Programmatic supply chain transparency study: reflections one month on
Sam Tomlinson, the author of a bombshell ISBA and PwC study exposing 'unattributable' supply chain costs, shares what he has learned since the report was first published
It's now just over one month since the ISBA programmatic supply chain transparency study (with AOP, by PwC) was published. This study uncovered important issues in the digital advertising supply chain, notably issues with data access and data quality, and some spend being unattributable.
In the one month since publication, we have presented the study and key findings 60+ times, to an estimated 180 different companies, across EMEA, North America, and Asia-Pacific. There's been a huge amount of press coverage, numerous webinars and podcasts, some lively debates on LinkedIn and Twitter, interest from government and regulators, and even an internet meme from someone with plenty of spare time on their hands.
As the dust settles, and the industry taskforce assembles to take the study recommendations forward, what are our reflections one month on since publication?
1. All study participants deserve great credit for taking part
More than 50 organisations participated in this study. Without them, these important industry-wide issues might have remained uncovered for many more years. All of them - the trade bodies, the publishers, the agencies, the tech vendors - deserve great credit for taking part, most particularly the advertisers that funded the study. These organisations, and their leaders, demonstrated courage and commitment in stepping forward. Everyone who will benefit owes them thanks.
2. Coverage has ranged from clickbait to cerebral
Inevitably, there were some lurid headlines in response to the study, sometimes by people who clearly hadn't read the study or press release. But much of the coverage - both when the study was published, and in the weeks since - has been really interesting, well-informed and thoughtful, challenging everyone in the industry to reflect on how programmatic advertising should evolve. Examples include these two recent pieces in Mediatel News and Campaign.
3. LinkedIn is growing in importance for informed business commentary
LinkedIn is now an important source for immediate and informed business commentary. In response to our study, it included thoughtful blogs, insightful agency responses, quick links to videos such as the typically forthright ID Comms Media Snack, and supportive comments (thank you!) from some high-profile industry figures e.g. Jerry Daykin, and Steven Pollack and Rachel Mervis.
If you're looking for LOTS OF CAPITALS, follow the POTUS on Twitter. If you're looking for informed business commentary, turn to LinkedIn.
4. Our study is not representative of all programmatic (nor was it ever designed to be)
We matched 31 million impressions end-to-end, just 12% of the impressions served from our study advertisers via the study tech vendors to the study publishers in the UK in January-March 2020. In the world of programmatic, 31 million is a drop in the ocean. As a result, we've sometimes been challenged on how we can know that our study is "representative". So, to be clear, we have never claimed that our study is representative of all programmatic.
Nor was it ever intended to be. This was a premium study to start with: prestigious advertisers operating disclosed programmatic models; the premium publishers who received approximately 20% of those advertisers' impressions; and the best-known tech vendors. Within that premium world, we could only match the 12% of impressions with the best data quality. Our study therefore represents the "super-premium" end of programmatic - a "best-of-the-best" scenario.
As analysts and auditors, we have no data from this study to indicate what's happening in long tail programmatic. But personally, as a statistician, I'd be happy to place a bet - and put it this way, I'd be surprised if more money reached publishers in the long tail.
(And, of course, 31 million in absolute terms is a big number. We can be very confident that our study accurately conveys the findings from those 31 million matched impressions.)
5. The real story is the challenges with data access and data quality
Much of the coverage and commentary has focused on the "unknown delta" (the 15% of unattributable spend), but this misses the more fundamental story, which is the myriad issues with data access and data quality. From when the advertisers, agencies and publishers signed up in June 2019, it took nine months to gather the data from the tech vendors.
Many of these delays were related to the complex permissioning and re-permissioning required to grant legal access to the data. Each individual component of the supply chain was behaving rationally, but the result is systemic irrationality - it should take a few days, not nine months, for advertisers and publishers to access and share their own data.
Compounding these challenges, the data quality was variable and challenging: different data definitions and taxonomies; missing data fields; some data only available at aggregate level; multiple date formats alone; and of course no transaction ID to match an impression from buy-side to sell-side. Without industry-wide protocols around data access and data quality, the unknown delta will remain exactly that: unknown.
6. "Working media" is an unhelpful term
"Working media" was used only three times in the main body of the public report, and always with quote marks to indicate that it is widely-recognised industry shorthand, not a term we endorse. It's a horrible phrase, because it implies agencies, DSPs and SSPs are not working, which of course they are. We don't yet have a better alternative, but once there is one, we would happily see the term "working media" consigned to the dustbin (along with its even more unhelpful cousin, the "tech tax").
7. Agency fees are not where advertisers should focus
Our study was never designed to be an exercise in benchmarking agency fees - you don't need a 15-month end-to-end supply chain industry study to do that. Astute brands will recognise that the complexity of programmatic advertising merits paying higher fees to secure the best quality of agency talent on their account.
The opportunity in programmatic is to execute sophisticated well-planned campaigns targeting the right audience in the right context at the right time, while minimising risks of spend leakage via the unknown delta. Those objectives are unlikely to be achieved by driving down agency fees.
8. Think carefully about where you would like your ads to appear
20% of impressions in our study universe were served to premium publishers, but conversely 14% were served to 'long tail' publishers. Clearly, small websites can still be publishing high-quality niche content, and those publishers need an avenue to monetise their ad inventory.
But the sheer volume of UK websites used in our three-month study period was nevertheless surprisingly high. Prestigious brands are understandably careful to restrict use of their logos and other intellectual property. Logically, the same care would extend to the curation of websites on which their brands and logos appear.
9. Whatever is driving the unknown delta, it isn't classic ad fraud
The unknown delta (of 15% unattributable spend) represents a discrepancy between DSP and SSP data at the level of a single impression or group of impressions. Numerous commentators quickly labelled the unknown delta as ad fraud. But our study waterfall analysed monetised matched impressions i.e. those where we could see an impression (or identifiable group of impressions) on both buy-side and sell-side.
If advertiser spend had been diverted to a fraudulent website, there wouldn't have been a sell-side impression with our premium study publishers to match with. Whatever is driving the unknown delta in our study waterfall, it isn't classic website ad fraud. For the same reason, it's also unlikely to be driven by discrepancies in impression counting.
(As an aside, all our experience in programmatic has indicated to us that the ecosystem is messy, not Machiavellian - it's simply grown very fast in the past decade, and governance is still catching up.
My personal guess is that the delta is the aggregate effect of multiple causes, each individually contributing only one or two or three percent, which aggregates to the 15%. The only way to know for sure is to first resolve the data access and data quality issues described earlier.)
10. Programmatic is already successful, with the potential to be even better
The promise of programmatic is the ability to target the right audiences, in the right context, at the right time. That promise remains hugely valuable to marketers across many sectors and products, and there is no question that programmatic is here to stay. All of us involved in the study are pro programmatic, provided it's done well.
As a brand or agency, if programmatic is already working well for you, then the 15% delta should be seen as good news: it means an already-successful channel could potentially be 15% even better.
Next steps: The taskforce will require collaboration and international support to succeed
In response to the study an industry taskforce has been convened comprising ISBA, IPA, IAB, AOP and JICWEBS, aiming to drive standardised protocols around data access and data quality. For this to succeed, it will require collaboration from all involved in the UK, together with international assistance led by the highly engaged US advertiser trade body ANA and the World Federation of Advertisers. The early signs are promising. There is a key role for everyone in ensuring this initial momentum is maintained.
As has been widely reported, this study cost a huge amount to deliver, many times what we were paid. But it has also been the single most interesting project most of us have been involved in. One month on, we remain optimistic that this study will be a positive catalyst for industry change.
Sam Tomlinson is Partner and Leader of PwC Marketing & Media Assurance