This guest post is written by Rich Jones, the Head of Product at Dataseat, a mobile-first DSP that was recently acquired by the Verve Group. I previously hosted Rich on the MDM podcast as part of the ATT, One Month In series.
Most in-app marketers will now be quite familiar with the main challenges posed by Apple’s App Tracking Transparency (ATT) privacy policy – and similar forthcoming privacy changes on Android – effectively spelling the end of both behavioral-based targeting and unrestricted measurement – requiring marketers to pivot campaigns toward new privacy-preserving methods.
Although most advertisers have made good progress navigating this change, one key driver of success has not been highlighted: in the post-ATT world, spending more with fewer advertising channels is better. This is for two reasons:
1: You are competing with yourself for the same impressions
Contextual targeting has significantly changed bidding dynamics. This has been observed via a surprising increase in average CPMs since the rollout of ATT. With more platforms using only contextual signals to determine the attractiveness of an impression, a much higher share of available impressions are attractive to buyers now.
- In the pre-ATT market, DSPs and ad networks were more likely to seek specific device IDs in the bid requests, allowing them to ignore far more of the available impressions. The logic behind running with multiple ad networks and DSPs at this time was that they each had their own unique database of users to target.
- Post-ATT, the net result is that a far higher share of bid requests will now result in bids, and therefore, more campaigns are competing for each impression than before.
So a higher bidding rate means you are competing more with others, but it also means for any single impression, your own campaigns may bid against each other via different DSPs and ad networks. In the case of in-app bidding / unified auctions (which continues to gain share-of-voice vs. traditional waterfall bidding): this competition can occur within the same auction.
Of course, this will not occur every time, but it’s very likely to happen when two ad networks are running the same campaigns and are targeting the same inventory – e.g. the same publisher apps. Since the crossover in supply across networks is very high, there are lots of opportunities for this to happen.
The more channels across which your budget is deployed, the more you are unnecessarily driving up your own CPMs. Buying on a CPI does not change this dynamic, as a higher CPM naturally leads to a higher CPI.
This dynamic also risks exposing users to far more ads than the advertiser intends: since frequency caps only exist at the per-network level, and the more networks are showing ads for the same campaign, the more you risk exceeding those limits.
The solution? Pick your preferred channel and stick to it. This has already happened in more mature programmatic verticals such as desktop branding, where The Trade Desk tends to have one disproportionately large slice of the pie vs. other networks.
But it’s not just performance that is affected here: measurement can also be worse – and is likely to get worse still – if you are running media with many different networks.
2: On iOS: the more networks you use, the less conversion data you get
Although most app UA campaigns are still using some form of MMP fingerprinting for install tracking, it’s generally expected that SKAdNetwork will be the only viable method on iOS in the mid-to-long term, with Apple likely to pursue methods that block probabilistic attribution.
Those already using SKAdNetwork extensively will be aware of the challenges posed by crowd anonymity / ‘privacy threshold’ restrictions, which often prevent marketers from receiving crucial conversion data, such as information about the source app that delivers the install.
But these restrictions are significantly alleviated if the same level of spend is directed to a smaller number of networks and campaigns: the fewer networks you spend with, the more conversions will be scored above the privacy threshold, and the more SKAN conversion data will be returned.
This is because crowd anonymity is based on the volume of installs generated by each network and campaign, over a rolling time period (approximately daily) – with higher volumes of installs constituting more privacy. And the more you spread your total budget for a given app over many campaigns and networks, the fewer daily installs you will get on any one campaign and network.
Very large spenders are naturally less affected by this issue and are thus better positioned to spend across many channels with less penalty. But even the largest advertisers will find that SKAdNetwork’s crowd anonymity restricts the visibility of some conversions.
This will become even more critical with the changes promised in SKAdNetwork 4.0, where the core campaign ID data itself (now represented by a hierarchical “Source Identifier”) will be subject to two levels of privacy constraints. Advertisers wanting to get the most out of the SKAdNetwork 4.0 features will be more restricted by crowd anonymity than ever before.
The solution here is to spend as much as possible under a single DSP or network, while also having a SKAdNetwork Source App ID strategy that will maximize visibility above the privacy threshold based on various characteristics of the campaign.
It may seem counterintuitive to suggest that right at this moment – when they are facing more challenges than ever before – that advertisers can benefit by doing less. But considering that in-app bidding is on the rise and here to stay, and privacy constraints are poised to make measurement harder, it’s likely that many can benefit by shifting their DSP and ad network strategy. This is especially true with networks offering full transparency, control, and customization options that can help marketers get the most out of contextual targeting and the new measurement constraints.
Photo by Museums Victoria on Unsplash