Google turns to AI in bid to move on from last-click attribution

The internet giant claims its “cutting edge” machine learning work will help advertisers better understand the consumer journey and therefore how their marketing leads to a sale.

Attribution remains one of the more difficult areas of advertising. As Andrew Willshire, founder of analytics company Diametrical, recently wrote for Marketing Week, it nearly always excludes inconvenient data or pointlessly analyses events a marketer can’t influence. Last-click attribution is the model that has dominated the internet, in part because it’s easiest to measure, but it has any number of shortcomings.

READ MORE: Attribution is broken, here’s how to fix it

Now Google is looking at new ways of working out what influences a consumer to make a purchase. Using its work in machine learning and artificial intelligence, it claims to be taking the “complexity out of a very complicated problem” with a new tool called ‘Google Attribution’.

The service is being announced at Google’s annual Google Marketing Next event in San Francisco today (23 May). Essentially, it aims to help marketers better understand how their marketing efforts work together to lead to a sale, and therefore how to make their marketing more efficient and effective.

Speaking on a press call to explain the new service, Google’s vice-president of product management Jerry Dischler, said: “There has never been a more exciting time to be a marketer, especially online, because [brands] can access their customers at any moment throughout the day as they bounce across devices including tablets, mobile and desktop.

“But this has created unprecedented complexity. There is more data, more platforms, more channels. On a daily basis Google Analytics processes half a trillion digital moments across devices. What we are trying to do is use tech and machine learning to help make ad platforms more useful, easier and better connected.”

Building a more complete attribution picture

Google Attribution

Google Attribution works by integrating with AdWords, Google Analytics and DoubleClick Search to bring together ad data from across Google to give a more complete view of performance. It is based on tech Google acquired when it bought attribution company Adometry three years ago.

The machine learning works by determining how much credit to assign each step in the consumer journey – whether a first search or a final click before purchase. It analyses each account’s unique conversion patterns to compare the paths of those who convert with those that don’t, meaning personalised results.

Babak Pahlavan, Google’s senior director of product management, claims that where current products on the market are “difficult to set up, require an expert to upload data, lose track of customers when they switch between devices and are not well integrated with bidding tools”, its offering solves these problems.

Google Attribution will roll out globally in beta over the next few months, and is free for any advertiser to use. There will also be an enterprise version ‘Attribution 360’ that Pahlavan says connects with more Google products, such as DoubleClick Campaign Manager and has more sophisticated reports and data import functionality.

The enterprise offering also monitors TV ads as they air and assigns to them any lift seen in search results and website visits.

Online and offline attribution

The second area Google is looking to tackle is online and offline attribution. Here, Google’s offering builds on its store visit tech, which it launched two and a half years ago. It has measured 5 billion store visits across 17 countries and is already used in search, shopping and display campaigns.

We want to help consumers online find what they are looking for in the physical world.

Jerry Dischler, Google

Now Google is expanding that to video and YouTube TrueView campaigns. This will help marketers better understand how video ads impact traffic to stores.

It is also adding the ability to measure in-store revenue, either by retailers importing data from loyalty programmes or through a partnership Google has with an (unnamed) third party that captures 70% of credit and debit card transactions in the US. This will only be available in the US.

Dischler explains: “The offline value of an online ad is just as high, if not higher, than its online value. We want to help consumers online find what they are looking for in the physical world.”

In early testing, Google says it has found that consumers who click on a Google search ad before visiting a store are 25% more likely to make a purchase and spend 10% more on average. And in a Virgin Holidays case study, it found that the brand was generating twice as much profit from its search campaign when it considered both online and store sales, rather than just online.

The challenge of measurement and attribution

Measuring the impact of digital marketing has only become more complex and even the platforms offering it are having problems. Facebook has now admitted to 10 separate errors in its measurements, the most recent of which resulted in it over-billing some brands.

This move from Google goes some way to improving measurement and attribution. But marketers will still be relying on Google to get its tech and data right.

Plus while the company claims it is “across all channels” it will still be very difficult to get data that allows a marketer to compare side-by-side using the same metrics how digital advertising compares to more traditional, or indeed how advertising on Facebook compares to Twitter or Snapchat or Google.

READ MORE: Keith Weed – We need comparable metrics in a digital world

Marketers including Keith Weed and Sir Martin Sorrell have previously complained about the big digital firms “marking their own homework”. And this does little to alleviate that concern.