In the US, several media agencies have recently put their faith in the same company, Placed, to offer a standard attribution model for gauging how mobile marketing influences purchases in stores. But can a standard for attribution really work, when brands’ marketing channels and customer experiences are so divergent and the combination of reasons for buying are unique to any given product?
“Attribution models had caught up until mobile device browsing and shopping exploded,” says Mobiles.co.uk digital marketing manager Jay Karsandas. “This changed how we view consumer behaviour and the models will need to evolve to incorporate this.”
Last click can still work extremely well, says Tom Lowes, head of online marketing at Sykes Cottages, however over time any marketer will inevitably want additional insight on the user journey. “For example, marketers are using software like Google Analytics to begin comparing channel performance under different attribution models, in order to understand the touchpoints in the journey and avoid having too much emphasis on last or first click,” he explains.
“There’s no advantage to adopting a standardised model, as doing so could be giving away a competitive advantage.”
Tom Lowes, Sykes Cottages
The customer journey is not just fragmented over channels either – it’s also across multiple devices. The speed with which both cross-channel and cross-device browsing have grown has “caught marketers out”, Lowes says, but the attribution technology is emerging to cope with that. “It’s not as though the technology isn’t there, it’s often the rate of adoption and technical difficulty of integration which delays implementation.”
Choosing the model to use can be a headache – the temptation being to revert to simpler models once again. These provide a hassle-free means of, for example, rewarding affiliates and identifying return on investment, but are they fair to the former and able to really help marketers maximise the latter?
There are so many paths people can take that allowing a single marketing channel to take all the credit for a sale doesn’t make sense anymore. Consumers are influenced over time by brand messaging – some more than others – as they move towards their purchase decisions. To ignore those previous touch points ignores the different role that each channel plays in the process.
A recent Forrester white paper summarises the current situation: “Customers control their relationships with brands. They know what they want, when they want it, and how — leaving companies with a fuzzy picture of what marketing efforts their customers are actually exposed to. This complex relationship makes it increasingly difficult [for brands] to understand their customers and identify the optimal purchase path.”
Jemma Jones, senior marketing manager at Honda Europe, agrees. “As marketers, we like to think – or at least hope – we are using our data effectively to make the right decisions about optimisation and allocation but in reality we are only seeing part of the total picture.”
Multichannel attribution models have therefore been receiving more attention. “For us, last click has had its day,” says Alix Satchell, online partnerships manager at House of Fraser. The retailer’s new model, developed with Rakuten Marketing, “looks at every interaction in the journey”. Recently, a multi-device aspect was added, which rewards affiliates more accurately.
“Without an understanding of our customers’ journeys across channels, we can’t reward valuable contributions from affiliates,” explains Satchell. “It’s really important that we understand the overall journey; it helps us to optimise our marketing activity and understand what influences our customers to buy.”
This is certainly an advance in affiliate attribution where there has long been a reliance on the last click model. Could standardisation work in the more advanced models that HoF is now using? “I don’t think there’s a ‘one model fits all’ rule,” says Satchell.
Others tend to agree. O2 innovation and capability lead Dan Michelson says: “Personally I don’t think [standardisation] is possible given there are so many variables, such as what data brands have access to, the channels used, relationships with agencies and so on. And that isn’t even accounting for business differences like products and KPIs.”
“Given that each brand is distinctive and takes a different approach to its marketing, there’s no advantage to adopting a standardised model, as doing so could be giving away a competitive advantage,” adds Sykes Cottages’ Lowes. “It would be very difficult for a CMO to successfully argue to the board that the brand has a marketing advantage over the competition when they’re both using essentially the same attribution model.”
Multi-touchpoint attribution models are far from perfect – and there is plenty more room for innovation. Mobile is the game changer for many, including Jones at Honda, who says: “Despite the rapid growth of mobile and tablet consumption, mobile ad spend is still lagging behind due to uncertainty about mobile attribution and how to accurately measure mobile campaigns.”
She adds: “Mobile bridges an important gap between online and offline channels so an accurate understanding of mobile interactions and their relationship with other touchpoints would provide a very important, currently missing, part of the jigsaw.”
O2’s Michelson expands on the point. “Marketing and the understanding of what works – and what doesn’t – has become incredibly complex. Attribution suppliers have come a long way, but we’re now all faced with a huge challenge of how to include mobile properly within attribution models.
“This includes how we deal with multiple screens [mobile, tablet and desktop, and even smart TVs], as well as how we calculate the value of organic social when it includes elements like sharable content, given the [buzz around] ‘content creation’ right now,” he adds.
Indeed, there is a long way to go in terms of data consolidation. Advertising platforms such as Google AdWords and Facebook are attempting to track consumers across platforms and devices, and do so to a certain degree, but these platforms are disconnected and information is not therefore synced, says Honda’s Jones.
Research shows that 63% of new car buyers have a brand in mind before they start searching but that only 20% of these consumers actually buy the first vehicle they research. “There is a huge window of opportunity to disrupt and lure the consumer into your brand during the shopping phase,” says Honda’s Jones. “Advanced attribution models will allow us to identify and assign a value to those consumers who have interacted with our brand and automatically apply this to future buying decisions in real time.”
Jones feels the models have “come a long way in a short space of time” and that modellers will increasingly tap into the opportunity offered by data to understand the role of every touchpoint and the relationship between those touch points, to deliver “not only an improved understanding of the value of each one, but most importantly how this information can be used for optimisation and future budget allocation”.
We currently use two forms of attribution models. A time-decay attribution model [where interactions closer to the sale are given more credit] is used to gain insight on our customer journeys and paths to purchase, using a half-life of seven to 14 days depending on seasonality and campaigns. This also drives our decisions on where to spend our marketing budget and the types of activity to progress. On the other hand we use a ‘last non-direct click’ model when determining payments to our partners and affiliates to keep a fair balance without data discrepancies.
We have three different attribution methods running concurrently: ‘first click’ internal attribution tracking, incorporating CRM data; ‘last three clicks’ within Marin Software, weighted mostly towards ‘last click’; and Adometry, a machine learning-based attribution technology that has no fixed click-weighting model.
We focus on cross channel convergence and convergence attribution. This essentially means identifying the direct and indirect relationships between channels within the consumer journey and their value in achieving a business outcome.
Until recently we used multiple methods including ‘last click wins’ to attribute online sales, and more traditional forms of measurement to track offline sales. Working with Aquila Insight, we have recently implemented an advanced decay attribution model, which covers both online and offline customer interactions.
Q: What are the most common forms of attribution models you use?
In the past, all of our digital performance was assessed on a last click or impression basis. This provided us with a decent view of performance, but we struggled to understand how social, display or mobile helped deliver on our key performance indicators (KPIs), as they appeared to underperform in both models. This just felt wrong.
Q: Have you changed your approach then?
Yes, we’ve been working with Visual IQ because we wanted a more appropriate way of assessing digital media so that we could make the right optimisation decisions based on which media were influencing the sale. We realised conversions weren’t based solely on one touchpoint, but on multiple touchpoints, and that we were potentially undervaluing the impact of our marketing efforts beyond direct response, such as brand communications.
Q: Does every brand need a multi-touchpoint attribution model?
Yes, if they want to assess the true impact of their media, especially considering the multitude of different consumer journeys. It’s no longer just about finding the channel that converts, but also what influences discovery and how to get your brand on a consumer’s consideration list. Prior to embarking on an attribution initiative, however, it is essential that you properly scope why you want attribution. If you aren’t able to act on the insights and recommendation derived from attribution, there’s a real chance you won’t get anything out of it.
Q: What are the benefits and disadvantages of multi-touchpoint attribution models?
One of the most basic, but most important benefits is that it allows you to understand which digital channels actually deliver against your KPIs. The disadvantage is it takes time and planning to get the most out of attribution, and it is also dependent on good data. If you have multiple data sources, you may struggle. We were fortunate in that we weren’t confronted with as many data collection issues as we could have been. We only had to collect data from the log files from our ad server, which included the majority of our paid and organic media. Without good planning and good data, you won’t get the most out of your attribution investment.