Where the results sit

The Trade Desk published a case study on Allwyn, the UK National Lottery operator. An agency ran the campaign with The Trade Desk. ThumbAd was not involved, and Allwyn is not a ThumbAd client. This is an operator's read of it. The case states no campaign period, so nothing here should be read as recent.

What was done

According to The Trade Desk's case, Allwyn has run the National Lottery since 2024; the campaign it describes, undated, ran on The Trade Desk's Kokai platform. The method, as described:

• A universal pixel on Allwyn's own website. One site-wide tag, letting the platform connect early-funnel exposure to end-of-funnel outcomes.

• Allwyn's first-party data plus Tesco retail data, combined to build the starting audiences.

• Connected TV (CTV), digital audio, online video and display.

• Cross-channel frequency control, meaning a limit on how often one person is reached across all channels together.

• Suppression of recent converters. People who had just converted were removed from the audience.

What was published

Every figure below is The Trade Desk's, quoted at the value its case study publishes. None is ThumbAd's.

• The Trade Desk's case reports a 64% reduction in time to conversion for Lotto when three channels were used instead of one.

• The Trade Desk's case reports a 30% lower cost per conversion when digital audio and display were combined, and 23% lower when CTV and display were combined.

• The Trade Desk's case reports 43% higher ROAS (return on ad spend) for audiences built on Tesco retail data than for standard prospecting.

• The Trade Desk's case reports a 30% lower CPA (cost per acquisition) for those retail-data audiences than for standard prospecting.

• The Trade Desk's case reports that cross-channel frequency controls freed 8% of the campaign budget, which was reinvested to reach 12 million more users.

What it does not prove

These are platform-reported outcomes: the platform that bought the media also measured what followed. The case names no control group and no holdout, so it shows the platform's own attribution of conversions to exposure: what was recorded, not what was caused. Whether those conversions would have happened anyway is an incrementality question, and that needs a controlled test, not a case study.

It is one advertiser's campaign, with no stated period or spend, and the case does not say what the two channel-pair cost figures were measured against. It shows what one campaign recorded, not what programmatic returns in general.

And lottery is not casino or sportsbook. The products differ in price, repeat play, player value, regulation and tax. None of the figures above can be carried into an operator's CPA, cost per first deposit or margin. The method question travels. The numbers do not.

What an operator should take from it

Look at where the results sit: each one somewhere a last-click report cannot see.

Channel combinations. Last-click credits the final click. People rarely click on CTV or audio, so a last-click report credits display or search and shows the rest as cost without return. The case reports its results for channel combinations: the unit last-click cannot measure.

Suppression. Money not spent on people who have already converted never appears in a conversion report. At most it shows up as a lower cost on the conversions that remain.

Frequency. Budget recovered from over-exposure comes back as reach, not as a click.

Underneath all three is the pixel on the advertiser's own site, connecting exposure to outcomes across every channel. It is what made the reported results measurable at all. Without it, the same campaign could have run and its results would have gone unrecorded, or been credited to whichever channel was clicked last.

That is the part any operator can copy, and the question it raises has nothing to do with lottery: is your own site and app signal connected to exposure across every channel you buy, so that if a combination of channels were working, or failing, you would see it?

For UK readers: the duty rise and 3 December

From 1 April 2026, Remote Gaming Duty, charged on operators' remote gaming profits, rose from 21% to 40%, as announced at the Autumn Budget 2025.

That leaves less margin for error in judging which players were worth acquiring. A player who looks profitable on last-click may not be: last-click hands the whole player to the final click, and the cost of everything before it sits elsewhere in the plan, unconnected to the player it helped bring in. At the old rate that error could hide inside the margin. At the new one it may not.

The first market-wide evidence has a date: the Gambling Commission's industry statistics for April to June 2026, the first quarter at the new rate, are scheduled on gov.uk for release on 3 December 2026 at 09:30. They will show what happened to the sector, not to any one operator's players. That answer exists only in an operator's own signal.

So the question to settle before that release is not whether programmatic works. It is this: if your players were costed today across every channel you buy, not just the last click, and valued after the new duty, would you still back the same players and the same channels?