Commentary
Targeting Options Divide Advertisers
- by Laurie Sullivan , Staff Writer @lauriesullivan, November 5, 2018
If you don’t believe marketers are still divided and confused by their options to use cookies or people-based measurement techniques to measure data, think again. A recent white paper measures the differences.
The data suggests cookies tend to overestimate reach, and underestimate frequency and conversions because of the way they technically operate.
More marketers today now build their media-buying plan around reach, but with a cookie-based approach, marketers underestimate reach.
Victor Wong, CEO at Thunder Experience Cloud, which helps marketers reach consumers through data targeting, said marketers actually gain half the reach. “If a marketer tries to reach 20 million people, we estimate the reach turns out to be around 10 million.”
The data revealed that a cookie-based approach tends to overestimate unique reach by 101%, underestimate the actual frequency by 50%, and find 18% fewer ad-attributed conversions.
advertisement
advertisement
Historically, most ad servers like Google DoubleClick and Sizmek measured (or still do) cookies, where they count how many they see. The problem is that people delete cookies from the browser and use multiple devices.
The study also found when applying an identity graph, or a unique identifier, to the data set it will match multiple cookies from multiple devices to one person because the log-in links to the same ID.
When marketers stay with the cookie-based approach, rather than move to a people-based one, they are missing the opportunity to reach more consumers.
The inability to associate all the correct sales to having seen an ad remains the problem. For example, if a consumer viewed an ad on a desktop and purchased it on mobile, the advertiser will not know that it’s the same person. It happens to me all the time. I may view something on a retailer’s website via my laptop and then later go back and purchase the item on a mobile device. If I’m logged in the retailer knows it’s me, but if I’m not I keep seeing the same ad for what seems like eternity.
People-based measurement counts ad exposure and conversions using a person ID that combines multiple cookies and device IDs rather than using a single cookie or device ID, explains Wong. He said this doesn't solve the problem of tracking in Apple Safari -- where the advertising identifier (IDFA) is used for each iOS device that mobile ad networks typically use to serve targeted ads -- but it does help.
An identity graph would allow marketers to attribute that conversion to an ad and measure the true performance of a campaign.
The study also found a 5% difference in unique reach measured for every 100 million impressions. The data also found bigger discrepancies occurring the longer the campaign runs.
Laurie Sullivan is a writer and editor for MediaPost. You can reach Laurie at lauriesullivan@gmail.com.
Hi Laurie.
I agree that there is a big issue with cookie-based measurement versus people-based measurement.
But I think you have a few things upside down.
Wong says "If a marketer tries to reach 20 million people, we estimate the reach turns out to be around 10 million.”, and "a cookie-based approach tends to overestimate unique reach by 101%, underestimate the actual frequency by 50%".
You correctly state that "The problem is that people delete cookies from the browser and use multiple devices." as being one of the core reasons for Unique Audience (UA/Reach) being 'inflated'.
But your second and third paragraphs state:
"The data suggests cookies tend to underestimate reach, and underestimate frequency and conversions because of the way they technically operate.
More marketers today now build their media-buying plan around reach, but with a cookie-based approach, marketers underestimate reach."
Is this not contrary to the TEC report and Wong's comments?
Cookie-based measurement is good for measuring total traffic but not audiences. It will always overtstate reach and understate frequency (cookies, multiple locations, multiple devices). This means that marketers are overstating whet they think their reach will be and end up buying more frequency.
As an example, a decade or so ago in Australia, our Unique Browser (UB - the lead metric back then) exceeded our population. It was dismissed as a glitch and the underlying measurement issues were put aside. That is until we got decent broadband and closer to universal interent access and the UBs rose to 5x our population. It took some hard talking but UBs were 'retired' and Unique Audience (UA) became the currency metric,
John, a related issue is how good are cookie measurements as a way to "control" one's frequency via "frequency capping"? Answer: not very good. Indeed, when an advertiser caps the frequency of its campaign at , say, three---to avoid excessive redundency and offending users, the actual frequency may be only 1.5 and, if we knew how many times the ad was really seen, as opposed to being "on screen", the figure might be .5.
Spot on Ed.
But the quirk is that as cookie-based measurement tends to over-state reach (due to not de-duplicating by platform, device, location or longitudinally over time) it tends to under-state the frequency. For example if I see the same ad on my phone and my laptop I appear to the website analytics cookie-counter as two people seeing the ad once instead of one person seeing the ad twice.
Sadly, this has the double-whammy effect of appearing to reach more people but with a lower frequency than actually occurs, which of course is also an incentive to spend more to get to the 'effective frequency' level that,the client requires.