CX Is a Statistical Science: Building a Customer 360 Across Nine Businesses

IN ONE MINUTE

Customer experience is often treated as an art. In my experience it is a statistical science. Customer behaviour follows life stages that repeat with remarkable consistency, and the data shows them, if you capture it well. In 2018–2019, in a Hong Kong conglomerate spanning property, retail, hotels, insurance, healthcare and education, I connected customer data from nine businesses into one consented membership. E-commerce sales more than quadrupled, and cross-selling brought about three-quarters more new customers to the group's businesses. The tools have since moved on to AI agents; the principle has not.

THE OBJECTIVE

Most large groups already own the data they need to serve customers better. It simply sits in separate businesses, systems and loyalty schemes, each seeing one slice of the same person. This issue shows how one customer view was built across a diversified group, how machine learning turned it into predictions of what customers would need next, and what each member of the leadership team should do to make customer experience measurable, repeatable and trusted.

WHAT THE RESEARCH SHOWS

Customers expect relevance. McKinsey found that 71% of consumers expect companies to deliver personalised interactions and 76% get frustrated when that does not happen. Done well, personalisation most often drives a 10–15% revenue lift.

The gap between leaders and the rest is widening. Deloitte Digital's September 2026 study found that 36% of brands with mature personalisation saw revenue grow by 15% or more from that investment, against 16% of low-maturity brands. It also found that 63% of US consumers used AI-powered search tools or assistants to find products or services in the past year, and one in five now start their search there.

Yet the overall experience is slipping. Forrester's 2025 Global Customer Experience Index found that 21% of brands declined and only 6% improved, with CX at an all-time low in North America.

And personalisation can backfire. Gartner's survey of 1,464 buyers and consumers found that personalised marketing created negative experiences for 53% of them, who were 3.2 times more likely to regret a purchase.

Academic research explains where the opportunity lies. As far back as 1984, Alan Andreasen showed in the Journal of Consumer Research that people going through major life changes are more open to new brands and preferences. Life events are when loyalties are formed.

The message: relevance pays, but only when it is accurate, timely and welcome.

THE CASE: ONE CUSTOMER, NINE BUSINESSES

The situation

In 2018–2019 I headed the digital and CRM division of an ambitious unit under the CEO, within a Hong Kong family conglomerate. Its listed and private companies spanned property, retail, jewellery, hotels and hospitality, insurance, infrastructure, healthcare and education, in Hong Kong and much of Greater China. My role was to coordinate the business verticals so the group could see each customer whole: a customer 360.

The scale was both the opportunity and the challenge. A family might shop in the malls, stay in the hotels, insure their home, choose a hospital for their first child and a school a few years later, all within the same group, and still be a stranger to each business. Every unit held a fragment. None held the customer.

Building the platform

We started with experiences customers valued, not with a database.

• An edutainment ecosystem. We combined the group's education and entertainment businesses into one proposition for affluent families and young parents: retail, social media, an online forum, e-commerce, education, intellectual property, food and beverage, a children's playhouse and publications.
• A digital launch for a new flagship destination. I led the digital launch of the flagship, setting footfall and ticket targets and integrating online booking, social media, digital marketing and conversion tracking with on-site operations.
• A measurable venue. We upgraded the venue's digital infrastructure with a mobile app, iBeacon indoor positioning and interactive wayfinding kiosks. Location-based, time-limited offers made each visit personal, and every offer measurable.
• One membership, one login. We connected customer data from nine businesses into one online-to-offline membership and e-commerce platform with single sign-on. Customers gained one account and consistent benefits; the group gained one view of each member.

Consent as a value exchange

None of this works without trust, so I led the redesign of the membership consent form. Before enrolling, every customer chose whether to share their data only with the business they bought from, or with the whole group, and the form named each company that would receive it. In return, a purchase anywhere in the group turned into points they could redeem across a far wider range of rewards. Customers could see the value of sharing their data, and enjoyed the benefits straight away, at no cost to them.

From data to prediction

I then led a team of PhD data scientists to build a machine-learning model that identified likely customers for products and services across the group, from property and education to insurance and consumer goods. The objective was single: understand customer needs well enough to cross-sell across the group.

What we found was that behaviour is highly predictive, because it follows a life cycle. After graduation, people invest in themselves and their personal brand as careers and incomes grow. A few years later, many start building a family, so they rent or buy a home. Then come baby products, hospitals, insurance and schools. The path repeats. Capture the right signals and serve the right offer at the right moment, and the sale follows.

I often say data is the most honest thing in business. Behaviour does not lie, because we spend our time and money on what we want and trust. If you and your family have no interest in gaming, you will never study the storage specifications of the next games console. The job was to capture the right data with minimal noise, categorise behaviour, understand the logic behind each action, and predict the next need from probability, not intuition.

What has changed since, and what has not

The tools have moved on since 2018. Propensity scores built by a data-science team have given way to real-time data platforms, large language models that read unstructured signals such as searches, chats and reviews, and AI agents that can act on a prediction the moment it is made. Discovery is shifting too: in Deloitte Digital's study, one in five consumers now start their search in an AI tool.

The principle has not moved. A large language model or an AI agent still needs to know who the customer is, what they have agreed to share, and where they are in life. The life-stage trigger, and the probability that a need follows it, is the same logic, now calculated faster and acted on automatically. What has changed is speed and scale. What decides the result is still identity, consent and measurement: the foundations this case was built on.

The results

Measure	Result, 2018–2019
Transaction value across the ecosystem	Close to US$200 million
E-commerce sales	More than four times higher
New customers won through cross-selling between the group's businesses	Up by about three-quarters
Repeat purchase	About a third of members
Cross-sell	About a quarter of members bought from a second business
Fulfilment cost	Down by about a tenth

Segmented welcome, reactivation and cross-sell journeys did the work. None of them depended on a clever campaign. They depended on knowing who the customer was, where they were in life, and what they were likely to need next.

THE FRAMEWORK: FIVE STEPS FROM DATA TO THE NEXT BEST OFFER

Step	The question	What it takes	How to measure it
1. Capture	Are we collecting the signals that matter, with minimal noise?	Consent at sign-up, purposeful data points, clean event tracking	Share of customers with complete, consented profiles
2. Connect	Do all our businesses recognise the same customer?	One membership, single sign-on, a shared customer ID	Share of transactions linked to a known member
3. Classify	Do we understand where each customer is in life?	Segments built on behaviour and life stage, not demographics alone	Segment stability and size
4. Predict	Can we anticipate the next need?	Propensity models, and now LLMs and AI agents, tested against real purchases	Prediction accuracy; uplift versus a control group
5. Serve	Do we act at the right moment, through the right channel?	Welcome, reactivation and cross-sell journeys, online and on site	Conversion, repeat purchase, cross-sell, cost to serve

Each step depends on the one before it. A brilliant model on unconnected data predicts nothing useful. Connected data without consent becomes a liability. And a prediction no one acts on is only a report.

THE PART THAT NEEDS CARE

The same data that makes an experience feel personal can make it feel intrusive. Gartner's finding that personalisation can triple the likelihood of purchase regret is a warning: relevance without restraint damages trust.

Regulation sets the floor. Hong Kong's Personal Data (Privacy) Ordinance, like the UK GDPR, requires clear notice and consent before personal data is used for direct marketing, and written consent before it is passed to another company for its own marketing. Within a group of separate companies, that is not a technicality. It shapes how the membership, the consent and the data flows must be designed from the start.

Trust sets the ceiling. My test is simple: would the customer, seeing why they received an offer, think it was helpful? If the answer is no, the offer should not be sent, however accurate the prediction.

WHAT THIS MEANS FOR CEOS AND BOARDS

Treat the customer view as group infrastructure. In a diversified group, no single business has the incentive to build it alone. It needs a sponsor at the top, a shared budget and a mandate across every unit.

Ask how many of your customers you actually know. Request the share of revenue linked to identified, consented customers. It is a better measure of customer strategy than any satisfaction score.

Govern trust as closely as revenue. Consent, data sharing across entities and the use of prediction in marketing belong on the board's risk agenda, not only the marketing plan.

WHAT THIS MEANS FOR CMOS AND CUSTOMER LEADERS

Design journeys around life events, not campaigns. Map the moments when your customers' needs change, and build welcome, reactivation and cross-sell journeys around them.

Prove uplift, not activity. Keep a control group for every journey. Report incremental sales, repeat purchase and cross-sell, not opens and clicks.

Prepare for AI-led discovery. With one in five consumers in Deloitte's study now starting product searches in AI tools, accurate, structured product and service information decides whether AI assistants recommend you at all.

WHAT THIS MEANS FOR CFOS AND FINANCE TEAMS

Value the customer, not the transaction. Lifetime value and cross-sell rate show what a connected customer is worth across the group, and justify the platform investment better than campaign returns.

Count the cost savings too. One platform and one login reduce duplicated systems, acquisition spend and fulfilment cost. In our case, fulfilment cost fell by about a tenth.

WHAT THIS MEANS FOR DATA AND TECHNOLOGY LEADERS

Start with identity. A shared customer ID and single sign-on matter more than the choice of analytics tool.

Collect less, but better. Every data point should have a purpose and a consent behind it. Noise is not neutral: it degrades every model built on top of it.

Test predictions against reality. Validate models on real purchases and control groups before scaling, and monitor them as customers' lives and markets change.

THE LAST WORD

Customer experience is not the art of guessing. It is the science of serving: predictable, repeatable, and only as strong as the trust behind the data.

If your customer's next life event happened tomorrow, which of your businesses would know first?

I will be exploring how AI is changing search, discovery and conversion on the AI in Retail panel at London Tech Week on 16 November.

—

Until next week,

Artificial Ian

→ Found this useful? Forward it to a CMO, CDO or customer leader.

→ Subscribe to The AI Executive Briefing, weekly in your LinkedIn inbox.

Sources: McKinsey & Company, "The value of getting personalization right—or wrong—is multiplying", 12 November 2021. Deloitte Digital, "Personalizing growth", 29 September 2026 (1,000 US consumers and 480 US B2C brand leaders, April 2026). Forrester, Global Customer Experience Index 2025, 24 June 2025 (275,000+ customers, 469 brands, 13 countries). Gartner, "Gartner Survey Reveals Personalization Can Triple the Likelihood of Customer Regret at Key Journey Points", 3 June 2025 (1,464 B2B buyers and consumers, November–December 2024). Alan R. Andreasen, "Life Status Changes and Changes in Consumer Preferences and Satisfaction", Journal of Consumer Research 11(3), 1984. Hong Kong Personal Data (Privacy) Ordinance, Part 6A (direct marketing). Programme results (2018–2019): the author's own programme data.