Exclusive: IAG, Columbus debate digital advertising measurement pitfalls and perils
- 05 December, 2017 07:41
Last-click attribution remains the bane of the digital marketing industry, many in the industry agree. Yet widespread use as a result of technology, along with growing applications of artificial intelligence (AI) in programmatic advertising, could actually see brands beholden to this misguided approach for even longer.
That’s the view of IAG director of media and technology, Willem Paling, co-author on the new IAG Digital Ad Effectiveness Playbook, being released in Australia this week. Paling talked to CMO exclusively along with IAB’s director of research, Gai Le Roy, and co-chair of the council and national head of strategy and products at Columbus Agency, Marcus Betschel, to mark the paper’s launch.
The playbook is the latest offering from the IAB’s Ad Effectiveness Council, which launched in March and presented its first glossary for digital advertising in June. It covers the basics of digital advertising measurement as well as attribution models, improving data quality, marketing mix modelling and brand strategy. It’s been created off the back of insights and debate from across the Ad Effectiveness Council’s 16 representatives: Columbus Agency, eBay, IAG, Experian, Facebook, Google, Yahoo7, Nielsen, Impact Radius, REA Group, ComScore, Kantar MillwardBrown, Nine, Fairfax Media, News Corp and Roy Morgan Research.
“There is a lot of bad practice out there – this is about trying to overcome that,” Le Roy said. “This playbook shares details from people living and breathing this stuff every day. People do things quite differently in different organisations. Having 16 people around a table looking at each other to ask if there’s a best way of doing things has been a healthy exercise.”
The fixation on last-click
One of the most troublesome metrics still in use is last-click attribution. Paling, who co-penned the attribution section of the paper with eBay, said one problem is marketers and agencies are still using multiple tools optimised to last interaction.
“It’s still a common currency when you compare between tools. It’s a real problem, because to do away with it, you introduce a whole lot of complexity,” he said. “Marketers think they’re making a meaningful comparison. The problem is it’s not meaningful, you’re comparing the same ‘lie’.”
A growing concern is DSPs using AI to further automate ad delivery against an outcome based on last-click attribution, Paling said.
“They’re optimising towards a goal that none of us really subscribe to anymore,” he said. “It’s a massive thing that needs to be fixed in the industry. That’s the point where people seem to be applying AI right now.”
What Paling would like to see is DSPs employing AI to better illustrate how they’re picking up ad fraud and optimising towards effectiveness in terms of attention. More widely, Paling also identified an opportunity to use AI to help read people’s expressions as they’re exposed to advertising to understand emotional responses and impact.
In the meantime, Paling advised advertisers to recognise which media messaging is contributing to their desired outcome, and which isn’t. That requires a better understanding of the distinction between attribution and effectiveness well before you employ a particular attribution approach, he said.
“We talk a lot about the assumptions that go into a means of measurement. Generally with attribution, is there is a short time period between exposure to an ad and the sales outcome,” Paling explained. “Yet if you go back to traditional marketing, it’s about building mental availability, brand awareness, and positive emotions towards the brand. That’s effectiveness.
“There are components of delivering effectiveness where attribution is the right thing. And there is a type of marketing you can deliver to the sales outcome. But there’s lots of marketing where it’s not. Effectiveness then comes back to principles, human knowledge and thinking about the opportunity for attention.”
Yet having become used to last-click attribution, it’s a real challenge for marketers to flip those results on their head, Paling admitted.
“It’s uncomfortable to have to say I’m now measuring properly and the results are worse than before,” he said. “It’s tempting to keep reporting these great results and not switch to actually adding value.
“We are lucky at IAG to have the support of our chief customer officer to look at this problem specifically, to find out what’s been going wrong in the last three years and change it if something’s wrong.”
Wherever the team can, IAG uses ‘experimental design’ as an attribution method. This is about taking two or more similar groups, creating different advertising conditions for each, and then measuring the different outcomes. These groups may be identified as groups of cookies, people, or geographic segments.
“If we’re able to measure sales impact, we need to create the dataset that shows us if we did or didn’t do the marketing then look at the sales difference between the two,” Paling said. “For Facebook, or marketing to customers, for example, we can create a treatment and control and look across these groups. In search, the time to outcome is short. So you can turn it on and off for a day, and see the sales change, so you get a stronger read of contribution of search beyond last interaction.”
At IAG, budgets are also split into activities with a known sales result, and those without. Paling said search is about the only channel that fits the first bill.
“For everything else, we’ve done the experiment to see the spend can’t be justified on such a short-term sales outcome,” he said. Paling also noted ongoing research by Peter Field and Les Binet on the importance of investing in long-term, brand building advertising activities. “In doing that, we create the space to invest in the brand work. That’s an ongoing process.”
Painting the wider picture
For Betschel, a lot of Australian companies have been “lazy” around metrics and measurement. He also expressed concern about relying on default metrics, such as clickthrough rates or cost per click, to identify effectiveness.
“No one metric in itself will give you the answer to the success of the business or if you’re spending in the best way,” he said. “It’s about understanding all that’s available, challenging marketing teams, agencies and people you’re working with to get a holistic picture of your advertising and what metrics are telling you about that activity.”
In addition, businesses using old-school metrics like last-click attribution are often only focused on short-term interactions, painting a very misleading picture of success.
“That leads to problems with their longer-term brand health. And that affects sales health as well,” Betschel said. “I’ve seen budgets going straight to performance channels, and that limits the spend and activity on driving brand awareness. That leads to long-term decline and makes it harder to perform. You can’t overweigh in one area.”
So where do you start? From the beginning, Betschel said. “What are you trying to achieve with digital spend, and what are the key drivers?” he asked.
Le Roy called on more marketers to embrace new ways of measuring their digital activity.
“Marketers shouldn’t be scared of getting rid of metrics if they’re not right. If this playbook stops 10 companies doing last-click attribution, we’ll have made ground,” she said.
Paling, Betschel and Le Roy also discussed the rise of people-based marketing and its impact on the way digital advertising is understood. Betschel saw strong opportunity in tracking people metrics to understand an individual’s interaction with a range of different messaging, and how they behave differently with each.
Being able to gain a more wholesome view of a person, their purchasing behaviour and other behavioural elements then allows marketers to create more meaningful ways of understanding different types or groups of people.
“That will change the way we advertise to people in the next decade I think,” Betschel said.
But again, Paling warned against falling into the trap of last-interaction attribution as people-based marketing becomes the norm.
“Because you’re able to tie all the devices together, people-based marketing has the potential to make last-interaction attribution a lot worse,” he claimed. “If an outcome happens on one device, you’re able to look across multiple devices to ask if you served them anything. While we all agree that’s broken to start with, it just further overstates the impact.
“Where you actually know who people are across multiple devices, and are able to implement a more robust attribution method, such as experimental design, you’re able to say here’s a group of people who did and didn’t see the advertising and what difference it made. That’d different to rating the devices that saw advertising or didn’t.”
Le Roy pointed out connected TV has seen consumers become more comfortable with logging into different devices, opening the door to people-based marketing techniques. Yet she agreed people-based marketing is not an easy path to travel.
“The advent of clients taking more control in-house of their data sets – like IAG, CommBank – helps them have control and oversight on the whole picture,” she added.