Shoppable CTV in Demand Gen: Set Up the Shopping Experience and Measure It

What needs to be in place

Google announced shoppable CTV as generally available in its January 2026 Demand Gen update. Its current QR-code guidance describes the setup: link Merchant Center, attach the product feed to the campaign and include TV screens in device targeting.

Google generates the QR code; do not bake one into the video. For Demand Gen skippable in-stream ads, the code appears when the skippable period begins, alongside a call to action and send-to-phone option. A TV that cannot support the required player resizing may not show the shopping experience.

The exclusions are specific. Google's guidance excludes YouTube TV and Google TV inventory from QR-code availability, and it does not support product feeds with QR codes on Shorts. Including TV screens therefore does not promise a product panel on every television impression.

For the feed itself, Google's product-feed guidance says to use images at least 500 × 500 pixels to serve product feeds on Connected TV. Check the products selected for the campaign as well as the image files. A valid feed full of the wrong collection is still the wrong shopping experience.

Choose what the viewer should want to buy

Consider a furniture retailer advertising a compact sofa. The video demonstrates how it fits into a small apartment; the product selection should help someone investigate that sofa, its relevant sizes or genuinely useful alternatives. A panel of unrelated dining tables would break the connection between the story and the next action.

Watch the creative at the distance a person would sit from a television. Can you recognize the product? Does the opening establish why it matters? Can you understand the offer without reading a paragraph of small text? These are design checks, not a requirement to replace every existing video.

Review the assets with and without sound, and inspect the available previews rather than assuming a vertical social edit will make equally good use of every placement. The aim is a clear proposition, not a television aesthetic imposed on an unsuitable product.

Follow through to the mobile product page. In the sofa example, the destination should make dimensions, delivery, price and the selected option easy to understand. A slow page, a missing size or an unexpected delivery charge is a specific problem to fix. It does not establish that every TV impression was wasted, but it gives you a better next action than changing bids blindly.

Trace the purchase before interpreting the report

Make a controlled test of the destination and measurement using your normal testing process. Confirm that the intended product opens, the right variant can be selected and a completed test order reaches the expected purchase event with the correct value. Do not assume a working browser page proves that conversion measurement is working too.

Then inspect what the campaign is optimizing toward. An add-to-cart event and a completed purchase are not interchangeable, even when both appear under a broad conversion total. If the account mixes them, isolate the purchase outcome before calculating revenue per advertising dollar.

Keep reporting windows and conversion delay visible. Someone may act later or on another device; the report reflects the measurement and attribution available to the account, not a complete diary of every viewer. Missing credit is possible, but it is not a license to assign unobserved purchases to TV.

Use a device breakdown for the TV-screen question. The API identifies that device as CONNECTED_TV. Do not confuse it with a YouTube channel total: a channel and the device on which its ads were seen answer different questions.

A TV row needs a cost, not just a share of clicks

Suppose a retailer's mature reporting period shows $3,000 of TV-screen spend and $9,000 of attributed purchase value. Mobile shows $5,000 of spend and $10,000 of purchase value. The reported ROAS is 3× for TV and 2× for mobile. These are invented figures, not a benchmark.

That comparison is more useful than saying TV received 9% of clicks or 14% of conversion value. A share without its cost cannot tell you whether the investment met the business's needs. Even the ROAS comparison needs comparable purchase values and enough time for returns or cancellations that materially affect the result.

It still does not prove TV is the more incremental investment. Attribution allocates credit; a controlled test asks what changed because of the advertising. Nor does 3× historical ROAS mean the next dollar on TV will return $3.

If you can run an eligible, well-designed experiment that isolates TV inclusion, use the combined business outcome to judge it. Otherwise, a before-and-after review can be informative, but account for promotions, inventory, other creative changes and the altered mix of available impressions.

Decide what the evidence warrants. That may be better product selection, a clearer video, a repaired destination, more observation time or a device-targeting change. Do not add an assumed pool of untracked sales just to rescue a disappointing result.