In a rather unexpected turn of events on Tuesday evening Google announced that it will replace FLoC (Federated Learning of Cohorts) with the Topics API, a new interest-based targeting proposal. Given that Google plans to deprecate third-party cookies at some point in 2023, what does this pivot in approach mean for the industry?

 

What is the Google Topics API Proposal?

Topics is the latest proposal in Google’s Privacy Sandbox initiative that aims to improve privacy for its users whilst enabling publishers and developers to build thriving businesses. At its core the API will provide up to five topics of interest based on the user’s Chrome browsing history over the past 3-week window and share them with participating websites.

 

How is Google Topics API Different from FLoC?

  • The Google Topics API does not group users intro cohorts, which addresses the Electronic Frontier Foundation’s concerns over the FLoC’s weakness to potential fingerprinting techniques.
  • The Topics API will associate topics with users on a weekly basis, but these will only be kept for three weeks. Under FLoC cohorts would have been assigned on a weekly basis based on previous week’s activity.
  • The Topics API will share three topics with websites, whilst FLoC planned to share a cohort ID.

 

What Are the Strengths of the Topics API Proposal?

On the surface Google’s Topics API has a lot going for it:

  • Privacy. Topics achieves a stronger level of privacy and user transparency when compared to the FLoC proposal.
  • Remarketing in a post-cookie world. Topics allows for directional remarketing based on previous site browsing behaviour.
  • Non-contextual monetisation. Interest-based advertising allows sites that perhaps could not be easily monetised via contextual advertising to display more relevant ads to the user. This should help fund more websites that the user visits, keeping with Google’s mission of supporting access to free content online.

 

What Are The 5 Main Limitations with the Topics API?

 

1) Interest Targeting Latency Will Lead to Wasted Advertiser Spend

According to Google’s proposal the API “returns an array of up to three topics, one from each of the preceding three epochs (weeks).”

As our purchase-making decision timescales vary from the short term (e.g. buying shampoo) to the long term (e.g. buying a car), a three week interest latency will likely lead to consumers being targeted with ads for old needs that have already been met. This will not only be annoying for users (picture being pestered by a product that you bought 2 weeks ago), but will also lead to a particularly large budget wastage for brands whose products have short-term purchase journeys.

 

2) Classification at Hostname Level Will Lead to Intent Over-Generalisation

According to Google’s proposal “The browser will leverage a classifier model to map site hostnames to topics…We propose choosing topics of interest based only on website hostnames, rather than additional information like the full URL or contents of visited websites.”

A hostname (e.g. www.example.com) rather than the full URL targeting approach (e.g. www.example.com/target-url) will inevitably struggle to accurately classify big sites and publishers which cover a wide variety of industries, interests and intents. Narrowing a website’s meaning to 3 interests will be restrictive, reducing the accuracy and effectiveness of future targeting as a result.

 

3) User’s Topic Order Randomisation Will Make Targeting Less Accurate

According to Google’s proposal “The returned array is in random order. By providing three topics, infrequently visited sites will have enough topics to find relevant ads, but sites visited weekly will learn at most one new topic per week.”

A random order suggests that interests will not be able to be optimised by weighting and/ or order priority. This reduces the relevancy and accuracy of advertising since ads for say car hire might appear for users who have visited just 5% of relevant sites in the past 1-3 week period:

  • Topic 1 – Model Airplanes – (70% share of sites visited)
  • Topic 2 – Woodworking – (15% share of sites visited)
  • Topic 3 – Car Rental – (5% share of sites visited)
  • Remaining share of 10% sites will not be displayed

 

4) Google’s Addition of Noise to the Topics (Might) Further Limit Targeting Accuracy

According to Google’s proposal “For each week, the user’s top 5 topics are calculated using browsing information local to the browser. One additional topic, chosen uniformly at random, is appended for a total of 6 topics associated with the user for that week/epoch…. There is a 5% chance that the random topic is returned.”

Whilst Google says this is to improve privacy and reduce fingerprinting, this in theory will lead to a further 5% decrease in targeting accuracy for the advertisers. From our current interpretation of the technical specification, this ‘noise’ will work in broadening targeting on Google’s platform, which seems quite misleading as an advertiser may promote car hire to people who were never interested in car hire over the last 3 weeks.

 

5) Option to ‘Opt-Out’ of Topics API May Be Used Against Google by Competitors and Regulators

According to Google’s proposal “If the user opts out of the Topics API, or is in incognito mode, or the user has cleared their cookies or topics, the list of topics returned will be empty.”

Whilst on the surface it’s great that Google plans to make opt-out features available, this may be used by privacy activists and competitors such as Apple to heavily promote toggling this option, reducing Google’s ability to access and sell data. In theory regulators could even mandate that the feature to ‘opt-in or opt-out’ is displayed front and center when setting up Chrome on your computer or phone, just like they did with Explorer and Microsoft Windows devices back in the 2000s.

 

How is the Nano Solution Different from Google’s Topics API?

  • No Targeting Latency. At Nano we offer advertising based on ‘live intent’, meaning that we can target users in the moment that they search for and consume the intended content, rather than intent that happened 1-3 weeks prior.
  • Precision Targeting. Nano’s understanding of intent occurs at the granular keyword and URL level, meaning that instead of targeting say an ‘Automotive’ topic we can target the specific automotive manufacturer and even model of car that the content is about. Our ability to overlay sentiment and semantic meaning over intent targeting enables a wide array of granular strategic targeting tactics.
  • No Predefined Category Limitations. Unlike Topics, which will be limited by Google’s choice of classification intents (starting at ~350 topics), Nano is able to create bespoke intent-based audiences that directly meet the client campaign objectives.
  • 100% Identity-Free. The original privacy issues with FLoC and the current workaround with Topics whereby users can manually amend or ‘opt-out’ altogether, highlight some of the issues we are discovering as an industry with cookie workarounds. Being identity-free means we are privacy first and fully scalable.

Take your campaigns to the next level by learning more about the Nano LIIFT™ platform here.