Trends & strategies

Forecasts on this subject have a poor record. The much-announced overhaul of advertising technology turned out differently from what was expected, and a considerable part of what counted as the future three years ago is now settled or obsolete. This page therefore tries less to predict the market than to name the developments that have already happened and have consequences for working with intent signals.

The cookie phase-out that never came

For years the expectation was that Chrome would abolish third-party cookies and thereby end a large part of data collection across other people’s websites. Google postponed that step repeatedly and finally declared it would keep the cookies.

That is not a free pass. Safari and Firefox have blocked third-party cookies by default for years, and the legal situation in Europe does not depend on what any given browser allows: consent is required for access to the end device, regardless of technical enforceability. The practical consequence for providers of intent data has therefore not changed. Anyone relying on broadly collected third-party data is working on a basis whose origin they do not control.

What follows is unspectacular: building your own data holdings remains the most stable route. Not because a deadline looms, but because your own data can be verified. The differences between the origins are under fundamentals.

AI answers change what becomes visible at all

The more noticeable shift is happening in search. Search engines increasingly answer questions directly, and a growing share of research runs through assistant systems that summarise sources instead of sending people to them. Visibility and clicks come apart as a result.

For intent signals that has two consequences. First, first contact shifts: someone who learns from an AI answer that your product exists leaves no trace with you. The share of research that stays invisible grows. Second, the nature of the accesses that still arrive shifts: they are rarer but on average more targeted, because the general questions have already been answered elsewhere.

In practice this means adjusting the yardstick. A drop in visits with the same number of enquiries is not a problem but the expected consequence. Anyone still carrying reach as a measure of success will draw the wrong conclusions from it.

AI in the analysis: what holds, what does not

Language models are good at classifying texts by topic. Automatically deriving from a set of pages which content belongs to which topic, and classifying accesses accordingly, works reliably and saves maintenance work. That is the use with the clearest benefit.

Predicting purchase probability is a different matter. A model meant to learn which signal patterns lead to closed deals needs many completed cases – won as well as lost. A company with a few dozen deals a year does not have that volume of data. What then gets sold as “AI scoring” is usually a rule set with weights that could just as well be written down by hand, and would at least stay comprehensible that way.

The sober rule of thumb: language models for understanding content, simple rules for scoring – until enough cases exist to justify more.

Regulation

The European framework remains the determining factor. The GDPR applies unchanged, the national implementation of the end-device rules sits in the TDDDG, and the attempt to harmonise the rules EU-wide through an ePrivacy Regulation made no progress for years. For practice this means there will be no simplification to lean on for the foreseeable future.

On top of that comes the EU AI Act, in force since 2024 and becoming applicable in stages. For most marketing applications it mainly produces transparency and documentation obligations; the more far-reaching requirements hit other fields of use. Where your specific use falls is an assessment that belongs with legal guidance.

The practical consequence is the same as at the start of any such project: document what is processed, why and for how long. Those records arise during setup anyway; reconstructing them later is the considerably more unpleasant option.

From the individual contact to the buying group

A development that has less to do with technology than with how the process is viewed: scoring individual contacts is increasingly being replaced by scoring the account. The reason lies in the matter itself – several roles are involved in a purchase, researching at different times and with different questions.

Score only individual contacts and you see five weak signals instead of one strong one. Score the account and you see the pattern. For outreach it follows that different roles need different material: the specialist department asks about function, IT about interfaces and operations, purchasing about terms and contract conditions. Most companies are missing this material for at least one of the roles.

Signal quality instead of signal quantity

The reflex to add further data sources when the results are unconvincing rarely helps. More sources mean more hits, not better ones.

What actually helps is less spectacular:

• Confirmation instead of addition. A signal arriving from two independent sources is considerably more dependable than two signals from one. Use additional sources to verify, not to lengthen the list.
• Build in decay. Signals lose value, and quickly. A score that does not fade turns within a few months into a ranking of the most active readers.
• Take negative signals seriously. Visits to careers pages, press pages or the legal notice usually point to something other than a purchase. They belong downgraded, not ignored.
• Establish feedback. The only dependable check is comparison against how the deals turned out. Without that feedback, every weighting stays an assertion.
• Maintain exclusion lists. The most boring point, and the one with the best ratio of effort to effect.

What does not change

For all the movement, three things stay the same. A signal from a company that does not fit the offering stays worthless. An approach with no recognisable occasion stays a nuisance. And an analysis built on faulty measurement stays wrong, however sophisticated the model on top of it is.

Anyone wanting to prepare is therefore well advised to put their own foundation in order first: clean collection on their own website, maintained master data in the CRM, a documented legal basis, defined events. What that looks like technically is under technology and integration. That work is useful regardless of where the market goes – and it is the precondition for any new tool to achieve anything at all.

The B2B intent signals page gives an overview of the whole series. If you want to work out which of these developments affect your business, drop us a line – you can also book a slot directly there.