenue quality.

Look at lead quality.

Look at margin.

If you use Smart Bidding such as tCPA or tROAS, Google may already consider many signals at auction time.

So you do not always need manual bid adjustments.

But adding audiences as Observation still has value.

It gives you visibility.

It helps you report.

It helps you identify patterns.

It helps you decide whether an audience deserves its own campaign, exclusion, creative angle or landing page.

The key point is this:

Audience data helps you stop treating every searcher as the same person.

That is where the financial value begins.

---

Part 3: Framework - The Audience Layering Onion

You should rarely launch a serious search campaign with no audience thinking at all.

Use the Audience Layering Onion to capture useful signals without restricting reach.

The goal is to add audiences as Observation first.

Then let the data show what matters.

Layer 1: The Core (First-Party Data)

• All Visitors (540 Days): The "catch-all" net.
• Cart Abandoners (30 Days): High intent.
• Pricing Page Visitors: Consideration phase.
• Past Purchasers: LTV plays.
• YouTube Viewers: Brand aware.

First-party data is the strongest layer.

These people have already interacted with you.

They may have visited the website.

They may have bought before.

They may have watched your videos.

They may have started checkout.

They may have read key content.

They are not strangers.

This makes them valuable.

But not all first-party audiences are equal.

An all visitors list is broad.

A pricing page visitor is stronger.

A cart abandoner is stronger again.

A past purchaser may be valuable for repeat sales, cross-sell or exclusion.

For lead generation, you might build audiences such as:

1. All website visitors.
2. Service page visitors.
3. Contact page visitors.
4. Form starters.
5. Form submitters.
6. Blog readers.
7. Returning visitors.
8. CRM qualified leads.
9. Closed won customers.
10. Disqualified leads.

These audiences can help with reporting, remarketing and exclusions.

Layer 2: The Adjacent (In-Market & Life Events)

• In-Market for [Your Category]: They are actively researching.
• In-Market for [Competitor Category]: They are comparing.
• Business Creation (Life Event): Useful for selected B2B services where available.
• Moving (Life Event): Useful for Home Services where available.

In-market audiences can be useful because they suggest active interest.

They are not as strong as first-party data.

But they can help you understand demand.

For example:

A person searching for "business insurance" who is also in-market for business services may be more valuable.

A person searching for "kitchen showroom" who is also in-market for home renovation may be more relevant.

A person searching for "hotel wedding venue" who has shown travel or event planning signals may be more useful.

Do not assume all in-market segments will work.

Add them.

Observe them.

Let the data prove it.

Layer 3: The Broad (Affinity & Demographics)

• Technophiles / Luxury Shoppers: Lifestyle indicators.
• Detailed Demographics:
  • Homeowners where available.
  • Education Level.
  • Parental Status.
  • Employment: Company Size, Industry where available and relevant.

Broad audiences are weaker signals.

But they can still reveal patterns.

For example:

Luxury shoppers may perform better for premium products.

Parents may perform better for family cars.

Homeowners may perform better for home improvement.

Company size may matter for B2B software.

Industry may matter for specialist services.

Do not over-trust these signals.

Use them as evidence.

Not as identity.

Action: Add relevant audience layers to your Search Campaigns as Observation. It costs nothing in reach. It gives you useful reporting.

This does not mean adding random audiences just to fill the account.

Add audiences that have a business reason.

Then review them properly.

The value is not in adding audiences.

The value is in learning from them.

---

Part 4: GA4 Predictive Audiences (The AI Layer)

Old-school remarketing relies on simple rules.

For example:

"Visited pricing page."

"Started checkout."

"Viewed product."

"Visited in last 30 days."

These are useful.

But they are based on what already happened.

GA4 predictive audiences try to go further.

They use predictive metrics when your property has enough eligible data.

If you have Google Analytics 4 (GA4) linked to Google Ads, and sufficient volume, you may be able to create "Predictive Audiences."

Key Segments to Use:

1. Likely to Purchase (7 Days): Users who haven't converted yet but exhibit behaviour similar to converters.
2. Likely to Churn (7 Days): Users who are likely to stop visiting. Useful for win-back thinking.
3. Predicted Top Spenders: High LTV targets where available.

These audiences are not available in every account.

GA4 needs enough data to calculate predictions.

If the property has low traffic or low conversion volume, predictive metrics may not be available.

That is normal.

Do not build a strategy that depends on predictive audiences unless the account has the data to support them.

Implementation:

1. Go to GA4 Admin → Property → Audiences.
2. Click "New Audience" → "Predictive".
3. Sync to Google Ads.
4. Add as Observation first and review performance before making stronger decisions.

You can also use predictive audiences for remarketing campaigns.

For example:

1. Show a stronger offer to likely purchasers.
2. Show reassurance content to high-intent users.
3. Exclude recent purchasers from acquisition campaigns.
4. Build YouTube sequences for likely high-value users.
5. Test different creative for likely churn audiences.

But be careful with value and privacy.

Predictive does not mean certain.

A user is not guaranteed to purchase.

They are more likely, based on available signals.

Use these audiences as a helpful layer.

Not as a perfect truth.

For SEO, AEO and GEO thinking, predictive audiences also teach a useful lesson.

The best marketing systems do not only ask what someone searched.

They ask what the person is likely trying to do next.

That is where audience strategy becomes more human.

---

Part 5: Customer Match v2.0

First-party data is king in a privacy-first world.

Customer Match allows you to upload customer data such as emails, phone numbers or addresses, subject to Google's policies and eligibility requirements.

This can be used for:

1. Remarketing.
2. Exclusions.
3. Upsell campaigns.
4. Customer retention.
5. New customer acquisition signals.
6. Audience observation.
7. YouTube campaigns.
8. Search audience analysis.
9. Performance Max signals where appropriate.

But Customer Match must be handled responsibly.

You need permission to use the data.

Your privacy policy should explain how customer data may be used for advertising.

You should follow Google policy.

You should not upload data you are not allowed to use.

Trust matters.

People gave you their data for a reason.

Do not treat it casually.

The "Match Rate" Problem

Advertisers often complain about low match rates.

For example, uploading 1,000 emails and getting 200 matches.

There are several reasons this can happen.

• Reason: Users register with work emails (john@company.com) but are logged into Google with personal emails (john@gmail.com).

Other reasons include:

1. Old data.
2. Typos.
3. Missing phone numbers.
4. Poor formatting.
5. Duplicate records.
6. Country or postcode formatting issues.
7. Users not signed in.
8. Emails not connected to Google accounts.
9. Low list size.
10. Data not refreshed often enough.

Google says most advertisers' match rates sit within a broad range and recommends using multiple sources of customer information in the same row to improve matching. :contentReference[oaicite:1]

The Fix: Enrichment

Before uploading your list:

1. Map Multiple Data Points: Don't just upload Email. Upload Email + Phone + First Name + Last Name + Zip Code where you have permission and clean data.
2. Normalization: Ensure formatting matches Google's requirements. Use SHA256 hashing if using the API, or use the interface where Google handles hashing.
3. Keep it Live: Use CRM integrations, scheduled uploads or specialist tools to keep lists fresh. A 6-month old list may be much less useful.

Freshness matters.

Google's Customer Match policy notes list membership has a maximum duration and recommends refreshing lists regularly. :contentReference[oaicite:2]

A strong Customer Match setup should not be one giant list.

Segment it.

For example:

1. All customers.
2. High-value customers.
3. Repeat customers.
4. Recent customers.
5. Lapsed customers.
6. Qualified leads.
7. Sales opportunities.
8. Closed won customers.
9. Disqualified leads.
10. Newsletter subscribers.

Each list has a different use.

High-value customers can support value-based bidding and similar signals.

Current customers can be excluded from acquisition.

Lapsed customers can be targeted with win-back offers.

Qualified leads can be used to study which audiences perform better.

Disqualified leads can be excluded from some campaigns.

Customer Match becomes powerful when it reflects business reality.

Not when it is treated as a dumping ground.

---

Part 6: Advanced Strategy - Combined Segments (The "Sniper")

What if you want to reach people who are "In-Market for SUVs" AND "Parents"?

That is where Combined Segments can help.

A combined segment lets you narrow by multiple audience conditions.

This can be useful when a keyword is too broad by itself.

The Recipe

1. Go to Tools → Audience Manager → Combined Segments.
2. Click New Combined Segment.
3. Group 1: "In-Market for SUVs"
4. AND (Click "Narrow your segment")
5. Group 2: "Parental Status: Parents"
6. Name it: "Parents Shopping for SUVs".

This is not for every campaign.

It is a specialist tactic.

Use it when you want strict control.

The "Generic Keyword" Sniper Strategy

Broad match keywords like best cars are usually too expensive and vague.

But if you layer them with a Combined Audience using Targeting mode, the economics can change.

• Keyword: best cars (Broad Match)
• Audience: "Parents Shopping for SUVs" (Targeting)
• Result: You show up for a generic query, but ONLY to a more qualified user.
• Benefit: Lower waste from broad keywords because the audience adds qualification.

This can work well for:

1. Automotive.
2. Higher education.
3. B2B software.
4. Finance.
5. Home improvement.
6. Luxury products.
7. Travel.
8. Specialist services.

But it can also limit volume.

That is the trade-off.

Sniper campaigns are not built for scale first.

They are built for control first.

Use them when broad reach is too wasteful.

Do not confuse this with your main Search campaign.

Your main Search campaign usually wants intent and volume.

Your sniper campaign wants strict relevance.

---

Part 7: B2B Special - Detailed Demographics

For B2B advertisers, Google Ads can feel less precise than LinkedIn.

LinkedIn has job titles, companies and professional data.

Google has search intent.

Those are different strengths.

But Google Ads also has audience and demographic signals that can help.

Hidden B2B Signals:

Under "Detailed Demographics" > "Employment", where available, you may see options such as:

1. Company Size: Small, Large, Very Large.
2. Industry: Technology, Financial, Hospitality, and other categories.

Availability varies by country, account and campaign type.

So check what is actually available in your account.

Pro Tip: If you sell Enterprise Software, you may test excluding very small company size segments if the data clearly shows they cannot afford you.

But do not exclude too early.

A small business founder may still be high value.

A startup may become a strong customer.

A consultant may search on behalf of a larger client.

Audience data is useful.

It is not perfect.

For B2B, Search intent is still one of the strongest signals.

A person searching "enterprise payroll software demo" may be more valuable than a person who merely fits a demographic profile.

So use demographics carefully.

A better B2B approach is:

1. Add detailed demographics as Observation.
2. Review performance by segment.
3. Compare with CRM lead quality.
4. Exclude only when there is enough evidence.
5. Use Customer Match for known high-quality leads and customers.
6. Use landing pages to qualify by company size, budget or use case.
7. Import offline conversions to teach Google which leads became real opportunities.

That is more reliable than relying only on demographic assumptions.

---

Summary: Your Audience Roadmap

Data is useful.

But careless audience settings can damage performance.

By defaulting to Targeting, you may restrict the campaign too far.

By using Observation, you can collect audience insight without narrowing reach.

That is why Observation is usually the right default for Search.

Targeting is for deliberate use cases.

Remarketing search.

Sniper campaigns.

Restricted tests.

High-value short-tail keywords.

Audience-qualified experiments.

The goal is not to add audiences for the sake of it.

The goal is to understand the human behind the search.

A searcher is not just a phrase.

They have history.

They may know you.

They may be new.

They may be ready.

They may be browsing.

They may be returning after comparing you.

They may be the wrong fit.

Audiences help you see that.

Your Action Plan:

1. Audit: Check every ad group setting. Switch "Targeting" to "Observation" unless it is a deliberate "Sniper" campaign.
2. Onion Layering: Add relevant Remarketing, In-Market, and Demographic audiences to campaigns as Observation.
3. GA4 Sync: Create predictive audiences where GA4 has enough eligible data.
4. Test: Launch a "Sniper" campaign for a broad keyword using Combined Segments.

Stop treating all searchers as anonymous text strings.

They are people with histories, jobs, needs and intent.

Use that data carefully.

Use it ethically.

Use it to make better decisions.

---

The "360 View" Audience Layering Framework

We do not guess which audiences work.

We add relevant audience segments on Observation mode and let the data decide.

Step 1: The "Kitchen Sink" Setup

When launching a new Search Campaign, add a healthy set of relevant audiences as Observation:

• 5-10 In-Market Categories relevant to your product and closely adjacent categories. Selling office chairs? Add "In-Market for Office Furniture" and relevant business-related categories where available.
• All Eligible Remarketing Lists: All Visitors, Cart Abandoners, Leads and key page visitors.
• Detailed Demographics: Homeownership, Education, Employment Industry or Company Size where available and relevant.
• 3-5 Affinity Audiences: Broad interests aligned to your buyer persona, such as "Business Professionals" where relevant.

This gives you a wider view.

But keep it sensible.

Do not add random audiences that have no commercial relationship to the business.

More data is not always better.

Useful data is better.

Step 2: Data Gathering

Let the campaign run for 3-4 weeks.

Do not make audience decisions too early.

You need enough data to see patterns.

For low volume accounts, you may need longer.

For high spend accounts, you may have useful data faster.

The goal is to avoid emotional decisions.

One conversion does not prove an audience.

One bad week does not condemn it.

Wait for patterns.

Step 3: The Analysis

After 30 days, look for patterns like:

• "People in 'In-Market for Real Estate' have a $50 CPA vs the $100 campaign average." → Consider a measured bid adjustment where manual bidding allows, or test a separate campaign.
• "'Job Seekers' affinity has high CTR but zero conversions." → Consider excluding this audience if the evidence is strong.
• "'Likely C-Level Executive' demographic has 3× ROAS." → Consider a standalone campaign, tailored landing page or higher value rule if supported by data.

Do not analyse only CPA.

Also check:

1. Conversion volume.
2. Revenue.
3. Lead quality.
4. Sales accepted leads.
5. Close rate.
6. Average order value.
7. Lifetime value.
8. Assisted conversions.
9. Search terms.
10. Landing page behaviour.

Audience data becomes powerful when it connects to business data.

The "Bid-Only" RLSA (For High-Value Short-Tail Keywords)

RLSA can unlock keywords that would otherwise be unprofitable.

Example: You sell Enterprise SaaS and bidding on "database software" costs $50 per click.

That keyword may be too broad for cold traffic.

Solution:

1. Run cheaper Display, YouTube, organic or content activity to build an "All Website Visitors" audience.
2. Create a Search campaign for "database software".
3. Set the audience to Targeting for "All Website Visitors" or a stronger remarketing segment.

Now you are bidding on a high-volume short-tail keyword, but your ad only shows to people who already know your brand.

Conversion rate can improve because the audience is warmer.

But use this carefully.

If the remarketing list is too small, volume will be limited.

If the list is too broad, quality may still be weak.

If the landing page is poor, the tactic will not save it.

Customer Match & The Privacy-First Future

With privacy changes and limitations on third-party tracking, First-Party Data becomes more important.

Upload your email list via Customer Match where you have permission and meet Google policy requirements.

Customer Match can help with:

• Observation: Understand how known leads or customers behave in Search.
• Exclusions: Exclude current customers from user acquisition campaigns where appropriate.
• Upsell: Target current customers with new feature announcements on YouTube, Demand Gen or Search where suitable.
• Retention: Reach lapsed customers with relevant offers.
• Value Signals: Support smarter bidding when lists reflect high-value customers.

Warning: Ensure your privacy policy and consent approach support your use of customer data for advertising before enabling Customer Match.

The Audience Hierarchy by value:

First-Party Data (Customer Match/Remarketing) → Custom Segments → In-Market & Life Events → Affinity & Demographics.

Build from the strongest data first.

Then use broader signals carefully.

The future of Google Ads is not only keywords.

It is intent plus data.

The keyword tells you what someone wants.

The audience tells you what context they may bring.

The conversion data tells you whether it mattered.

Use all three.