Why O2 Scrapped Last Click Attribution

Customers are increasingly using a combination of online and offline channels as well as multiple mobile devices to research and buy products. This makes it difficult to gain a single view of the customer and his/her touch points. Consequently, advertisers and their agencies are having to work harder than ever to build an accurate view of the customer journey and to measure the value of different marketing channels and activities on the path to purchase.

Like many advertisers, O2 historically used last click digital attribution to help measure and optimise online marketing spend. 18 months ago Dan Michelson, Innovation and Capability lead at O2, kicked off a project to reassess how they were using digital attribution.

Knowing your star players and strikers

Describing the challenges at London’s Festival of Marketing last week, Michelson explained that by relying on last click – a technique whereby the value of an online purchase is attributed to the last click or online touch point before the sale is made – they were optimising themselves out of digital marketing.

By attributing value to only the last touch points in the customer journey the business was concentrating spend with just a handful of digital platforms. Michelson was keen to understand the role of other channels and whether they played a critical role in supporting conversions. Using a football analogy he pointed out the importance of understanding the chain of events leading to the conversion – “We knew who our strikers were but who are our star players?”

To shed light on the answers to these questions Michelson introduced a more advanced digital attribution methodology – algorithmic attribution. Unlike last click or other rules-based attribution methodologies which rely on certain assumptions (for example, that 100% of the value of the purchase should be attributed to the last touch point), algorithmic attribution uses statistical models to analyse and weigh the contribution of different touch points to the purchase.

Simply put, the model looks to identify identical customer journeys and their conversion rates to understand the average value of that sequence of touch points. By looking for examples of that sequence with a minor variation and the difference in conversion rate, for example the same sequence with no display ad touch point, the model can then understand the average value of that display ad touch point. This process is repeated across a high volume of different sequences and variations in touch points to build a more robust view of the value of each channel and individual ad without the same reliance on internal assumptions as with rules-based methodologies.

Digital attribution illustration

The results

The results of the first 18 months of algorithmic attribution were very different to those from the last click model and had a significant impact on business decisions:

  • Brand communications showed 10x impact compared to product-focused communications (where brand was previously an area of spend being squeezed by finance due to lack of measurable impact).
  • Mobile display showed 2x impact compared to mobile search.
  • Cost per acquisition for display was reduced by 37pc whilst investing 78pc more.
  • Conversions for affiliates and search, channels that typically show high conversation rates in last click models, reduced by 20pc.

Importantly, the reduced conversion rates for affiliates and search did not necessarily lead to lower spend:

“It’s as much an insight tool as it is an optimisation tool. Conversion rates went down but that doesn’t mean we are investing less. We still need that spend, it just means we better understand their role in the path to purchase and which media support our strikers.”

What next for O2?

There is still a way to go for brands to be able to measure and optimise the end to end customer journey. For O2, Michelson outlines their next steps with digital attribution as:

  • Marketing Mix Modelling: Integrating the two methodologies to bring together offline and online measurement.
  • Cross-device: Stitching together user activity cross device to gain a more complete view of the customer journey.
  • CRM integration: Integrating CRM data (for example, retail and customer service data) in order to optimise across all marketing activities, including understanding when lower cost direct marketing channels are the best option.
  • Activation of attribution: Automating online spend via programmatic buying, optimising the value of placements in real-time according to insight from the digital attribution model.
  • Closer collaboration with key media partners: Including affiliate and social media partners.

The same challenges exist across industry, although some brands will be more advanced than others. With over half of marketers in the UK still using last click methodology, the first step for many will be to evaluate the best toolkit and underlying measurement techniques required to achieve their business objectives. The second step will be to gain buy-in from all the relevant internal stakeholders across the business, involving them throughout the process to ensure the measurement framework will meet their needs and to align expectations. 18 months in to the project Michelson notes that selling the new concept into his business as a two-year initiative was difficult – particularly in a department where decisions are made quarter by quarter. However, he also says it was necessary in order to give enough time to test and refine the KPIs and algorithms and to involve all stakeholders at every step of the journey.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s