Connected TV Advertising Explained: A Practical Guide for Brands

A first connected TV test can deliver beautifully and still leave you unable to say whether it made any money. Reach, frequency and completion rates tell you the ads ran. They don't tell you what the ads added.

The lesson I'd pass on from our own CTV work is simple: agree how you'll judge the test before the first one runs. This guide explains what connected TV is and how it's bought, then uses a hypothetical example of two tests to show what each kind of measurement can and can't prove.

What connected TV is

Connected TV, or CTV, is TV content streamed over the internet to a television: smart TVs, streaming sticks and boxes, and games consoles. When someone watches a broadcaster's streaming service or the ad-supported tier of a streaming platform on their TV, the ads they see are CTV ads. You'll also hear OTT (over the top), which covers streamed video on any device, phones included. The two terms get used loosely.

For advertisers, the difference from broadcast TV is that CTV ads are delivered household by household over the internet. That makes targeting and measurement possible, and it means you can start with a test in selected regions or audiences instead of a national campaign. Ads usually play full screen and often can't be skipped, so high completion rates are normal on CTV.

How connected TV is bought

Directly from broadcasters and streaming platforms. You agree a campaign with the broadcaster or service. Minimum spends are usually higher, in return for premium content and sometimes better placement.

Programmatically, through a DSP. A demand-side platform lets you buy CTV inventory from many publishers at once, set targeting and budgets, and adjust while the campaign runs. Our guide to programmatic advertising and DSPs explains how the buying works.

Within programmatic, you can buy in open auctions or through private marketplaces, where a publisher offers selected inventory to invited buyers at agreed floor prices. For a first test I'd use private deals or a curated inventory list. It costs more per impression, but you can see where every ad ran.

A worked example: two CTV tests

This is an illustration, not a client case study. It imagines a US consumer ecommerce brand testing CTV. All budgets and figures are hypothetical, and the budgets cover media only: production, management and measurement costs are extra.

Test one: good delivery, unclear impact

• Budget and length: $5,000 of media over four weeks, streaming inventory on connected TVs, aimed at selected audiences
• Price: a $25 media CPM, buying about 200,000 impressions
• Reach: around 50,000 households, an average frequency of four
• Completion: about 95% of ads watched to the end

Website sessions rose from 8,000 in the four weeks before the test to 8,800 during it, a 10% increase. That looks like a win, but search, social and affiliate activity were running at the same time, so there's no way to say how much of the rise came from CTV.

The measurement report also linked 100 site visits to exposed households within seven days. Those are attributed visits: people who saw the ad and later visited. They aren't proof of visits the ad caused, because some of those people would have come anyway.

Test one was judged only after it had run, with no comparison group, so all it could show was that the ads were delivered.

Test two: adding a comparison group

The second test used another $5,000 of media over four weeks, with a refined audience and new creative. This time it ran in some regions and not in a matched set of comparison regions.

Website sessions grew by 8% in the test regions and by 5% in the comparison regions. The comparison regions show what would probably have happened anyway, so the difference, about three percentage points, is the better estimate of what CTV added.

That points to a positive contribution, but a three-point gap in site visits isn't enough to establish a reliable effect on sales. The sensible next step is more measurement before raising the budget substantially.

The one decision to make before you book anything

Decide which question the test has to answer.

• "Did we reach the right households, often enough?" Any CTV campaign can answer this. Reach, frequency and completion come straight from the platform.
• "Did it bring extra visits and searches?" You need to compare activity where and when the ads ran against where they didn't.
• "Did it make extra sales?" This needs a comparison group from day one, and usually more than one test.

In the example, test one only answered the first question. Test two began to answer the second.

How to set up a comparison group

1. Pick matched areas. Choose regions that are similar in size, sales history and the other marketing they receive. Run the ads in some and keep the rest as your comparison.
2. Record a baseline. Note sales, site visits and brand searches in both sets of areas for a few weeks before launch.
3. Keep everything else equal. Avoid launching other campaigns or offers in only one set of areas while the test runs.
4. Run it long enough. Plan for the full flight plus a short period afterwards, because people often act days after seeing a TV ad.
5. Compare the change, not the totals. What matters is how much more the test areas grew than the comparison areas, against the baseline, as in test two.

How to measure CTV

Nobody clicks a TV ad, so last-click reports will always make CTV look weak. These work better, roughly in order of how much they prove:

1. Reach, frequency and completion. Delivery only.
2. Site visits and search lift before and after. Useful, but other channels muddy it, as the 10% rise in test one showed.
3. Household matching. Exposed households linked to later visits or purchases, like the 100 attributed visits. Treat these as attributed, not caused.
4. Test and comparison areas. The most reliable of the four. It also picks up sales that arrive later through search, retargeting or affiliates, which last-click reports credit to those channels.

What you need before launch

• A TV-ready ad. Most run 15 or 30 seconds and have to work with nothing to click. Clear branding, one message and an easy next step, such as a memorable web address or search term.
• Clearance, where required. In the UK, broadcaster platforms usually need ads cleared by Clearcast before they run.
• Your measurement agreed before the first impression. The comparison regions, the baseline and how you'll judge the result. It's the one step I'd never skip.

If you're thinking about testing TV for the first time, see how we plan and buy connected TV and DSP-led media, or start a conversation.