Attributing a sale to a particular piece of marketing has long been a headache for brands – one that increases proportionally with the growing number of marketing channels and technology platforms they use to reach their audience. But current models of attribution are so unfit for purpose that at the Advertising Research Foundation’s Re:Think conference in New York last month, Facebook vice-president of global marketing solutions Carolyn Everson proclaimed that marketing needs “new measurement for a new era”, in order to cope with consumers’ behaviour across all kinds of connected devices.
On this side of the Atlantic, News UK, publisher of the Sun, The Times and The Sunday Times, has commissioned research to develop new metrics to measure the success of advertising across different online and offline platforms.
Yet as marketers seek to move beyond outmoded last-click attribution, where a sale is credited to the last website from which the buyer clicked through to the vendor, the picture of what a one-size-fits-all attribution model might look like remains hazy.
One driving force for change is the divisive nature of last-click data. Simon Jarratt, digital acquisition and media manager at Center Parcs, says: “In the past we spent money on channels based on last-click data, our own booking patterns and experience, but it was clear that this was not the best use of budget, as cost per acquisition [CPA] of a customer continued to climb.”
He acknowledges that last-click attribution may work for some businesses where the journey to purchase is short, but is clear that it is not useful for a considered purchase, such as a holiday. “We have thousands of bookings where the consumer has been exposed to over 10 types of media before a booking was made. All had influence on the booking that was finally placed.” Center Parcs now uses this detailed attribution data as the primary measurement of the effectiveness of its digital marketing, working with technology platform IgnitionOne.
Others are also moving away from the last-click model. Argos worked with Summit Media to set up its latest attribution model to work across different marketing channels as well as within each one. Digital marketing controller Matt Roberts says: “Within paid search we are able to see which terms initiated and contributed to a sale, as well as the term that finally converted. That means there can be more value associated with search terms used by a buyer early in the process, which in last-click attribution wouldn’t get credited with anything, and so we can adjust our media buying accordingly.”
TUI Travel, which owns Thomson and First Choice, has invested in an attribution model that covers all of its online channels, as well as some of its offline activity, allowing it to move away from a model that saw all channels and steps towards the purchase being given the same weighting.
“We had all sorts of claims on the same booking and we were probably spending too much money paying out twice,” says web analytics manager Brian Bunyan. At the heart of its new model is ‘tag’ management, which involves using snippets of code that are triggered by certain actions. For example, if a customer clicks on a display ad on a third party page, a display tag lets TUI know which action had been completed and where.
“With last-click you can have two or three people trying to claim credit for a booking. A tag management solution allows you to identify who owns that booking and that is the person who gets credit for it.” TUI has increased its overall digital budget since using tag management through provider TagMan, including increasing its search optimisation spend six-fold, since it can measure the true contribution of natural search.
House of Fraser is also an advocate of understanding the customer journey across channels. It is working with Rakuten Marketing and DC Storm to bring together its online channels into a measurement hub. One benefit of improved attribution has been the ability to better measure and analyse the effectiveness of its display ads.
“We are also using attributed sales data to power the commercial terms of our affiliate programme,” says affiliate manager Margaret Herrera. “This new payment model not only gives us the confidence that we’re paying affiliates a fair price for their contribution, but it also encourages them to contribute further up the sales funnel and be rewarded for it.”
The uncertainties of the future are adding to the confusion. With the number of channels and platforms growing at pace and fundamentally changing the customer journey, it is difficult for brands to know in which areas of attribution to invest.
Center Parcs’ Jarratt says: “When we were first looking at technologies, the touch points that guests used to research and book were usually personal and work computers. In a relatively short period of time, users have become accustomed to utilising a range of different devices when researching and booking breaks, so it has become nearly impossible to tie this all up. As a result, it becomes difficult to understand what channels help most with
the different stages of the purchase funnel.”
Some echo Jarratt’s view that last click still has a part to play in shorter purchasing decisions. “In impulse campaigns, last click is still relevant,” says Simon Birkenhead, global advertising sales director at Telefónica.
He gives the example of a SMS text campaign for Unilever’s Cornetto brand, prompting consumers to buy from nearby vendors. He acknowledges that previous marketing messages will have created a brand perception of Cornetto, “but the trigger for that purchasing decision was the fact that you received a message with an offer that was contextually relevant”.
Birkenhead insists that taking a mobile-first approach to attribution will become more prevalent. “If you solve attribution for mobile, you are solving it for the future; if you are solving it for the desktop, it has limited shelf life because in five years time people will not be using desktops.”
Telefónica is placing particular emphasis on tracking the journey from mobile use to retail stores through Weve, its joint venture with Vodafone and EE, which has announced two deals to connect mobile marketing with the in-store experience. The first deal was with MasterCard, where purchase information from MasterCard is tracked back to offers that Weve has sent to customers. The second deal is a beta test at Eat stores for its Pouch app, which manages people’s loyalty cards from a single mobile phone. The loyalty cards are automatically identified by Bluetooth technology in store, enabling the retailer to match a customer’s purchases with previous marketing or promotional activity.
House of Fraser is also making strides in connecting online and in-store activity, thanks to its store card. “It makes it more straightforward to connect our offline sales with the online user journey,” says Herrera. “This is a bonus in helping us to understand the online/offline relationship.”
But for brands not offering loyalty or payment cards, joining the dots between online and wider offline activity can be trickier, and while a number of possible solutions have been tried, such as ‘contactless’ NFC chips and QR codes on outdoor and print advertising, a fool-proof solution has not been found, and brands need to avoid falling into the trap of losing sight of the consumer in their quest for an answer. As Birkenhead says, many of the techniques being tried are designed to help the media owner – not the consumer.
“Advertisers have to realise that if they are asking consumers to do something, such as scan a bar code, they need to provide a value proposition to the consumer to do that. They need to think about why someone would scan it first. Too many brands think the user is going to be interested in their product, but actually they’re not, and that is why they get poor results.”
Jarratt says Center Parcs recently worked with its media agency Walker Media to use econometrics – applying maths, stats and computer science to economic data – to understand the relationship between its offline spend and bookings. “While it is difficult to attribute each pound we spend directly to a specific booking, the combination of tools has given us a good understanding of which channels work at which stage of the journey, and we have distributed spends accordingly in order to reduce CPA and maximise returns.”
It seems that unless there is a direct call to action in a print or outdoor ad, prompting someone to visit a specific web address or call a phone number for example, attributing a later sale to that activity is nigh on impossible at the moment. As Roberts at Argos says, the ability to accurately connect online and offline behaviour is becoming increasingly important for brands, and he sees light at the end of the tunnel. “I think that the growing number of significant data owners – companies like Google, Facebook and retailers with large volumes of customer data – and the ability to measure physical interactions in store with digital data, will be able to unlock this increasingly complex customer picture.”
Case study: Phones 4u
Phones 4u uses a number of marketing channels to generate sales and traffic on its website, including search, social media, online display ads and retargeting, as well as referrals from websites in its ‘partner programme’. Before implementing advanced attribution, reporting the sources of its traffic was a manual and time-consuming process that required collecting performance data from a number of places.
The company was also using last-click data, but knew it was not allocating its budget efficiently.
Phones4u wanted a solution that would give accurate metrics and reduce the cost of acquiring a customer (CPA), as well as allowing it to integrate its own demographic, location and cost data into the measurement model. “Our focus is on spending more efficiently,” says Adam Cartlidge, head of online marketing. It now works with attribution software company Visual IQ. “Attributed metrics enable us to easily identify which channels, campaigns, ad groups, creatives and keywords are being over- or under-valued compared to the last click and where we can better allocate our spend to maximise performance.
“For instance, based on the recommendations from our attribution software, we increased our display spend by 86 per cent and decreased our paid search spend by 53 per cent month-over-month. We were also able to review the return-on-investment performance of all of our partner organisations, which served as the catalyst for stopping some of our partnership activity and reinvesting elsewhere. As a result of the recommendations, we are seeing a double-digit percentage decrease in CPA.”
Positive intentions to move beyond ‘last-click’ abound, but analytical challenges remain.
Companies have more data at their disposal, which can be used to understand the complexities of human interaction. Making sense of this data and, crucially, choosing which to analyse and which to discard, has always been a challenge for marketers without heavy-duty analyst resource.
While it is likely that most businesses running multichannel marketing campaigns will be working with attribution, many companies default to the last-click model. The beauty has been its simplicity and transparency, which has helped to fuel the experimental growth of performance marketing by providing a clear business case for heavy investment in channels such as search engine marketing, comparison shopping engines, affiliate marketing and direct response display advertising.
With the rapid rate of change within digital marketing, this reliance on last-click attribution is not keeping up. Investing in channels at the top of the sales funnel is integral for the success of those at the bottom, and marketers have almost unanimously agreed that relying on the last-click model exclusively is flawed.
But marketers are often stumped by the complexity of attribution models that are more complex than last-click. As a result, uptake of attribution modelling has not been as swift as predicted.
Reasons for the delayed uptake include the analytical challenges faced in setting up a solution and a lack of willingness to move away from the last-click model, which could reduce the budgets and rewards for teams. Multiple datasets, conversion types, teams and often partner companies all need to come together before a model is produced, and even then I’ve heard marketers express distrust in the final model.
The modern marketer needs to take into account the customer journey over multiple devices and channels, the time of day and year, the influence of other consumers and a myriad of other factors. Accounting for the offline part of the journey is further from being fully solved by technology.
Econsultancy research has shown that 65 per cent of consumers have checked a product in store before buying online, which presents the question of where in the journey the attribution model should begin analysing. The majority of marketers only apportion credit to the online part of the journey.
Most marketers, however, are still trying to overcome the challenges associated with attribution. As data gets exponentially ‘bigger’ and more complex, attribution models will need to be agile within businesses, adapting to the continual flux of the customer journey.
Director of Create
Telegraph Media Group
Advancement of data helps answer attribution questions, but it will not answer them alone. Data is powerful if it is used the right way, but can be misleading if it is not.
People assume offline is less accountable than online but this is not strictly true. Data telling us millions of people have seen a banner can sometimes be just as misleading as circulation informing us that millions have seen a print ad.
There is a place for last-click attribution but at the bottom of the sales funnel. However, it is only one of a host of attributions we need to assess. It is also important to remember the rest of the AIDA sales model: attention, interest and desire, before you get to action. All clients want to prove that their marketing activity helped them to sell stuff but that doesn’t come from last-click alone.
We need to be creative with attribution.