‘We have to test or you’re dead next year’: Overheard at the Digiday Programmatic Marketing Summit

Participants were asked ahead of the discussion to share some of the challenges they’re facing at work for Digiday’s challenge board, which was used to help spark conversation within the group. The bolded headers below are a few of the challenges shared by attendees and the quotes that follow are highlights from the resulting discussions. 

The quotes have been lightly edited and condensed for clarity. 

From the challenge board: ‘Google deeming traffic from DSPs as invalid’ 

“We don’t really know what is happening when our ad servers are bringing back a good 20 [to] 30 [to] 40% of impressions, and even 90% of clicks, as invalid. We have a couple of… theories as to why. It’s usually around Google touching the media at some point to validate the different media that’s coming through, but in instances where it’s not touching it at some point, that’s really where we’re seeing the invalid go through. When that happens, though, we have trouble seeing what is actually valid versus what isn’t and it’s usually what Google’s seeing and their bias.”

“It’s [happening at] different points of where Google can touch the media, whether it’s on the supply-side, the demand-side, the exchange itself; we don’t even get that level of granularity. And so those biases see if Google was touching the media, but like I said, it’s definitely just a guess on our behalf because all we can do is guess that Google is doing this as a self-serving type of ploy because they want to be able to touch more of the media and take more of the media dollars.”

“We’re seeing it mostly with invalid clicks. That’s kind of the biggest problem. It happens with impressions too, but we have these great CTRs in the DSPs, and then we get the DCM report and it’s a really bad click-through rate.” [Editor’s note, basically the DSPs are saying the click-through rate is good, but the ad server is saying otherwise.]

“Basically, we’ve just been building these really extensive blocklists to combat it and finding which sites that DCM seems to not like for some reason, which we can’t really figure out what it is, but they seem like good sites [but then the] clicks aren’t coming as valid but the traffic is. So we’ve just been building a lot of blocklists and trying to get it better, but it’s hard to optimize in the DSP or trust the DSP’s documentation because we use DCM DoubleClick as our source of truth.”

When asked by Digiday if they’ve raised this issue with Google or the DSPs, the participant said: 

“Google hasn’t given us great [insight] and honestly, neither really have [the DSPs]. Both blame each other.” 

“I haven’t reached out to the sites [that have been put on the blocklists as a result of the invalidation] because we were mostly running on the open exchange. We switched to PMPs a little bit to try and combat this.” 

From the challenge board: ‘Attribution in the cookieless environment’ 

Participant: “In b2b, when we’re talking about enterprise scale, account-based marketing, things of that nature, user graphs that have both deterministic and probabilistic [models] are building on top of IP [addresses] and other solutions that don’t need or don’t rely necessarily on cookies … And we’ve been using the solutions versus a cookie-based approach [for years].” 

Digiday: “Not to jump two-three years ahead, but are you worried about the IP address being taken off the table? That’s personal information in California.” 

Participant: Yeah. But for right now it’s not. 

“I would actually love a monopoly. I think that’d be great. Just one [alternative identifier]. And I think it should be Google since they took it away. They gotta give us our ball back.” 

“I think the next big wave everybody’s going to see is IP bridging. A lot of companies now, they’re sticking their graph into a cleanroom or some other shared space, and you can take your set of first-party users, and then choose a partner – or multiple partners, but I don’t really suggest that – and then you can bridge those IDs so that you can effectively get more of the marketplace than you would have just going through a single provider … it is essentially a cleanroom that’s not for activation. It’s literally just for connecting more IDs to your ID.” 

“I have to ask the question, how are we making a better value exchange with consumers? Because so much of the conversation here is [framed around] ‘how do we as an industry,’ [but not] ‘how are we talking to consumers?’ How are we making sure that they actually understand the value exchange of offering up an email, signing up for a text message, logging into a website, downloading an app? It all seems very pushed on people right now.” 

“[In consumer surveys] we actually discovered there was this huge thing about consent and when you looked at Gen Z versus looking at the older consumers, the older consumers have this defeatist mentality of like, ‘[the companies] will get [your data] anyway’ [but] Gen Z was like, ‘no [and] this is going to hurt you and I’m going to run with it.’” 

“From the agency side, we’re getting a lot more pressure a year-and-a-half, two years ago, on figuring out a cookieless solution. Now, it’s kind of us pushing it a little bit more in the background and that has affected the resources, the teams [and] the time we’re putting into figuring this out.”

“Isn’t part of the problem that no brands are good at first-party data?” 

“Giant brands never had to deal with [first-party data collection, while] medium-sized and small brands don’t have the budgets to have heavy investments in data cleanrooms and CDPs … but big brands and small brands are faced with the same challenge: They’re now having to do something they’ve never had to do before.” 

For one participant’s client, they “basically just stripped all cookies, all the tags, off of everything on programmatic. We were like, “OK, let’s just put those impressions out there. Welcome to 2002. Just throw the money at the wall.” 

From the challenge board: ‘Client hesitancy for new formats’ 

“If you have a CMO who is super aggressive and interested in testing, getting really entrepreneurial [but] has a board of directors that … [is] just not interested in tech fanatics … the tenure is short.” 

“You can’t say you want to be innovative and we have to hit our KPIs based upon historicals. It just kind of goes nowhere.” 

“How do you tell the client, ‘We have to test or you’re dead next year?’ Maybe that’s just how we have to say it.” 

“Even when you’re talking about new media [channels, clients say] that TikTok would be amazing, but they’re just so worried about privacy issues and where the data is going, that [TikTok] was a non-starter.” 

“One issue that we’ve also had is not necessarily new formats, [but] we’ll recommend something to a new client and they’re like, ‘No, that hasn’t worked for us in the past’ because it was done poorly by a previous agency.” 

“For our clients, we set aside budget tranches and say, ‘this is your testing budget, but you don’t really have a choice,’ … because the platform changes, the demographics change, something that didn’t work nine months ago can certainly work [in the future].” 

“Our biggest challenge sometimes is creative costs.” 

From the challenge board: ‘What are some novel uses of AI right now? Not old machine learning.‘

Participant: “The ability to leverage AI as a tool to create iterative content based on something that works that we can leverage on another platform, we’re trying to figure out if there’s a capability of that in the long term. Because for us, those creative costs are just really expensive. Some of our clients, sometimes they keep us from doing what they need to do.”

Digiday: “Does that become a slippery slope, though, where if it’s a video campaign, [you say], ‘Let’s have Runway ML generate something that we can use and we’ll test it.’ [You] test it, it performs well, then the client’s like, ‘Wait, Runway ML is only 15 bucks a month? What do I need my agency for?”

Participant: “All of us that have used AI probably have realized that it usually gets you like 70 to 80% of the way there and never gets you all the way to the line. So if I can [help] my creative team [have] less work to do because the AI portion is getting them that much closer to the final product, in the end that maybe helps us and helps our clients feel … because they’re going to have more ability to test, more ability to optimize, more ability to win. It’s a double-edged sword.” 

“AI will not steal anyone’s job. But someone who knows how to use AI effectively will seal your job.” 

“You still need marketing prowess to optimize the AI.”

“We are using AI especially on the copy side, because there’s nothing worse than having to copyright from a blank page. It takes too long. AI is able to get us a head start on that process.”

“One of the things I hope for AI is that we actually use web crawlers in a good way for change, and utilize variables that are understandable from the perspective of like, what can we tell is a made-for-advertising site, or a bad news site or something like that.” 

“We just came out of annual planning with the goal of how do we productize, streamline, become more competitive with pricing, with the use of AI. So we’ve got our billing team using ChatGPT to streamline invoicing and they’ve reduced invoice times by I think seven to 10 days, which just turns more cash flows for the agency.”

“[AI] saves me a lot of time [on] RFP responses. We’re starting to use Beautiful AI, which is like a deck builder. I know we all probably hate spending all of our time building decks and that has turned an eight hour project into a two hour project where I’m just really providing the value and not searching for images and graphs. AI can do all of that for me.”