What audiences do in a search campaign
A search campaign targets searches. Its keywords and their match typesSettings that decide how closely a search must match a keyword for the ad to enter the auction. decide which queries the ads can enter, and the keyword match types guide covers that decision; the search campaigns hub shows how keywords, ads, bids and settings fit together around it. Nothing about an audience segment changes the query. What a segment adds is a second description of the auction: who is searching, as Google estimates it from behaviour or as the business knows it from its own records.
Google's audience targeting documentation describes segments as groups of people with particular interests, intents and demographic characteristics, as Google estimates them, alongside segments built from the advertiser's own data, and they can be added to campaigns and ad groups. On display and video they are usually the targeting itself, which is why the display targeting guide and the YouTube audience targeting guide treat them as the main decision. In search they are a modifier. The same query typed by a long-standing customer and by a stranger is the same query, but it is not the same prospect, and the segment is how the account can tell the two apart.
That gives audiences two jobs in search. The first is learning: showing which kinds of searcher convert better or worse on the same keywords. The second is acting on what was learned: restricting a campaign, building a separate campaign for a valuable group, excluding people who should not see an ad, or, under manual bidding, changing bids. The marketing funnel framework is a useful way to hold the two descriptions together: the keyword says where the search sits in the journey, and the segment says whether this person has already met the business.
Observation versus targeting
Every segment added to a search campaign or ad group carries one of two settings, and the choice matters more than which segments are chosen. Google documents targeting as narrowing the campaign's reach to the people in the selected segments, and observation as leaving reach unchanged while reporting on those people. The YouTube guide states the same distinction for video campaigns.
| Setting | What it does to delivery | What the report shows | When it fits in search |
|---|---|---|---|
| Observation | Nothing: the campaign enters the same auctions as before | Results for people in each segment, beside the campaign's totals | The default for almost every search campaign; free audience research |
| Targeting | Restricts the campaign to people in the selected segments | Results for those people only, because no one else is reached | A campaign built for one group, such as past visitors on broader keywords |
The commonest error is a segment added with the wrong setting. A search campaign that suddenly spends a fraction of its budget, or stops showing for colleagues who are not on a remarketing list, has often had its audiences set to targeting by accident; the diagnosis for a campaign not spending its budget lists an audience applied as targeting instead of observation among the settings that cap reach. The account audit checks the setting campaign by campaign and looks for campaigns that both target and observe the same list, which is a sign that nobody decided.
Observation costs nothing in reach, so there is little reason to be sparing with it. Add the segments that might plausibly differ, let the report accumulate, and decide later. Narrowing without that report is a guess.
Which segments are worth adding
The segments available in search are the ones the display guide catalogues. What each adds to a keyword differs a great deal, because a search already carries strong evidence of intent and a segment has to tell the account something the query does not.
| Segment type | Built from | What it adds to a search | Practice view |
|---|---|---|---|
| Website and app visitors | The Google tag or app events: all visitors, product or service page viewers, basket abandoners, converters | Whether this searcher already knows the business, and how far they went | Usually the most useful segments in search; add several by depth of visit |
| Customer lists | Contact data the business uploads, hashed and matched by Google to signed-in users, as the display guide describes | Whether the searcher is already a customer | Valuable for exclusions and for separating new from returning buyers; size depends on the Match RateShare of the records in an uploaded customer or account list that an ad platform matched to its own users or companies. |
| In-market | Google's inference that a person is actively researching a category | A weak confirmation of intent the query usually already shows | Worth observing for the categories you sell; rarely worth targeting |
| Affinity | Durable interests and habits | Little that a commercial query does not already say | Observe if curious; seldom changes a decision in search |
| Detailed demographics and life events | Inferred attributes and moments such as moving home | Context where the product follows a life stage | Observe where the offer depends on the stage |
| Custom segments | Interests, keywords, sites or apps the advertiser chooses | A description of the searcher's reading or wider interests | Mainly a display and video tool; the YouTube guide covers how they are built |
The practice view in the table rests on one observation: segments built from First-Party DataData a business collects directly from its own customers, visitors and users. record a real relationship with the business, while inferred segments describe a probability. A person who viewed three product pages last week and now searches a generic category term is a different prospect from a stranger typing the same words, and the remarketing list is the only place the account can see that. Inferred segments are still worth observing, because occasionally one differs sharply, but they rarely carry the weight that site visitors and customer lists do.
Two practical limits apply to your-data segments. A list needs enough members before Google will serve against it, and a new tag has no history, so lists should be created early, before they are needed. And a customer list reaches only the share of uploaded records that Google could match; the glossary's match rate entry explains how to read that share, and the remarketing entry sets out how site and app lists are built.
Reading the audience report
The audience report shows, for each segment, the same columns as the campaign: impressions, clicks, cost, conversions, conversion rate, cost per conversion and, where values are tracked, conversion value. The reading that matters is the comparison of each segment's row with the campaign or ad group total over the same period.
Three cautions keep that comparison honest. The rows do not add up to the total, because a person can belong to several segments and many searchers belong to none that you added. Volumes are often thin, especially for your-data segments in a small account, so a difference built on a handful of conversions is noise until it holds over several months; the sample size entry explains why a small group's rate swings so far. And recent days are incomplete because of conversion lag, so read closed periods rather than the last week.
| What the report shows | Likely meaning | Response |
|---|---|---|
| Past visitors convert far better than the total, with steady volume | The keyword works harder for people who know the business | Consider a separate campaign for them on broader keywords |
| Past customers convert well but mostly on brand terms | Returning customers navigating, not new demand | Read with the brand campaign; consider excluding them from acquisition campaigns |
| An in-market segment matches the total | The segment adds nothing the query did not already say | Leave it in observation |
| A segment converts worse on the same keywords | A group the offer suits less well | Check the ad and page for that group; exclude only if it persists and the cost is material |
| Differences vanish when the period changes | Noise from small numbers | Keep observing; decide nothing yet |
The comparison is a description, not proof of cause. Past visitors convert better partly because they were already closer to buying; the segment identifies who they are, not what the ad did for them. That matters when a segment's results are used to justify spending more on it, because a campaign for people who would have bought anyway can report well and add little. The prospecting versus remarketing comparison explains why remarketing results need a holdout before they are believed.
When to switch a segment to targeting
Switching a segment to targeting inside a working search campaign is rarely right, because it removes everyone else from auctions the campaign was winning. The usual case for targeting is a new campaign built for one group.
The classic version is a remarketing search campaign: past site visitors, set to targeting, on keywords broader than the business would buy for strangers. A generic term that converts too rarely to pay for cold traffic can be worth buying for people who have already looked at the products, because their conversion rate on that term is different. That campaign needs its own budget and its own target, set from what its audience converts at rather than borrowed from the prospecting campaign, and its broader keywords should not also sit in the main campaign, so the two do not compete for the same searches.
A second case is an offer meant for one group, such as a returning-customer promotion, where showing it to strangers would mislead. A third is a test: a segment set to targeting in a duplicated campaign, compared against the original, answers a question the observation report cannot. Outside those cases, keep segments in observation and let the keywords do the targeting.
Exclusions
An excluded segment removes its members from the campaign whatever they search. In search, exclusions are mainly for people the ad should not reach: existing customers in campaigns whose offer is for new customers only, recent buyers of a product that is bought once, and staff or job applicants who inflate clicks. Excluding a segment is a stronger decision than observing it, so make it from the report rather than by default.
Two cautions apply. Excluding customers from a brand campaign can leave them looking at a rival's ad above your organic listing, so brand campaigns are usually left open to everyone. And an exclusion list built from a customer upload is only as complete as the match, so some customers will still see the ads. How to run a full existing-customer exclusion strategy across campaign types is a larger question than search alone, and the guide to excluding existing customers from prospecting covers defining an existing customer, building the lists and applying them by campaign type; the display targeting guide covers exclusions on the Display Network.
Audiences and Smart Bidding
Under Smart BiddingGoogle Ads' automated bid strategies that set each auction's bid from conversion signals., audience membership is already one of the signals the strategy reads in each auction. The Smart Bidding strategies guide lists device, location, time, query and audience membership among the auction-time signals Google documents, so the strategy can bid more for a searcher it predicts is likely to convert without being told to.
That changes what observation is for. Under manual bidding, observation segments are the basis for bid adjustments: a segment that converts better on the same keywords can be given a higher bid by hand. Under Smart Bidding, manual audience adjustments are not the lever; the account audit reads the audience report for adjustments that no longer make sense under automated bidding and treats them as cleanup. Observation segments then earn their place in three other ways: the report shows where results differ, which feeds structural decisions such as a separate remarketing campaign; the same lists support exclusions; and the comparison of segments over time shows whether the account is reaching new people or mostly the ones it already had.
What Smart Bidding cannot do is decide whether a segment deserves its own budget, offer or keywords. Those are structure and business decisions, and the report is where they start.
Segments added to an equestrian store's account (illustrative)
An equestrian store finds the segment that earns its own campaign
Saddleglint Equestrian, an invented riding-equipment store, runs a generic search campaign on product terms such as riding boots, saddle pads and horse rugs, bidding to a target CPA of £30. It adds four segments in observation: past buyers from a customer list upload (a match rate of 52%), all site visitors, basket abandoners and an in-market segment for equestrian equipment. After three months the report reads:
| Segment (observation) | Clicks | Conversions | Conversion rate | Cost per conversion |
|---|---|---|---|---|
| Campaign total | 9,400 | 282 | 3.0% | £29 |
| Past buyers | 610 | 41 | 6.7% | £14 |
| All site visitors | 2,150 | 118 | 5.5% | £17 |
| Basket abandoners | 380 | 30 | 7.9% | £12 |
| In-market: equestrian equipment | 3,300 | 101 | 3.1% | £28 |
The in-market segment matches the total and stays in observation. Past buyers convert well, but most of their conversions come from searches containing the store's name, so the team leaves them alone in the generic campaign. Site visitors are the finding: they convert at nearly twice the campaign rate on the same keywords. Saddleglint builds a second campaign targeting site visitors on broader terms it had rejected for cold traffic, such as equestrian clothing and stable supplies, with its own £15 daily budget and a target CPA of £30. In its first quarter the new campaign records 64 conversions at £24 each, and the team plans a holdout on part of the list to check how many of those sales it actually caused.
Common mistakes
- Leaving segments on targeting by accident. The campaign shrinks to the list and the budget goes unspent.
- Adding no audiences at all. Observation costs no reach, and without it the account cannot see who converts.
- Narrowing on inferred segments. In-market and affinity segments rarely add much to a commercial query.
- Reading thin rows as findings. A difference built on a few conversions needs months before it means anything.
- Treating remarketing results as caused by the ads. Past visitors were closer to buying; test before scaling.
- Excluding customers from brand campaigns. They search for you and meet a rival's ad instead.
- Keeping manual audience adjustments under Smart Bidding. The strategy already reads audience membership; the adjustments are leftovers to review.
