Data-Driven Personalisation Is Broken. Here Is How to Fix It

Data-driven personalisation works when it connects the right message to the right person at a moment that genuinely matters. Most brands are not doing that. They are segmenting by postcode and calling it personalisation, or firing retargeting ads at people who already bought, or showing “recommended for you” carousels that recommend nothing a human would actually want. The infrastructure exists. The intent is there. The execution is where it falls apart.

This article is about closing that gap. Not with a technology checklist, but with a clearer way of thinking about what personalisation is actually supposed to do and why so many programmes underdeliver despite significant investment.

Why Most Personalisation Programmes Underdeliver

A few years ago I sat through a pitch from a large network agency. They had built an AI and machine learning platform that, they claimed, could deliver personalised creative at scale with dramatic improvements in cost per acquisition and conversion rate. The numbers they quoted were striking. I pushed back. When I dug into the methodology, it became clear that the comparison point was their client’s previous creative, which was genuinely poor. They had replaced bad work with slightly less bad work and attributed the improvement to the platform. That is not personalisation success. That is a low baseline dressed up in technology language.

This is the single most common failure mode in data-driven personalisation: the technology gets the credit, and the underlying strategic weakness stays hidden. Personalisation layered on top of weak creative, a confusing value proposition, or a leaky funnel will produce some uplift, because almost anything produces some uplift on a low baseline. But it will not produce the step-change results that the vendor decks promise, and it will not scale.

Before any organisation invests seriously in personalisation infrastructure, it needs to answer a harder question: what are we actually personalising, and does the thing we are personalising deserve to exist in its current form? If the answer is uncertain, fix the fundamentals first.

What Data-Driven Personalisation Actually Means

Personalisation has become one of those words that means everything and therefore means nothing. In a marketing context, it spans a huge range of activity: from inserting a first name into an email subject line to dynamically assembling a webpage based on behavioural signals, firmographic data, and real-time intent. These are not the same thing, and treating them as equivalent creates confusion about what success looks like.

A useful working definition: data-driven personalisation is the practice of using structured data about an individual or a cohort to deliver a more relevant experience than a generic one would provide, in a way that measurably improves a business outcome. Three components matter here. The data has to be structured and reliable. The experience has to be more relevant, not just different. And the improvement has to be measurable against something that actually matters to the business, not just a vanity metric.

This definition rules out a lot of what gets called personalisation. Showing someone a banner ad for a product they viewed three weeks ago, on a site they have already left, for a purchase they have probably already made elsewhere, is not personalised marketing. It is automated stalking. The data is there. The relevance is not.

Personalisation that works tends to operate at the intersection of three things: what the brand knows about the person, what the person needs at that moment, and what the brand is actually able to deliver. When those three things align, relevance is almost automatic. When they do not, no amount of machine learning will compensate.

The Data Foundation: First-Party or Nothing

The shift toward first-party data is not a trend. It is a structural change in how digital marketing works, driven by browser privacy changes, regulatory pressure, and a gradual erosion of the third-party cookie ecosystem that much of the industry built itself on. Organisations that treated first-party data as a nice-to-have are now finding themselves with a personalisation capability gap that cannot be papered over with a technology purchase.

First-party data comes from direct interactions: website behaviour, purchase history, email engagement, CRM records, customer service contacts, loyalty programme data. It is data the brand owns, with consent, and it is the only kind of data that will remain reliably available as privacy standards tighten. Forrester’s work on intelligent growth models has long emphasised that sustainable customer relationships are built on direct knowledge of the customer, not on rented audience data from third parties.

Building a first-party data asset requires a value exchange. Customers share information when they receive something worth having in return: relevant content, better service, exclusive access, genuine personalisation that improves their experience. The brands that have built strong first-party data positions have done so by making that exchange obvious and worth taking, not by burying consent checkboxes in terms and conditions.

The practical starting point is a data audit. Most organisations have more first-party data than they realise, scattered across CRM systems, email platforms, e-commerce databases, and analytics tools that have never been connected. Before investing in new data collection, the question is always: what do we already have, where does it live, and can we actually use it?

If you are thinking about how this fits into a broader commercial growth framework, the Go-To-Market and Growth Strategy hub covers the strategic context that makes personalisation a revenue driver rather than a cost centre.

Segmentation vs. Personalisation: A Distinction That Matters

There is a meaningful difference between segmentation and personalisation, and conflating them leads to wasted effort. Segmentation divides an audience into groups based on shared characteristics. Personalisation delivers a distinct experience to individuals within those groups, or ideally at the individual level. Most organisations are doing segmentation and calling it personalisation. That is not a criticism. Segmentation is genuinely useful. But it is not the same thing, and the gap between the two is where the most significant commercial opportunity sits.

Effective segmentation is the prerequisite for effective personalisation. You cannot personalise at scale without a clear model of who your customers are, what they need, and where they are in their relationship with the brand. The segments need to be built around behaviour and intent, not just demographics. Age and location tell you very little about what someone needs from you today. Purchase history, content consumption patterns, and engagement frequency tell you considerably more.

When I was running iProspect and we were scaling the team from around 20 people to over 100, one of the things that changed our client work significantly was moving away from demographic segmentation in favour of intent-based segmentation. We stopped asking “who is this person” and started asking “what is this person trying to do right now.” The difference in campaign performance was not marginal. Relevance is a function of timing and intent, not of who someone is on paper.

The practical path from segmentation to personalisation runs through progressive profiling: building richer individual-level data over time through repeated interactions, rather than trying to capture everything in a single transaction. Each touchpoint is an opportunity to learn something that makes the next interaction more relevant. This is how personalisation compounds over time rather than plateauing after the first campaign.

Where Personalisation Creates Real Commercial Value

Personalisation is not uniformly valuable across all parts of the customer experience. There are places where it creates significant commercial impact and places where it is largely irrelevant. Knowing the difference is what separates programmes that deliver ROI from those that consume budget without a clear return.

The highest-value applications tend to cluster around three moments: acquisition, where relevant messaging reduces cost per conversion; onboarding, where personalised guidance reduces churn in the critical first weeks of a customer relationship; and retention, where timely, relevant communications extend lifetime value. These are the moments where a more relevant experience changes behaviour in a way that shows up in revenue.

Mid-funnel personalisation, particularly in email and on-site content, is where most brands start and where the evidence base is strongest. Triggered email sequences based on specific behaviours, dynamic content blocks that reflect purchase history or browsing patterns, and product recommendation engines that use collaborative filtering all have well-documented commercial cases. Tools that support growth and personalisation workflows have become more accessible, which means the barrier to entry is lower than it was, but the strategic thinking still has to come first.

The lower-value applications are often the ones that get the most attention: personalised ad creative at the top of the funnel, where audiences are too cold for individual-level data to be meaningful, or hyper-personalised landing pages for low-intent traffic. The effort-to-return ratio is poor. This is where I have seen significant budget disappear without a clear business case, usually because the technology made it possible and no one stopped to ask whether it was worth doing.

Growth frameworks that emphasise experimentation are useful here, because they force a discipline of testing assumptions before scaling investment. Personalisation should be treated the same way: run a controlled test, measure the actual business outcome, then decide whether to scale.

The Technology Layer: What You Need and What You Do Not

The marketing technology landscape for personalisation is genuinely complex, and vendors have a strong commercial interest in making it seem more complex than it is. Customer data platforms, dynamic creative optimisation tools, real-time decisioning engines, personalisation overlays, AI recommendation engines: the stack can grow very quickly, and with it the cost, the integration complexity, and the dependence on specialist technical resource.

Most organisations do not need the full stack. They need a reliable first-party data foundation, a CRM or marketing automation platform that can act on that data, and a clear brief for what personalisation is supposed to achieve. The technology should serve the strategy, not define it. When I see organisations buying a customer data platform before they have a coherent data strategy, I know the programme is going to struggle. The platform will sit underused, the data will remain siloed, and the organisation will conclude that personalisation is harder than it looks, when the actual problem was sequencing.

The right sequence is: define the use case, identify the data required, assess what you already have, build the minimum viable capability to test the use case, measure the outcome, then invest in more sophisticated infrastructure if the use case proves out. This is slower than buying a platform and hoping the use cases emerge, but it produces results that compound rather than costs that accumulate.

BCG’s research on evolving customer needs in financial services makes a point that applies broadly: the organisations that get the most value from personalisation are the ones that connect it to a clear understanding of what customers actually need at different life stages, not the ones with the most sophisticated technology.

Measurement: The Part Most Programmes Get Wrong

Personalisation measurement is where the wheels come off for most programmes. The temptation is to measure engagement: click-through rates, open rates, time on site. These metrics are easy to collect and they tend to improve with personalisation, which makes them attractive as proof points. The problem is that they do not necessarily connect to commercial outcomes. A more relevant email gets opened more often. That is good. But if the open does not lead to a conversion, or if the conversion does not contribute to lifetime value, the business case is not there.

Having judged the Effie Awards, I have seen a lot of effectiveness cases. The strongest ones share a common characteristic: they trace a clear line from the marketing activity to a business outcome, with honest acknowledgement of what was and was not measured. The weakest ones stop at the engagement layer and hope the reader infers the rest. Personalisation programmes need to hold themselves to the same standard.

The measurement framework should be built before the programme launches, not retrofitted after. That means agreeing on the primary business metric (revenue, retention rate, customer lifetime value, cost per acquisition), identifying the control group methodology, and setting a time horizon that is long enough to see real effects. Short-term A/B tests on engagement metrics are a starting point, not an endpoint.

Forrester’s analysis of go-to-market challenges consistently highlights measurement as a gap between strategic intent and commercial delivery. Personalisation is no different. The capability to deliver relevant experiences is increasingly available. The discipline to measure their actual value honestly is rarer.

Incrementality testing is the gold standard. Rather than asking “did people who received personalised messaging convert at a higher rate,” ask “did personalised messaging cause conversions that would not otherwise have happened.” The distinction matters because high-converting customers often receive personalised messaging simply because they are already engaged, which means the personalisation is capturing intent rather than creating it. True incrementality testing isolates the causal effect.

The Consent and Trust Dimension

Personalisation that feels invasive does more damage than no personalisation at all. There is a well-documented phenomenon where customers who feel that a brand knows too much about them, or has used data in a way that feels unexpected, pull back from the relationship. The line between “helpfully relevant” and “uncomfortably knowing” is real, and it is not always where marketers expect it to be.

The practical implication is that personalisation should be grounded in data the customer knowingly shared, used in ways that are consistent with the context in which they shared it. Using browsing data to recommend related products on a retail site is expected and accepted. Using the same data to reference specific browsing behaviour in an email feels like surveillance. The data is the same. The context is different, and context is everything.

Consent architecture matters here. Not just for regulatory compliance, which is a floor rather than a ceiling, but because customers who actively opt into personalised experiences are more valuable than those who are included by default. They are more engaged, more likely to share additional data, and more likely to respond positively to relevant communications. Building a consent framework that makes the value exchange explicit is not just good ethics. It is good commercial strategy.

BCG’s work on go-to-market strategy in regulated industries makes a related point: trust is a commercial asset, and programmes that erode trust in pursuit of short-term conversion gains tend to underperform over longer time horizons. Personalisation is a trust-intensive capability. It should be treated that way.

Building a Personalisation Programme That Actually Scales

Scaling personalisation is a sequencing problem as much as a technology problem. The organisations that do it well tend to follow a similar pattern: they start narrow, with a single high-value use case and a clear measurement framework, prove the commercial case, then expand. The organisations that struggle tend to try to do everything at once, spread resource across too many use cases, and end up with a programme that is broad but shallow.

Early in my agency career, I was handed a whiteboard pen in a Guinness brainstorm when the founder had to leave the room. The instinct was to freeze, because the stakes felt high and the expertise in the room was significant. The lesson I took from that moment was that the people who move forward with clarity, even when the situation is ambiguous, tend to produce better outcomes than the people who wait for permission or perfect information. Personalisation programmes need the same disposition. Start with what you know, test it rigorously, and build from there.

The scaling phase is where organisational capability becomes the constraint. Personalisation at scale requires alignment between marketing, data, technology, and creative teams. It requires a content production model that can generate enough variants to make personalisation meaningful without collapsing under its own weight. And it requires a culture of measurement that treats learning as valuable even when the results are disappointing.

The broader strategic context for this kind of programme sits within go-to-market planning. If you are building or reviewing your growth strategy, the Go-To-Market and Growth Strategy hub is worth working through. Personalisation does not exist in isolation. It is one capability within a commercial system, and its value depends on how well that system is designed.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions