Customer Demand and Behaviour Intelligence:
understand what creates, redirects and suppresses demand

What is changing in customer need, behaviour and choice?

Demand intelligence must explain behaviour, not merely describe audiences. The central question is not “Who is our customer?” but “What outcome is a person or organisation trying to achieve, under which circumstances, through which decision process, and what prevents or accelerates action?”

A reliable system combines observed behaviour, stated motivations, commercial outcomes and contextual change. It treats search, social and platform data as partial signals rather than population truth. It also distinguishes a change in underlying need from a change in channel, timing, affordability or measurement.

1. The demand model

Analyse demand through six connected layers:

1. Context: economic, social, regulatory and category conditions.
2. Trigger: the event that moves a latent need into active consideration.
3. Outcome sought: functional, emotional and organisational progress.
4. Choice system: users, buyers, approvers, influencers, channels and substitutes.
5. Friction: financial, behavioural, operational and trust barriers.
6. Observed result: adoption, use, repeat, expansion, switching or abandonment.

This prevents a common mistake: interpreting purchase as the whole customer journey. In B2B markets, the user, economic buyer and risk approver may value different outcomes. In consumer markets, household constraints, platforms and social context can alter realised demand.

2. Evidence architecture

Use four evidence families and reconcile them:

• Behaviour: transactions, usage, journeys, service contacts, search and switching.
• Voice: interviews, ethnography, complaints, communities and surveys.
• Economics: price, income, total cost, availability, substitution and retention.
• Context: demographics, regulation, channel and competitor change.

Official statistics provide essential baselines. Hong Kong’s 2024/25 Household Expenditure Survey uses a scientifically designed sample and expenditure diaries; its results update CPI weights and support study of consumption behaviour.¹ This is more defensible for population patterns than platform anecdotes, but it cannot explain every motivation. Primary qualitative work supplies mechanism; behavioural and experimental evidence tests whether it predicts action.

UNCTAD estimates business e-commerce sales of US$27 trillion in 2022 across 43 economies and notes that digital adoption, logistics, packaging, returns and consumer behaviour jointly shape outcomes.² The implication is analytical: “digital demand” should not be treated as a channel count alone.

3. Research sequence

Discover

Map the decision journey and recruit by situation, not only demographics. Interview recent adopters, rejecters, switchers and non-consumers. Ask for concrete past events before hypothetical preferences.

Quantify

Estimate incidence, frequency, value and heterogeneity. Design surveys around decisions to be made; disclose sample, weighting, field dates and uncertainty. Separate awareness, stated intent and actual conversion.

Validate

Use pilots, choice tests, pricing experiments or phased releases where feasible. The UK Magenta Book stresses that evaluation should be planned early and matched to the question; experiments estimate effects, while theory-based and process evidence explain how, why and for whom results occur.³

Monitor

Create leading indicators tied to triggers: consideration, qualified search, enquiry quality, trial, time-to-value and early repeat. Lagging revenue alone identifies change too late.

4. AI-native application

AI can classify high-volume feedback, translate multilingual evidence, retrieve analogous cases and propose competing explanations. It should not silently infer representativeness, causality or emotion. Require source links, sampled human verification, subgroup error checks and explicit confidence. Personal data must be purpose-limited and governed.

5. APAC and Hong Kong application

Use a common outcome taxonomy but localise language, buying roles, channels, affordability and trust. Do not infer Hong Kong behaviour from “Greater China” or APAC averages. Hong Kong household expenditure, retail sales and population statistics provide local baselines; cross-border travel and purchasing may require separate evidence because local retail statistics do not capture every purchase made by residents outside Hong Kong.⁴

6. Decision outputs

• demand map by situation and outcome;
• trigger and friction model;
• choice-system map for B2B or household decisions;
• demand signal dashboard with thresholds;
• evidence-backed opportunity hypotheses;
• experiment and learning agenda.

7. QA and performance

Track predictive validity, segment stability, time from signal to decision, research coverage of non-customers, and the gap between stated intent and observed action. Review whether samples exclude low-visibility groups and whether a change reflects demand, supply, price or measurement.

8. Decision playbook

Use the evidence differently by decision. For proposition design, prioritise outcomes, alternatives and friction. For demand forecasting, quantify triggers, eligibility and conversion. For retention, examine realised value, habit, service failure and switching conditions. For channel strategy, distinguish discovery, evaluation, transaction and support: a channel may dominate one stage but not the whole journey.

Create a demand confidence matrix. Rate each major proposition by behavioural evidence, qualitative explanation, population evidence and causal validation. A strong qualitative insight with weak prevalence should prompt quantification; a strong correlation without mechanism should prompt diagnostic research; a claimed causal effect without comparison should prompt a test.

9. Common failure modes

• treating customers as the market and excluding non-consumers;
• confusing satisfaction with future retention;
• using social listening without a coverage model;
• asking leading questions about a proposed solution;
• aggregating different buyers, users and situations;
• attributing a sales decline to demand without testing price, availability or execution;
• allowing dashboards to obscure the reference period and denominator.

10. Review questions

Before action, leaders should ask: Which observed behaviour supports the conclusion? Which group is missing? What alternative mechanism fits the same data? Is the effect large enough to matter economically? Which next signal would change the decision? These questions turn customer research into an accountable commercial input.

Sources

1Hong Kong Census and Statistics Department, *2024/25 Household Expenditure Survey*. https://www.censtatd.gov.hk/en/page_1351.html

2UN Trade and Development, *Digital Economy Report 2024*. https://unctad.org/publication/digital-economy-report-2024

3HM Treasury and Evaluation Task Force, *Magenta Book: Central Government guidance on evaluation* (2026). https://www.gov.uk/government/publications/the-magenta-book/magenta-book-central-government-guidance-on-evaluation-html

4Hong Kong Census and Statistics Department, *Retail Sales and Consumer Spending*. https://www.censtatd.gov.hk/en/page_213.html