Andrew Hood is the founder and CEO of independent analytics consultancy Lynchpin, and is a long-time friend of Econsultancy, both as a trainer and a go-to source for insight on analytics.
Here, he gives his expert thoughts on what 2025 will hold for data and measurement: why Google’s U-turn on cookies hasn’t changed the direction of travel all that much for first-party data; the skills needed to tackle marketing measurement; and how the ‘hype cycle’ around generative AI could give way to more focused applications.
Andrew Hood: Algorithms are spending everyone’s money, and the data used to evaluate the results are increasingly modelled rather than observed.
Keyword choices, audience targeting and advertising placements are optimised by algorithms that promise reach into granular areas of relevancy that would have been unachievable for manual planning approaches.
But then the evaluation of performance of these optimised campaigns is increasingly based on modelling – using machine learning and synthetic data to plug the gaps arising from lack of consent for directly observed measurement.
If those models are designed by the same networks or platforms designing the spending algorithms – as is often the case – then there is a real risk of bias emerging in the data (and hence the perception of performance) without independent measurement.
Ultimately while Google has U-turned on third-party cookies, Apple has not – so for most brands the overall direction of travel is still very much around maximising the opportunity of first-party data, not least to gain competitive advantage.
In line with this, there is a vendor focus on first-party data activation, and CDPs are being pushed heavily as an independent category at the same time as CDP functionality is increasingly becoming embedded into other incident platforms such as CRM.
There is a risk of overinvestment in technology versus people and process.
There is a risk of overinvestment in technology versus people and process. Organisations that focus in on the specific first-party data activation use cases that will drive value – and the data flows required to enable them – are better placed to get a return on their overall investment.
Marketing measurement is becoming more complex – legally, technically and analytically. A new blend of skills is increasingly required to successfully triangulate attribution, marketing mix modelling and incrementality testing, and to address the nuances of data collection itself. But equally critical is the business experience to align these techniques with the commercial backdrop to make sure the results can drive effectiveness.
In the digital ecosystem, the collection of data is increasingly shifting from client to server side, which means a much broader range of engineering and infrastructure skills and experience are coming to the fore as critical for success.
Start thinking more about data management, and specifically capturing data in a way that makes it ready for AI. While humans fill in missing details using experience, algorithms build up from details and use labelling for context. Investment in data quality now is going to be a massive future value enabler.
Continue treating first party data as a shared asset between you and your customers – with safeguarding and segmentation as well as activation core to the remit. If first-party data is being shared, then it needs to be consented. And effective segmentation is key to understanding the value of that data and being relevant to customers.
Avoid deploying new technologies in a way that bypasses transparency, control and hence trust in how customer data is being used and shared. For example, a hashed email might seem anonymous on the face of it, but it’s more permanent than a cookie as an identifier – when was the last time you “cleared out” your email address?
All hype cycles start with a technology trigger, and in the first instance vendors will be selling into that cycle aggressively in 2025.
Every data, marketing and productivity platform will have more GenAI stuffed into it, licence costs will go up accordingly based on a promised return… until the reality and real cost of effectively operationalising it all likely hits. And then a second wave of far more focused applications will likely emerge.
I wish the AI buzzword was talked about in a more functional and holistic sense, especially in data and analytics. Recent advances in GenAI and LLMs are exciting, but that doesn’t make them the right solution to every problem.
I wish the AI buzzword was talked about in a more functional and holistic sense…
AI still includes non-generative machine learning, optimisation, simulation, knowledge graphs, heuristics… and often these techniques can be more efficient, more explainable and use a lot less compute and electricity.
A safe bet is an increasingly divergent approach to AI regulation between the EU and US, with the UK likely somewhere in the middle in every sense. Particularly for companies and markets operating globally, it’s going to make universal AI adoption and governance a headache in the immediate term.
In less safe (or at least less predictable) bets, Trump meets Google could change the whole face of search and advertising… or not. The DOJ already thinks the world’s largest (by market share) advertising network, digital analytics tool, web browser and mobile operating system should not be owned and operated by the same business.
If, as currently proposed, Google was forced to sell off Chrome, then their stranglehold on the digital advertising market could fall away quite quickly, just at the same time as search engine results pages themselves are being disrupted by AI.
Read more of Econsultancy’s expert predictions for 2025.