81% of web users say the potential problems with personal data collection now outweighs the benefits.
Concerns about security
59% of consumers say knowledge of a data breach would negatively impact buying intentions.
Data from the most common analytics tools is 80% broken.
What solutions are there for the cookieless future?
FLoC, or ‘Federated Learning of Cohorts’, was pitched as a privacy-focused solution for delivering relevant ads “by clustering large groups of people with similar interests”.
This was still based on browsing history but instead of using cookies to send it to third parties, the browser uses your history to decide what you might be interested in.
Accounts are anonymised, grouped into interests, and user information is processed on-device rather than broadcast across the web.
However, following privacy concerns over FLoC, Google have since ditched this proposed solution in favour of their new Topics API.
Similar to FLoC, Topics uses categories of content a user is interested in for targeting. However, Topics will have a far more general level of categorisation and will only house approximately 350 ‘topics’ for targeting purposes.
These limited targeting options remove the possibility of granular bidding for advertisers. For example, if you are a specialist coffee provider, you won’t be able to bid on niche terms like “specialty coffee”.
Instead, you’ll only have a more generic topic like “food and drink” to bid on which will severely increase bidding competition and ad costs.
Corvidae includes a central piece of logic (which we call a ‘hub’), in place of cookies, that collects pixel and clickstream data when a customer either visits your website or views your advertising
This hub connects all devices and locations you would expect to see in a ‘normal’ research and purchasing experience
We then rebuild that journey to solve any potential issues with cross-device and cross-location problems
We then use these comprehensive paths, containing all touch points for all devices in a user journey which are processed by our proprietary AI model – tracing every route through the complicated network of touchpoints to either purchase or to ‘fall off’
It then uses the intersection between journeys to generate a weighting which is translated as how much an event has influenced the outcomes – establishing how effective a marketing touch point has been in driving users towards conversion
The average deployment of Corvidae achieves results like these for our clients:
With a first-party pixel. Our pixel uses an Event Stream architecture, meaning it can be deployed in any conceivable environment, including: • website • app • ad impression • and in custom middleware.
Using our AI model.
Impressions that are not associated with a click are associated using the model’s assessment of the click and click behaviour and the probability that the visit would have seen the impression. This plays to the strengths of AI and is measurable when training the AI.
We test and validate accuracy of this stitching by asking the model to classify conversion journey we feed it as converting or non-converting. We then compare its prediction to the true conversion state of the journey. No model is released which scores less than 85% predictive accuracy.
No, it uses no cookies at all.
Our model uses AI to understand the difference between a transacting and a non-transacting path to discover which touchpoints are influencing in your customers’ decision-making process. Our model is trained to understand each individual customer, demystifying which interactions were important to their decision to buy.
Allowing Corvidae to tune itself to your customer’s behaviour – rather than making you decide what to value in advance.
You can think of the scoring of the AI as a measurement of the incrementality of each event in the conversion path towards a conversion – or non-conversion. This is the basis of Corvidae’s attributed value.
We use anonymised test and train sets from our customers’ historic data to build and validate our models.
These sets are made up of thousands of transacting and non-transacting journeys.
The transacting journeys have transaction events removed.
Given these user journeys (impressions, click throughs and visit hits) we create a vector-based model that predicts whether a transaction would happen or not.
This is then compared to the complete journeys, including transactions, to test how accurate the model is at predicting transactions, and hence judging the engagement of an individual user at different points in their user journey.
Given that the model can weigh each event on how important it is for the transaction or not, that weighting would be used as part of the attributed share.
No. We trialled using device graph and other ID graph data in the early days of Corvidae’s development but found that those data sets added little-to-no incrementality to the model’s accuracy. Today, we use only first-party pixel, AdTech, and other client data, all compliantly collected.