3 ms·

If you've ever worked with FB ads, you know that you can directly attribute sales ($ hitting your account) to ad groups ($ leaving your account) You bid to hit

by jdross 7y ago

If you've ever worked with FB ads, you know that you can directly attribute sales ($ hitting your account) to ad groups ($ leaving your account) You bid to hit a certain CAC (customer acquisition cost) and to the extent any numbers are inflated, it would just quickly result in a lower bid price per inflated number

mochomocha 7y ago
... Which isn't a causal estimate of ad effect at all, as most conversions would happen anyway regardless of whether or not the ad was seen. Incremental conversions unfortunately are usually 1 or 2 order of magnitude smaller than raw number of correlationally-attributed conversions. Which makes most incrementality-based ROI numbers of advertising campaigns be negative.
- adamsb6 7y ago
  I don't follow. How would the conversion get measured if the user never saw the ad? They'd need to follow a link to a landing page that is associated with that particular campaign, and that registers the conversion after the user performs the desired action, right?
  - mochomocha 7y ago
    You can run "ghost ads": pretend that you did bid and won an ad auction, log this information, and measure what would have happened in terms of conversions for this counterfactual set of users. Both Google and FB have experimental products to do that, as well as a couple of third party companies. It's very difficult to set up properly, and the ROI numbers don't look good at all which is why big platforms are not super reluctant to offer it broadly, even though it would be a better and more honest way of measuring advertising effectiveness.
    - adamsb6 7y ago
      Ah, I think I understand. So, you'd measure whether this person made a purchase without ever having viewed the ad. Thanks for explaining. I can see why platforms would be reluctant to offer this feature. If you're running ads on multiple platforms and only one offers this feature, it will make itself look uniquely bad. However, the other platforms probably also provide little extra value compared to running ads on a single platform.

by jdross 7y ago

mochomocha 7y ago

... Which isn't a causal estimate of ad effect at all, as most conversions would happen anyway regardless of whether or not the ad was seen. Incremental conversions unfortunately are usually 1 or 2 order of magnitude smaller than raw number of correlationally-attributed conversions. Which makes most incrementality-based ROI numbers of advertising campaigns be negative.

adamsb6 7y ago
I don't follow. How would the conversion get measured if the user never saw the ad? They'd need to follow a link to a landing page that is associated with that particular campaign, and that registers the conversion after the user performs the desired action, right?
- mochomocha 7y ago
  You can run "ghost ads": pretend that you did bid and won an ad auction, log this information, and measure what would have happened in terms of conversions for this counterfactual set of users. Both Google and FB have experimental products to do that, as well as a couple of third party companies. It's very difficult to set up properly, and the ROI numbers don't look good at all which is why big platforms are not super reluctant to offer it broadly, even though it would be a better and more honest way of measuring advertising effectiveness.
  - adamsb6 7y ago
    Ah, I think I understand. So, you'd measure whether this person made a purchase without ever having viewed the ad. Thanks for explaining. I can see why platforms would be reluctant to offer this feature. If you're running ads on multiple platforms and only one offers this feature, it will make itself look uniquely bad. However, the other platforms probably also provide little extra value compared to running ads on a single platform.