Marketers are so busy trying to understand the customer's path to conversion, but have they missed a step in actually allowing themselves to be converted? AdNews partnered with AdRoll to find out.
This may sound like a riddle but it's a reality: countless brands and marketers are still focused on last click attribution.
Giving the last click all the credit for the sale or conversion is old hat and fails to reflect the sophisticated and complex marketing landscape. It’s incredibly common, but nonetheless is widely believed to be an outdated methodology.
It also goes without saying that marketers are juggling a l ot, with the onslaught of ad tech stacks, digital transformation, marketing clouds, viewability and adblocking. Wanting things to be simpler and more streamlined is probably high on the wish-list.
Mixed model and multi-touch attribution are alternatives, but it’s not simple. Not only does it require a change in practices and potentially staffing, but more importantly, it requires a huge change in thinking.
Hot on the heels of the AdNews roundtable with marketers, in partnership with AdRoll, we next turned to media agencies for their perspective.
A total of eight agency professionals joined us for a high level conversation about the must knows of attribution.
“Attribution is a really difficult topic,” AdRoll MD Ben Sharp said. “We’ve all got much more complex marketing programs that we run for our clients and we've noticed when speaking to marketers directly that they’re, to a degree, at a loss in terms of how best to put in place the right attribution model for their businesses.
“There’s lots of different technology solutions, but there’s also the strategy piece, like how do you plan for attribution in the first place which is something that is quite difficult.”
As a result of another hearty roundtable debate, 10 key lessons emerged as pointers to help marketers break free of the last click chains.
1. Get your data house in order
Cadreon's director of data and technology James Wood said getting your data house in order is “critical for attribution modelling” as the whole concept is about modelling how a series of events contributed to an outcome.
“If marketers are not tracking all these events then the model will be limited or, in the worst case, flawed,” he said. “Additionally, a robust attribution model needs to look outside paid media and incorporate data from any touchpoint that can influence action.”
He added this is where a data management platform steps in and helps ingest and organise data from both paid advertising and broader marketing initiatives (e.g. CRM) to provide a single customer view for holistic attribution modelling.
2. Data hygiene
Having data is one thing, but clean, usable data is another. Data hygiene, that being the process of ensuring data is accurate, complete and representative, is half the battle when it comes to advancing in attribution, explained GroupM's director of tech operations, Timothy Whitfield.
He said there are several data sources that need to be matched together - campaign management; ad server data and website traffic data etc.
“The problems are often simple like matching client and campaign names, however they get more complicated like media booking types and now even more complicated with programmatic as each 'trader' uses his/her own naming conventions for the various media strategies,” Whitfield said.
He added there is often “ too much data” and the client simply needs to see a clear report. Ikon's chief knowledge officer Rob Pardini agreed good data hygiene is one of the key challenges when advanced attribution solutions are implemented.
“In particular, blind spots within data sets can undermine multi-touch attribution models given these assign some of the credit for an acquisition or lead to every touch-point along the customer’s online journey,” he said.
“Good data hygiene can be onerous when implementing advanced attribution solutions as there are many factors that can distort the relevant data sets. For example, Safari blocks third-party cookies by default, cross-device behaviour may be challenging to match to the same customer, and PCs at home often have multiple users.”
3. Adjust your expectations
Attribution is a tough nut to crack, there’s no denying that. But the stigma that follows makes it difficult for marketers to embrace it. OMD's head of interactive Dan Robins said there’s a stigma around the complexity of getting attribution modelling right and that it will be impossible to find insights that can create value.
“While it is definitely not easy, and does require extended commitment, by breaking up the challenge, brands can start to dig out the insights which, as a collective, will drive performance improvements. Overall, it’s about taking a test, learn and re-assess approach as there is no set-and-forget with attribution,” Robins said.
Columbus Digital's national display director Jamie Moran said the common stigmas or
barriers are “will attribution drive incremental ROI from their advertising campaigns” and sometimes “external” factors (such as climate, economy, sales, offers) cannot be taken into account within an attribution model.
“The best head space would be to understand that attribution isn’t a s ilver bullet to understanding how each channel performs, but a tool and methodology that will continue to adapt and improve to hopefully help in delivering insight into the true impact of our marketing efforts,” Moran said.
Whitfield added he would like to see a more simplistic method introduced, such as an FMCG model, a retail model and a finance model emerge that works well for the market.
“This would be an easy way for clients to buy into the costs and bank the dividends,” he said.
ZenithOptimedia's chief digital officer Geraldine McDonagh added: “Australia is also quite risk adverse as a culture, especially within media” which can also add to the barriers of executing new models and new ways of thinking.
4. Beware of variable costs
When it comes to the type of costs marketers look at for this, Robins said costs are much like many areas of analytics, in that you need to pay for the attribution technology, though in many cases this is built into other techs such as your web analytics or ad server packages, coupled with resourcing analysts to prise out the insights for media buyers.
“Pricing can be as little as a few thousand dollars for a one-off piece of analysis, running up to the hundreds of thousands per year for a point solution incorporating tech, visualisation and analysis,” Robins said.
“The key for brands is to decide where on the spectrum of commitment to the testing they sit before committing funds as the analysis will only provide value if suitably implemented into media planning.”
But Ikon has seen clients that have dived in and rolled out multi-touch attribution solutions experience an overall cost per acquisition fall by at least 15% within the first year following implementation.
Pardini said the uptake of advanced attribution solutions is being held back because these solutions are complex and costly to implement and it’s not always apparent that a positive business return will follow.
“Across the clients we’ve worked with to roll out multi-touch attribution solutions, our experience is that the overall cost per acquisition typically falls by at least 15% within the first year following implementation,” he explained.
“For organisations investing heavily in digital marketing, this level of efficiency gain is compelling, but light spenders generally won’t see sufficient gains to justify the cost.”
5. Revise your structure
Introducing and implementing an attribution strategy can change the structure of a business, impact day-to-day norms and alter the way some channels work together. Going from there, the prerequisite training and education will also need to be different.
“This is a specific skill set with its own different language,” Whitfield explained.
He said people talk about “rim weighted, linear regression, normalised models” and a number of other mathematical models that don’t often make sense to clients.
“This is a skill that they don’t teach you at university,” he added.
ZenithOptimedia's McDonagh agreed attribution is challenging for clients to navigate as it almost requires them to shake up their structures to make sure they’ve got the right expertise on their side to actually tackle it. She added that to get it across the line it's not just the CMO that requires buy-in, but all relevant stakeholders from the client side are needed to be able to push it through.
The panel noted that more sophisticated attribution needs the right people, with deep expertise, to be able to integrate that and a talent crisis in this area only adds to the challenge.
Carat's head of data and analytics Kaine Hickson said for clients with performance objectives, as the use of attribution modelling becomes more extensive, it alters the existing structure of the media planning team within an agency.
He said it requires a different skill set to= analyse, build attribution models and optimise the media investment, utilising the abundance of data agencies currently have available.
The more clients that demand agencies to work with them deeper on attribution means agencies need to hire more candidates in this field, as well as upskill current planning staff to have a better understanding of attribution modelling and its findings.
“An attribution strategy will involve adding a much greater level of analytical thinking into the business structure to manage it. Building the right capabilities is vital as it’ll drive the strength of the model as well as its ongoing degree of success. For the majority of clients, it will require much less investment within marketing to allow their media agencies to manage it, as they have the scale across many clients to invest in the right people as well as ensure they’re trained in the latest developments in this evolving area,” Hickson said.
6. Let go of the fear
It was also discussed that many marketers tend to have ‘the fear', which is why attribution models that aren't last click are often put in the too hard basket. Key points raised were on why marketers should lose the fear and that agencies should help with this transition.
AdRoll head of marketing, JAPAC, Cat Prestipino, said one of the things that came up when speaking to the marketers in the first attribution roundtable was the uncertainty around introducing more complexity in the new attribution model.
“There was a fear that they would be upheld to unrealistic new targets or new quotas or new expectations around ROI. There might be some fear that changing the status quo could potentially be really damaging,” she said. Pardini echoed Prestipino’s comments, adding that costs were also part of the fear factor.
7. Have patience
Marketers are used to getting what levers they want pulled and expect a consistent ROI within a set time frame, and on budget. However, attribution, multi-touch or varying mixed models touted need patience and as Whitfield said, it shouldn't be about why they need to do it, but more about how they can do it. They need to be ready to wait and see when exploring diverse models.
Bohemia’s data and attribution director Stephen Wright said by their very nature attribution models alter the user environment through proposed increases or decreases in budgets across channels, sites or creative messaging. For this reason, he said it’s vital that marketers have patience and know exactly what it is they are optimising and measuring in order to gain value from attribution.
“Models are generally built retrospectively and so hypotheses need time to be tested and validated in live communications to see improvements in business metrics,” he said.
8. Take it easy
With a good budget at hand to advance within attribution, and an array of tech partners at the helm, agency execs believe there is a real danger of marketers running before they walk – with all advising it's vital for marketers to start experimenting with one or two models.
Wright said the major attribution technologies build models to single business outcomes, and for good reason.
“There is a danger that modelling against too many outcomes could muddy the waters when it comes to making decisions if there is no understanding of the interplay between business metrics,” he said.
“Narrowing the focus to a single outcome ensures that any change from implementing the recommended optimisations are not negatively impacted by other models and a true evaluation of business impact can be measured.”
9. The role of agencies
Agencies need to demystify multi-touch attribution and ensure it's passed on coherently from the marketer to other points of contact.
Something that also helps the subject matter advance is if agencies don't “over-sell” attribution.
“The education piece is massive,” Whitfield said. “We also need to start talking about this and making some headway with models and do somemyth busting.
“The industry could do with an expert in this field. A new type of company that consults on this and understands how it can be implemented and make it work. It’s currently just as specialised as a creative agency.”
Despite agencies believing they have a vital role in helping advance the attribution discussion, some external attribution specialists believe agencies should not do attribution as well as media planning as it's deemed a conflict of interest or like “grading your own homework”.
AdRoll’s Sharp added that the education piece from agencies back to clients is critical to make sure the same conversations surrounding attribution aren't happening in 12 months’ time.
“It really sounds like there’s an opportunity for those in the agency to really drive the ongoing understanding of how to put an attribution model in place with clients,” Sharp said.
10. The future of attribution
Cadreon's Wood said going forward the dependence on cookies will fade and we'll see more modelling against persistent, peoplebased identifiers - allowing complete visibility on the path to purchase.
“Additionally, as data accessibility increases marketers will be able to fuse new data sets into the model including traditional offline channels and audience behavioural data,” he said. “The inclusion of audience data will enable the models to optimise audiences as well as channels, for example, creating a segment of consumers with a high propensity to buy and predicting the next best action a marketer can take to drive the required action.”
“Of course, there’s also still the cross device challenge obviously. Facebook is trying to make that a lot easier, but how do we integrate offline data into online data to have a holistic view on attribution? ” McDonagh added.
“That’s obviously the holy grail, but we still have many challenges around that.”
Click here to download the AdRoll Agencies and Attribution discussion paper.
Have something to say on this? Share your views in the comments section below. Or if you have a news story or tip-off, drop us a line at adnews@yaffa.com.au
Sign up to the AdNews newsletter, like us on Facebook or follow us on Twitter for breaking stories and campaigns throughout the day.