How a CTV ad actually works.

A brand — say, a car company — wants to reach people watching live sport. They work with a media buyer (an agency or their own in-house team) to book ad slots inside a streaming service.

They agree on how many people they want to reach, how much they'll pay per thousand viewers, which shows or channels are acceptable, and how long the campaign runs.

At this point, no ad has been shown to anyone. This is the planning stage.

Once a deal is agreed, someone has to actually put the ad into the system. This is called trafficking — a word borrowed from broadcast TV.

The ad operations team uploads the video file, sets the flight dates, tells the system which audiences it's allowed to reach, and sets rules like "never show this ad more than 3 times to the same viewer in one day."

Think of it like loading a playlist before a concert — the songs are queued, ordered, and rules are set before the show starts.

Not every viewer watching the same channel is equally valuable to every advertiser. A car company might only want adults 25–54. A baby-food brand wants parents of young children. A local restaurant wants viewers in a specific city.

At the moment an ad break starts, the system checks: who is this viewer? Do they match the advertiser's criteria? Is it the right time of day? Has this viewer already seen this ad 3 times today?

All of that happens before a single frame of the ad plays.

This is the step that makes live TV advertising technically remarkable. The viewer is watching a live match. The broadcast reaches an ad break.

In the time it takes to blink — roughly 200–300 milliseconds — the system has to decide which ad this viewer should see, retrieve it, and stitch it seamlessly into the video stream so the viewer never notices a cut.

This is called server-side ad insertion: instead of the viewer's app fetching an ad separately (which can cause buffering), the streaming server blends the ad directly into the video before it reaches the viewer.

Once the ad is stitched in, it has to travel from a server to the viewer's TV, phone, or tablet. This is the delivery step.

The video is stored on a CDN — a Content Delivery Network — a global network of servers placed close to where viewers actually are. When someone in Calgary requests a video, it comes from a server in Calgary, not a data center in Virginia.

Without a CDN, every viewer in the world would pull from the same single server, which would immediately collapse.

This is the only step the viewer actually experiences. Everything before this — the buying, targeting, stitching, delivery — is invisible. The viewer just sees the ad play inside their show.

From the ad-tech system's perspective, this step is when measurement begins. The player tracks: did the ad start? Did the viewer watch 25%, 50%, 100%? Did they mute it? Did the ad take up the required share of the screen?

Once the ad has played and the beacons have fired, billing happens. The advertiser is charged for impressions that were verified as valid.

The money flows through the ecosystem: from the advertiser's DSP, through the exchange and SSP, to the publisher's account — minus the fees each intermediary takes along the way.

The streaming service earns revenue. The advertiser's campaign shows results. And the data from this campaign feeds back into the next campaign's targeting.