TLDR: 76% of consumers report frustration when a company fails to personalize their experience. Yet most teams are running rule-based tactics that barely move the needle (McKinsey & Company, 2021). There are three levels of personalization maturity: rule-based, segment-based, and algorithmic. Level 2 is where the real commercial impact begins, and reaching it requires a first-party data foundation, a CDP, and deliberately defined behavioral segments before writing a single rule.
The bigger mistake is measuring personalization solely by conversion rates. That framing keeps it small. I argue that personalization is the structural mechanism through which CXO delivers what CRO alone cannot: an experience that responds to who your visitor actually is, rather than what the average visitor might prefer.
Why personalization and CXO belong together
Personalization is the primary mechanism through which Customer Experience Optimization (CXO) delivers on its promise. Without it, CXO is mostly a rebranding exercise. When you make that shift, you commit to something harder than running A/B tests against a single control: you commit to building experiences that respond to who your visitor actually is, not just what the average visitor might prefer. Personalization is what makes that commitment real.
CRO optimizes for the mean. It asks: what version of this page converts better for most people? CXO asks a different question: what experience is right for this person, right now, given what I know about them? Those who’ve worked with me over the past years know I’ve always presented CRO and Personalization as two sides of the same coin. CXO changes that dynamic and puts personalization front and center – where I believe it should be. Personalization is the technical and strategic apparatus that makes that second question answerable. You cannot do CXO properly without it.
| CRO | CXO | |
| Primary question | What converts better for most visitors? | What does this visitor need, right now? |
| Optimization unit | Page or flow variant | Individual or segment experience |
| Primary metric | Conversion rate | Conversion rate Task completion Return rates |
| Data input | Aggregated session data | Behavioral data CRM data Purchase histories |
| Time horizon | Session | Customer lifetime |
| Role of personalization | Optional enhancement | Core mechanism |
What personalization actually means
Personalization means delivering a different experience to a visitor based on what you know about them—their behavior, segment, purchase history, and declared preferences. A handful of tactics routinely get mistaken for it:
- Showing someone’s first name in an email subject line.
- Geo-targeting a homepage banner to display a country name.
- Running a “customers also viewed” carousel using a generic off-the-shelf algorithm.
Those are personalization tactics, and shallow ones at that. Genuine personalization is a strategic program built on first-party data, deliberate audience segmentation, and a systematic test-and-learn process — which is exactly the same infrastructure that powers a mature CXO practice.
Salesforce’s State of the Connected Customer (7th edition, 2024) tracks this over time. In 2020, 65% of customers felt treated as numbers rather than individuals. That figure barely shifted through 2022 (56%) and 2023 (61%) — four years of marginal progress. Then in 2024, something shifted sharply: 73% of customers now report feeling treated as unique individuals, up from 39% the prior year. Unfortunately, the report doesn’t explain what drove the reversal, but Salesforce is using the results to focus on its Agentic AI strategy, as is its prerogative.
Taking these results at face value (the report is based on ± 16.000 surveys, equally distributed across the Americas, Asia, and Europe, as well as evenly distributed across male/female and age groups), it shows that consumers expect and see personalized experiences from the businesses they interact with. The reversal possibly comes from first-party investments accelerated by cookie deprecation and/or widespread AI personalization adoption.
What the 2024 data does make clear is where the remaining gap sits. Only 49% of customers feel brands use their information beneficially. Feeling seen and feeling served are not the same thing. Brands have learned to recognize customers. Fewer have learned to make the data exchange feel worth it. That gap — between recognition and genuine value — is what CXO is built to close, and what CRO, optimizing for the session, will consistently miss.
The three levels of personalization maturity
Most teams that say they “do personalization” are operating at level one. Understanding the full maturity model matters because each level has different data requirements, tooling, and an impact ceiling.
Level 1 — Rule-based personalization delivers a different experience based on simple conditions: new visitor vs. returning visitor, device type, traffic source, geo-location. This is low-effort and low-risk, but the ceiling is limited. Most A/B testing tools — VWO, Convert — support this natively, without a CDP.
Level 2 — Segment-based personalization requires you to define meaningful audience segments grounded in behavioral signals or CRM data: high-intent visitors who haven’t converted, loyalty members who haven’t repurchased in 90 days, visitors who read three or more blog posts but never viewed a product page. It demands a Customer Data Platform (CDP) — tools like Spotler, Bloomreach, Braze, and yes, Salesforce — to unify data across touchpoints and make those segments actionable in real time.
Level 3 — Algorithmic / 1:1 personalization is where the system adapts the experience dynamically for each individual, drawing on real-time behavioral signals and historical data simultaneously. Three barriers have traditionally kept this in enterprise territory: significant traffic volume, clean first-party data, and engineering resources to build and maintain the ML pipeline. The data quality and segmentation requirements are the two most likely to derail a Level 3 attempt before it produces anything commercially useful.
Those two barriers are what rAnalytics Activate — a tool I’m co-developing — is built to address. It handles the clean data pipeline and ships with pre-built segmentation signals, removing the dependency on a dedicated data engineering team. For mid-market teams with sufficient traffic, that meaningfully changes the Level 3 calculus. Volume still matters: below around 100,000 monthly sessions, the behavioral models won’t function reliably regardless of data quality. But for teams that clear that threshold, the data and engineering barrier is no longer the blocker it was.
The most damaging mistake I see is teams attempting Level 3 without the data infrastructure or audience understanding to support it. At one agency client, the brief was clear: deep personalization, and fast. We had just implemented a CDP and went straight into building algorithmic campaigns. However, we had spent insufficient time with the data to understand which signals were reliable and which segments were meaningful. The first campaigns launched had no measurable impact on the experience quality or conversion outcomes. The client’s confidence in the whole personalization program took months to recover.
The business case: what personalization actually moves
When personalization is implemented with a clear data foundation and measurable hypotheses, the commercial evidence is strong. In 2021, McKinsey’s research “The value of getting personalization right — or wrong — is multiplying” found that companies excelling at personalization generate 40% more revenue from those activities than their average-growth counterparts. The same 2021 study found that 76% of consumers said that receiving non-personalized treatment made them frustrated. That figure is now approximately five years old — and as the Salesforce data shows, the industry has since moved. But it established the baseline: the personalization gap was real, measurable, and commercially costly.
At citizenM, we knew American guests came with a specific hesitation: rooms are compact by US standards, and that gap in expectations showed up in the data. We built a personalized video campaign for the American market. The video greeted guests by name – in both video and audio – and addressed the room size question directly. NPS scores improved after we launched this campaign, while cancellation rates remained steady: arriving guests knew exactly what they were choosing. The 40% revenue uplift from the McKinsey document comes from this kind of exchange: removing a real concern for a specific visitor.
Epsilon’s Power of Me research (2017) found that 80% of consumers are more likely to buy when a brand offers personalized experiences. That figure is now nearly a decade old, but it has remained directionally consistent across every major consumer expectation study published since, including McKinsey’s own 2021 tracking data. The expectation of personalization has shifted from differentiator to baseline. Failing to meet it now creates measurable friction.
The trust problem: personalization that helps vs. personalization that creeps
Personalization always had the creepiness factor. The distinction between helpful and creepy is contextual, though. As soon as customers lose sight of the signals you’re using to personalize their experience, it becomes creepy. Showing a returning visitor the product category they browsed in their last two sessions feels useful. Referencing precise location history, browsing behavior across unrelated sites, or data the visitor wouldn’t expect you to hold feels intrusive.
The numbers back this up. Salesforce’s State of the Connected Customer (7th edition, 2024) found that only 39% of customers are comfortable with brands using AI to understand their needs. Despite the fact that most people feel businesses treat them as a client rather than a number. The report attributes the gap to “apprehension around data protection and privacy.” In other words: personalization can be helpful and is even expected, but as soon as people see the data used and do not feel in control of their data being used, they become worried, and rightfully so. This gap is exactly where personalization goes from useful to invasive, and where CXO thinking — with its focus on long-term trust, not just session conversion — becomes the right frame.
The test I apply: would the visitor understand and approve of why they are seeing this? If the logic is transparent and the benefit to them is obvious, it passes. If it serves your conversion goal more visibly than their need – or relies on data signals they would not expect you to have – it fails. GDPR compliance is necessary but insufficient. The higher bar is whether the experience feels designed for the visitor, rather than extracted from them.
In a CXO framework, this matters more than in pure CRO. CXO optimizes for experience quality and long-term customer value. A personalization tactic that erodes trust is anti-CXO by definition, regardless of what it does to the session conversion rate.
The bottom line
Personalization, when treated as a structural program, grounded in first-party data and measured against experience outcomes, is the mechanism that allows CXO to deliver on what CRO alone cannot. An experience that responds to individual need, reduces friction at the right moment, and earns loyalty rather than simply extracting conversions.
If you are shifting from CRO to CXO — or building the internal case for why that shift is strategically necessary — personalization is the capability that determines whether the program produces durable results or a sophisticated version of what you were already doing.
The data foundation, the audience segments, the experience metrics: these are the preconditions for doing any of this properly, and they deserve attention before you write the first rule.
Frequently Asked Questions
What is the difference between CRO and CXO personalization?
CRO personalization identifies which variant of a page or message converts better for a given audience segment. CXO personalization asks which experience best serves that visitor’s needs in the moment — and measures success by experience-quality metrics such as task completion, satisfaction, and return rate, not just session conversion rate.
Do you need a CDP to do personalization?
Not at Level 1. Rule-based personalization runs in most A/B testing tools without a CDP. To reach Level 2 — segment-based personalization using real-time behavioral and CRM data across multiple touchpoints — a CDP is effectively required. Platforms like Spotler, Bloomreach, and Braze are built specifically for this use case.
How much traffic do you need before personalization makes sense?
Rule-based and segment-based personalization can work from around 10,000 monthly sessions. Below that, segments are too small to generate statistical confidence. Algorithmic 1:1 personalization generally requires 100,000 or more sessions per month for the behavioral models to function reliably.
How do you measure personalization ROI?
Against a holdout group that receives no personalization. Track CVR uplift, revenue per session, and — critically — repeat purchase and return visit metrics. McKinsey & Company’s 2021 research found personalization leaders generated 40% more revenue from these activities than average players. Build your measurement framework before launch, not after.
Where should you start if you’re new to personalization?
Start with your highest-intent, lowest-converting segment: visitors who clearly signal purchase intent but don’t complete the purchase. Use session recordings, exit surveys, and funnel drop-off analysis to define the specific friction they’re experiencing. Build a single personalization rule that directly addresses that friction. Measure it properly. One well-executed Level 2 personalization experience — built on real audience insight and measured correctly — is worth more than twenty rule-based banners.