YouTube Audience Targeting: When Customer Data Beats Contextual Targeting
A lookalike audience built from qualified property enquiries achieved a 64.33% video completion rate at €0.01 per view, outperforming contextual targeting using virtually identical creative. Here’s what the comparison revealed about audience selection, customer data and the limitations of video engagement metrics.
YouTube offers advertisers an enormous range of targeting options. We can reach people based on their interests, demographics, search behaviour, website interactions, the content they consume and similarities to existing customers.
With so many options available, it’s tempting to assume that the more precisely we define an audience, the more relevant our advertising becomes. But what happens when an audience built from actual customer behaviour performs better than one targeting seemingly relevant content?
I recently had an opportunity to explore this while managing YouTube advertising for an overseas property exhibition. The campaigns were promoting the same type of event, using virtually identical creative, but approached audience selection differently.
The results were striking, an audience built from high-quality property enquiries achieved a 64.33% video completion rate at an average cost per view of €0.01, compared with 56.56% and €0.03 for an audience based on keywords, websites and placements.
The comparison raises an interesting question about how we approach YouTube targeting.
Are we better off trying to identify relevant content, or identifying people who resemble those already demonstrating meaningful commercial interest?
The campaign background
The advertiser was promoting an overseas property exhibition aimed at people interested in purchasing property abroad. The objective was to reach prospective buyers, introduce them to the exhibition and encourage attendance.
This wasn’t a completely new advertising activity. A previous agency had run YouTube advertising for the exhibitions, giving us an opportunity to review the earlier approach alongside the more recent campaigns.
One particularly useful aspect of the comparison was the creative. The exhibition videos followed an established template, with the date and venue details updated for each event. The underlying creative was virtually identical, meaning we weren’t comparing completely different advertising messages or production approaches.
That doesn’t make this a controlled experiment. The campaigns ran at different times and used different configurations, which matters when interpreting the results. However, it does remove one obvious complication: substantially different creative wasn’t the primary distinction between the approaches.
The previous approach: Websites and expanded targeting
The previous agency had used a Demand Gen campaign with Maximise Clicks bidding, restricted to YouTube inventory. Its targeting included website-based signals alongside expanded or optimised targeting. On the surface, that isn’t an unreasonable starting point.
If we’re promoting an overseas property exhibition, identifying relevant websites and content associated with property abroad, overseas homes and related interests appears logical.
The difficulty is that relevance to a particular website or topic doesn’t necessarily establish that someone is a suitable prospective property buyer.
Someone consuming content about overseas property could be researching a purchase, browsing aspirationally, investigating relocation or simply interested after having seen something on TV. Those behaviours can overlap, but they don’t necessarily represent the same commercial opportunity.
The previous campaign’s viewing metrics also raised questions. Its website-targeted audience recorded a 4.49% video completion rate. The expansion and optimised targeting segment recorded just 0.79%. The latter also had a higher reported cost per view.
| Previous campaign audience | Avg CPV | Video played to 50% | Video played to 100% |
|---|---|---|---|
| Website targeting | €1.45 | 10.26% | 4.49% |
| Expansion and optimised targeting | €2.11 | 1.59% | 0.79% |
These figures are worth examining, but they shouldn’t be treated as proof that expansion causes poor performance.
The campaign was using Maximise Clicks, so its optimisation objective was different from simply generating completed video views. Its audience delivery and inventory mix may also have differed from the subsequent activity. Nevertheless, the results prompted a useful question: could a different approach to audience selection produce stronger engagement with the same type of creative?
A different starting point: People rather than content
For the more recent campaign, I wanted to compare two approaches to reaching prospective overseas property buyers.
- The first was contextual targeting, using relevant keywords, websites and placements.
- The second used a lookalike audience built from first-party conversion data.
Importantly, this wasn’t simply a list of everyone who had ever visited a property website or submitted an enquiry. The source audience came direct from the client, an overseas property business with an established enquiry process.
When prospective buyers submit an enquiry, the information they provide is used to assess the quality of that opportunity. Factors such as funding arrangements and travel plans help distinguish people who may be closer to purchasing from those whose interest is less immediate.
The lookalike audience was built from enquiries classified as high-tier conversions. That distinction matters. A conversion in Google Ads tells us that somebody completed an action we’ve chosen to measure. It doesn’t automatically tell us how commercially valuable that action was.
Two people can submit the same property enquiry form while having very different levels of purchase readiness. One might have suitable funding and be travelling imminently. Another might still be exploring the possibility of purchasing abroad without any immediate plans.
Both represent enquiries, but they may not represent equivalent commercial opportunities.
By using the high-tier conversion group, the intention was to give Google’s audience modelling a starting point informed by stronger commercial signals rather than enquiry volume alone.
The contextual audience remained useful as a comparison. It allowed us to examine whether reaching people through relevant content would produce similar engagement to reaching people modelled on qualified enquiries.
Lookalike beats contextual by 66% on CPV
The campaign ran from August to October 2026, with a clear difference in the reported viewing metrics between the two audiences.
| Metric | High-tier lookalike | Contextual targeting |
|---|---|---|
| Average CPV | €0.01 | €0.03 |
| Average CPM | €3.98 | €13.63 |
| Average viewable CPM | €5.08 | €16.38 |
| Video played to 25% | 86.45% | 80.05% |
| Video played to 50% | 73.78% | 68.03% |
| Video played to 75% | 68.17% | 60.38% |
| Video played to 100% | 64.33% | 56.56% |
The lookalike audience achieved a 64.33% video completion rate, compared with 56.56% for contextual targeting. That’s a difference of 7.77 percentage points, or approximately 13.7% higher relative completion.
The cost difference was even more pronounced.
At €0.01, the lookalike audience’s reported average cost per view was approximately 66% lower than the contextual audience’s €0.03.
Its average CPM was also approximately 71% lower.
The lookalike performed better at every reported video completion milestone, from 25% through to 100%.
Most of the campaign’s delivery went towards the lookalike audience, so the overall campaign figures were naturally much closer to its performance. That doesn’t mean the contextual audience was unsuccessful. A 56.56% completion rate is not inherently poor, and without an agreed benchmark or commercial outcome, we shouldn’t describe it that way.
What we can say is that, within this campaign, the audience modelled on qualified property enquiries generated stronger viewing engagement at a substantially lower reported cost.
The audience informed by qualified property enquiries achieved higher video completion rates and lower viewing costs than the contextual audience using the same creative.
Why the quality of the source audience matters
The interesting part of this comparison isn’t simply that a lookalike audience performed better. It’s the information used to create it.
First-party data is frequently discussed as though collecting more of it is automatically beneficial. But the quality of that data matters just as much as its volume. If we build audiences from every conversion, we’re effectively asking the platform to find more people resembling everyone who completed that action.
That may include valuable customers, poorly qualified enquiries, casual browsers and people who were never likely to become commercially meaningful prospects. The algorithm doesn’t inherently understand the commercial differences between those people unless we’re providing signals that help distinguish them. In this case, the business already had a method of identifying stronger enquiries.
Using that qualification to inform audience selection created a closer connection between the advertising strategy and the characteristics of the prospects the business actually wanted to reach.
It also highlights an opportunity that can easily be overlooked. We spend considerable time deciding which audiences to select inside Google Ads but sometimes the more important decision happens before we open the platform.
Which customer behaviours should we use to define the audience in the first place?
Why contextual targeting still has a place
I wouldn’t interpret these results as a reason to abandon contextual targeting. There are situations where contextual relevance can be particularly valuable, especially when a business has limited first-party data, is entering a new market or wants to reach people consuming content closely related to its product or service. Contextual targeting also provides a different route to discovery.
Someone might be researching overseas property for the first time without resembling the customers who have previously submitted qualified enquiries.
An approach focused exclusively on existing customer characteristics could potentially overlook those people.
There are also practical limitations to lookalike modelling. Audience size, source-data quality, campaign eligibility and the ability to reach genuinely new prospects all need consideration.
The question isn’t which targeting approach is universally better. It’s which one the evidence shows is right for your campaign.
The question shouldn’t be whether contextual targeting or customer-data-led targeting is universally better. It’s whether you have enough evidence to understand which approach is more appropriate for the campaign you’re running.
In this example, the lookalike audience produced stronger viewing results. That makes it a valuable finding, not a universal rule.
Putting the results into context
The objective of the video campaign was to reach relevant prospective property buyers, encourage them to consume the exhibition video and ultimately take action.
With that in mind, the viewing results are particularly encouraging. The high-tier lookalike audience delivered substantially more completed views relative to impressions, at a significantly lower reported cost per view than contextual targeting. More people consuming the advertising message was an important part of what we were trying to achieve.
There are still some differences worth acknowledging when interpreting the results. The previous agency used Demand Gen with Maximise Clicks bidding, while the more recent activity took a different approach. We therefore shouldn’t attribute the entire performance difference between the campaigns to audience targeting alone.
The comparison between the lookalike and contextual audiences within the recent campaign provides a clearer indication of how the different targeting approaches performed.
Of course, video engagement is one part of the customer journey. We’d also want to understand how that engagement contributed to exhibition registrations, attendance and ultimately property enquiries.
But that doesn’t diminish the value of what the viewing data tells us. When the goal includes getting relevant people to consume your advertising message, the ability to generate substantially more completed views at a lower cost is a meaningful performance improvement.
What I’d take into the next campaign
The biggest takeaway from this experience is that audience planning should begin with the customer information a business already has, rather than the targeting options available inside Google Ads.
Before building audiences, I’d look at the conversion data and ask whether all those conversions represent equally valuable prospects. If they don’t, there’s an opportunity to use that distinction to inform targeting.
In this case, separating high-tier property enquiries from general enquiries gave us a more commercially relevant starting point for audience modelling. That doesn’t necessarily require an elaborate scoring system. It could be as simple as distinguishing qualified enquiries from unqualified ones, identifying customers who actually purchased or separating higher-value transactions from lower-value activity.
I’d also continue testing customer-data-informed audiences alongside contextual alternatives rather than assuming either approach will always perform better. Using comparable creative, campaign objectives and measurement settings makes it easier to understand what is driving the differences.
The results reinforced the importance of evaluating audiences against the campaign’s actual goals. Video completion rates and viewing costs were particularly relevant here because encouraging prospective buyers to consume the exhibition message was an important part of the strategy. Subsequent actions and commercial outcomes remain valuable additional measures.
Ultimately, I’d resist the temptation to equate more targeting options with a better strategy. Google’s audience capabilities are extensive, but sometimes the most useful targeting information isn’t another option inside the platform.
It’s what the business already knows about its most valuable customers.
Better audience targeting starts before Google Ads
In this campaign, an audience modelled on high-quality property enquiries delivered a 64.33% video completion rate at just €0.01 per view, outperforming contextual targeting on both engagement and cost.
The difference wasn’t more complicated targeting or additional layers of audience selection. It was the decision to use commercially qualified customer data as the starting point. That doesn’t make contextual targeting redundant, but it demonstrates why audience strategy shouldn’t begin and end with the options available inside Google Ads.
We spend considerable time deciding how to reach people, but perhaps we should spend more time understanding which people we’re actually trying to reach and what makes them valuable to the business.
The most valuable targeting insight might not come from Google Ads at all. It might come from what your business already knows about its customers.
