4 ms·

ah but how do you really know if any of those numbers are causal? marketers need to account for multi-touch attribution, if a customer sees an ad on one platfor

by huac 7y ago

ah but how do you really know if any of those numbers are causal? marketers need to account for multi-touch attribution, if a customer sees an ad on one platform and converts on another, or if they do it ten minutes outside of your conversion window. there certainly are a lot of numbers in adtech and marketing. whether or not those actually relate to causal impact, across the industry, is not definitively answered. very few players (on the supply and demand side) have the scale and ability to properly do incrementality testing. personally, i think people don't want to know the real results. if the numbers say it's not broken, why risk your job?

bluGill 7y ago
Those are real concerns that I have no insight on if they are accounted for. I know that they can be accounted for and have been. TV and newspaper ads have been doing it for years where for obvious reasons they can't track where or when you saw the ad. McDonald's runs ads on those platforms anyway.
- huac 7y ago
  yeah you can try, for print/OTT the most common way to estimate lift is with DMA segmentation, compare lift in the DMA's which saw the ad versus the DMA's which didn't. but how do you pick the right DMA's to pair, etc. it's hard and i think most people don't do it right; again, it requires very large n to do correctly, which requires very large budgets. mcdonalds is actually one of the savvier players here, they know that it's hard to quantify brand advertising if you only go off in-store purchases, so they spent very heavily to promote their app and develop a consumer-brand relationship there. now, if you see a mcdonalds ad on youtube, they can (try to) match your youtube ad impression to a purchase done through the app / credit card / other PII.
summerlight 7y ago
> very few players (on the supply and demand side) have the scale and ability to properly do incrementality testing. https://support.google.com/google-ads/answer/6394265 https://support.google.com/google-ads/answer/6394265 Actually Google is preparing it as a product, so I would guess the result was generally positive? FB also has a similar product as well.
- huac 7y ago
  google and facebook are not "most players" -- they own walled gardens for measurement, it's much harder to do measurement on open web auction placements. the google page you linked also says, "you won't be able to continue using this model if your data drops below 10,000 clicks on Google Search or below 400 conversions for the conversion action within 30 days." depends on what your market/CTA is, but a lot of SMB likely don't hit these thresholds. similarly with FB, my understanding is that you have to go through an account rep to set it up and that the minimum spend is fairly high to qualify. i think that these measurements are possible to do with a reasonable level of certainty, but they require a particular set of circumstances to pull off.
- sundance_X 6y ago
  Incrementality testing in an additive model makes YouTube spend show stupendous ROI (very cheap, but targeted CPMs)for consumer electronics, at least.

by huac 7y ago

bluGill 7y ago

Those are real concerns that I have no insight on if they are accounted for. I know that they can be accounted for and have been. TV and newspaper ads have been doing it for years where for obvious reasons they can't track where or when you saw the ad. McDonald's runs ads on those platforms anyway.

huac 7y ago
yeah you can try, for print/OTT the most common way to estimate lift is with DMA segmentation, compare lift in the DMA's which saw the ad versus the DMA's which didn't. but how do you pick the right DMA's to pair, etc. it's hard and i think most people don't do it right; again, it requires very large n to do correctly, which requires very large budgets. mcdonalds is actually one of the savvier players here, they know that it's hard to quantify brand advertising if you only go off in-store purchases, so they spent very heavily to promote their app and develop a consumer-brand relationship there. now, if you see a mcdonalds ad on youtube, they can (try to) match your youtube ad impression to a purchase done through the app / credit card / other PII.

summerlight 7y ago

> very few players (on the supply and demand side) have the scale and ability to properly do incrementality testing. https://support.google.com/google-ads/answer/6394265 https://support.google.com/google-ads/answer/6394265 Actually Google is preparing it as a product, so I would guess the result was generally positive? FB also has a similar product as well.

huac 7y ago
google and facebook are not "most players" -- they own walled gardens for measurement, it's much harder to do measurement on open web auction placements. the google page you linked also says, "you won't be able to continue using this model if your data drops below 10,000 clicks on Google Search or below 400 conversions for the conversion action within 30 days." depends on what your market/CTA is, but a lot of SMB likely don't hit these thresholds. similarly with FB, my understanding is that you have to go through an account rep to set it up and that the minimum spend is fairly high to qualify. i think that these measurements are possible to do with a reasonable level of certainty, but they require a particular set of circumstances to pull off.
sundance_X 6y ago
Incrementality testing in an additive model makes YouTube spend show stupendous ROI (very cheap, but targeted CPMs)for consumer electronics, at least.