CTV Attribution Headaches – How Are You All Measuring True Incrementality?

[AccomplishedTart9015]

we don’t try to “fix” ctv attribution inside the platform. for clients we treat view-through as directional only, then validate impact outside it.

we start with a simple geo holdout (on vs off markets) and judge on total conversions and revenue over a few weeks, not attributed conversions. once there’s enough spend history and variation, we layer in mmm to quantify ctv’s contribution over time and sanity check the test results.

[ppcbetter_says]

I would bet that your views/impressions are 90%+ fake/bots so there is no incremental revenue.

I see advertisers using on screen QR codes with accompanying call to action to try to drive users to a tracking page. I’d love to see the traffic data on those pages, it’s probably close to zero.

The only thing I’d recommend is adding a bot detection tool into the creative on the server side. That could help you confirm or deny my claim that 90% of your traffic is fake.

[NiceStraightMan]

Geo holdouts are your best bet. Pick similar markets then run CTV equally and measure total conversions. Way more reliable than platform attribution nonsense.

[ppcwithyrv]

At that point its view through conversions—> similar to TTD or Google Display (prog). Keep in mind their midfunnel…..never first or last click attribution. Its tough to show true conversion on CTV.

At that point its DDA—-> data driven attribution where the weight of the conversion is decided by Google. Not first/last.

Side Note: CTV is CPM driven as 95%+ of the inventory is a full video completion.

[InevitableImpress850]

Been there. We ditched platform attribution entirely and went full MMM after 6 months of data. Takes patience but gives you actual incrementality numbers leadership can trust. Geo tests work shortterm.

[oreynolds29]

What’s worked for us is treating CTV like upper-funnel by default, then proving directional lift. We ran clean geo tests and paired them with post-exposure trends. Some managed marketplaces (think Vibe-style setups) make testing easier, but you still need discipline in design.

[BillelKarkariy]

For CTV, the only credible answer is an incrementality test. If you have budget, do a geo holdout. If not, at least shorten VTC windows and compare exposed vs non‑exposed cohorts over 7–14 days. Modeled conversions without a test will always get questioned.

[Electronic_Branch901]

I would suggest running geo holdout experiments (aka “incrementality tests”). But please be cautious about things like Synthetic Controls, MDE, Confidence Intervals. Because if done incorrectly you might get very misleading insights.

If you are asking for specific vendors, happy to suggest a couple of credible ones

[Available_Cup5454]

Run geo based holdouts with fixed impression caps and measure post exposure conversion deltas outside the view through window only

[MySEMStrategist]

We use attribution tools like triple whale or north beam. We also use lift studies for our video ads in Google.

[kubrador]

tbh if you’re seeing lift but can’t prove it, leadership doesn’t actually want proof. they want permission to keep the budget. run a holdout test anyway and watch them suddenly find reasons why it’s “not representative of your audience.”