Blog article in response to Digiday article on challenges of programmatic advertising
At the Digiday Programmatic Summit, top brands anonymously shared their biggest challenges with programmatic advertising. You can read the piece here. At Conversant, we discussed some of these issues with Elliott Clayton and Mayur Kshetramade who respectively lead our Agency and Direct to Brands businesses in the UK. Their responses to the questions and challenges can be found below.
IMAGE SOURCE: DigidayWe need to tie real-time spend to real-time insight, how do we do this?
[MK] The strategic concept here is the real-time optimisation of decisioning. In programmatic (or cross-device display) advertising, every customer makes themselves available to be spoken to 10/15 times a day, across thousands of advertisers and publishers out there. It is important to assess that and see who the person behind those cookies and devices is, what this person really means to the brand right now because of where they have been in the journey with that brand. This helps answer the question: should I advertise now or not? And if I advertise, what should I show them and why? What is the motivation for me to show something to them? That is real-time decision-making and the crux of what we do at Conversant.
[EC] To do this effectively, it is critical to activate the full CRM of that advertiser (full-site visits, customer transactions, email programme, direct mail programme, etc.). That is, it is critical that real-time programmatic advertising takes into account everything we know about that person as it relates to that brand.
[EC] One of the challenges in bringing this to life is that some advertisers’ databases are often not in one place, they are in several places, and often not joined up. Conversant solves for this challenge and creates a single customer view for the advertiser.
[MK] Finally, all advertisers should demand full accountability of their advertising investment. Conversant gives advertisers 24/7 access to an online portal. They can log in any time they want and see how many ads have been delivered against different groups of people and what kind of incremental sales we are generating for them.
If we understand the customer, have one view of them and one way to measure it, the last bit is the incrementality. How do we ascertain whether a conversion is truly incremental? Do algorithms need to change to bid for incrementality rather than just bid for the conversion or click?
[EC] Let’s address what incrementality based-bidding is. Whenever bid becomes available, you need to understand who the real person behind that bid is. How does that person index against all other audience for that brand at that time? Is the person potentially going to convert anyway?
[EC] To optimise for incremental sales, we need to understand how to measure the incrementality. And the most scientific way to do is not via attribution but by test vs. control measurement that measures the net impact of that specific advertising channel.
[MK] A lot of solutions claim to do a test vs. control. But there are varying levels of scientific integrity that’s used. And that causes a lot of bias and over-estimation of incrementality. In programmatic advertising, a few rules are a must to do this right:
Test and control need to be at an individual profile level, not cookies or devices. That means for any given individual, all their browsers, cookies, devices are either in test or control for that advertiser.
Assignment to test vs. control needs to be random, as you go, and persistent. What that means is anytime a user becomes available for the first time to be advertised to, they should be randomly assigned to test or control and that’s where they should stay forever. This allows for a true measurement of advertising at an individual level over lifetime – if advertising works, it should work over time and it should change behaviour. More people should engage with that brand more often.
The decision to show an ad, what to show, and what to pay – all should happen before the real-time bidder and server knows if that person is in control or test for that advertiser. That is why Conversant shows a charity ad for those in control and pays for those ads using our own money. This allows for true baseline measurement by looking at the conversion rate for the control group.
[EC] Incrementality is a great measurement because it puts the onus on the media supplier to deliver the right campaign – if the audience or media volume are wrong, it won’t work.
[EC] Measuring incrementality hasn’t properly caught up yet. For the past 15 years the entire industry has been peeling the onion finer and finer, trying to get closer to the conversion until you now have suppliers who are showing an ad on the client’s own site and claiming a conversion from that, which is clearly all wrong.
What if people want to bring programmatic advertising in-house? What are the challenges and risks?
[EC] You need to be involved with a DMP, matching partner, DSP, basically a lot of different systems. This means there are a lot of set-up costs, a lot of partners, and a lot of heavy-lifting. But even then, you lose scale as these tools are not joined up. The simple fact that one tool calls a customer 111 and the other refers to them as ABC can regularly lose audience. And without real-time incrementality based bidding; performance can be a problem too.
[MK] The underlying motivation to bring programmatic advertising in-house arises from two key factors – control and accountability. Given the overall low level of accountability in programmatic advertising, no doubt more advertisers want to have more control and visibility into their investments and programmes.
[MK] Conversant solves for control and accountability. 24/7 visibility into impressions delivered and against whom, on-going incrementality measurement, and match-back to real customer transactions as a feedback loop deliver control and accountability.
So that is one aspect of measuring: reaching out, identifying, but how about the issue of scaling it? There is sometimes the perception that even if we can scale it, it is a lot of work and we don't have the resources to do that – especially in the area of producing dynamic creative. How can we address this perception?
[EC] There’s a need to be able to build dynamic ads that cover brand, loyalty and direct response and which can be executed across operating systems. If you could create a process that could achieve that scale in real time and decide which of the strategies to employ, that would be fantastic. However, it should not be worth investing three thousand man-hours in, it should be done quickly with technology and in real time. This is true dynamic content.
[MK] Conversant has a vertically integrated tech stack that includes all functionalities of DMP, DSP, dynamic personalisation, incrementality-based real-time bidding and serving, and test vs. control measurement.
[EC] The Conversant personalisation engine builds creative iterations in real-time, at scale. It allows that creative to either appear as a direct response ad or a brand ad and it can build with the assets put together.
[MK] What you get is a personalised cadence of ads based on what we know about that person up until that time and what we want to achieve as the next business outcome. Every impression on purpose, with a purpose.
When we use Google’s bid manager platform, we’re able to see how the user converts. But once we set ads to devices, there’s no way of connecting these ads with acquisition. Even if we see conversion in the end, we cannot track that to a device. How do we get around this problem?
[MK] This is an issue that Google is addressing, but they are facing significant anti-trust issues that are making identifying an audience cross device; and then marketing to them problematic.
[EC] Yes - but the issue is already being addressed by the market through person-first media delivery. There are a number of suppliers looking at this – large siloed partners like Facebook, and then point solutions like Adbrain, Drawbridge and Tapad. We put ourselves in the deterministic-match pool but done in a privacy-consistent way and focused on keeping the internet open.
[MK] The best way to solve for this problem is to work with a partner that can activate a clients’ own first party data for tracking and delivery across devices– the output is that all numbers match clients’ own figures.
[EC] Conversant has the most robust model for matching a user and conversion across devices; we’ve solved for challenges and the gaps where conversion leaks occur and their cost to advertisers. Our work with a large fin-tech platform shows that Marketers are losing around 30% of the value they are generating by not resolving this issue.
Final observations on the programmatic advertising landscape?
[EC] An ongoing challenge is how we educate organisations on the power of programmatic advertising. What does programmatic advertising do? Right now it still has lots of untapped potential. It is mainly used as a cost-saving tool (buy as many impressions for as little cost as possible). If executed properly, it can become an incremental revenue-generating solution.
[MK] Yes, and it can do that by working across the entire funnel. It can account for loyalty marketing, promo, campaigns, retargeting. Programmatic advertising is more often used in acquisition but it’s in customer retention where it can be even more effective but is currently underutilised.
[EC] It is a game-changing tool if you use it properly.