Research insights
Big Data and Customer Behaviour: From Records to Decisions
Big-data analysis examines large or complex collections of records to identify patterns relevant to a decision. Customer transactions, digital events and feedback can provide useful observations, but scale alone does not make them complete or representative. More records can repeat the same blind spot.

On this page
Analysis does not itself change behaviour. A business may use a finding to revise an offer, recommendation or service; the effect of that intervention then needs evaluation. Keep observation, prediction and action separate.

Define what a record actually represents
A device event, account, household and person are different units. A purchase can also contain several users or purposes. Establish the unit before calculating frequency, value or retention, and document how duplicate or incomplete records are handled.
For an illustrative retailer, multiple accounts may belong to one household, while one account may serve several people. Treating accounts as unique individuals can distort the customer story. Linking records requires appropriate authority and safeguards, not just technical ability.
Understand coverage and missing behaviour
Website activity shows what happens within the measured environment. It may not capture purchases elsewhere, offline decisions or people who do not use the channel. A transaction log records what was bought, not every reason behind it.
Check time coverage, changes in event definitions and data quality before comparing periods. A tracking change can look like a behaviour change. Unusual values need investigation rather than automatic deletion or acceptance.
Use predictions with a clear outcome
A model needs a defined target, observation window and evaluation on data not used to build it. Predicting a likely purchase differs from explaining why someone buys. A high score does not establish causal understanding or guarantee the prediction remains useful when conditions change.
Compare the model with a relevant baseline and examine errors across important groups. The consequences of an incorrect recommendation differ from those of a high-impact decision. Avoid inferring sensitive personal circumstances from ordinary shopping activity.
Connect patterns with explanation
Customer interviews, observation and surveys can investigate why a pattern appears. They also have limits, so the aim is to combine evidence appropriately rather than declare one source universally superior. Keep the sample and question consistent with the business decision.
For suitable FMCG purchase questions, iPanel INTAGE is INTAGE Vietnam’s shopper panel and consumer behaviour data platform. Its relevance must be checked against the category, coverage and definitions required; no specific coverage or predictive performance is assumed here.
Maintain a clear distinction between events produced by a system and actions confirmed by a customer. An automatically refreshed page, retry or duplicate transaction record can change a count without a new decision. Agree those definitions with the relevant data owners before presenting a behavioural trend, and retain the explanation in the report so that later teams can reproduce the comparison.
Test the action, not only the analysis
If a pattern suggests a recommendation change, compare the revised experience with an appropriate baseline. Measure completed outcomes, unwanted effects and customer understanding. A visually impressive dashboard does not establish that the action improves the experience.
Qualitative and quantitative research can clarify which evidence answers each question. Explore INTAGE Vietnam’s market research services or discuss a customer-data study with the decision, available records and limits. Responsible analysis begins with a well-defined question rather than collecting everything possible.
Related reading
Latest posts
- Market Research: Building a Market Research Report for Investment Funds
Survey Sample Size: How Many Respondents Are Enough?
Survey Data Collection Methods in Vietnam: How to Choose- Technology Adoption in Enterprises: How to Research Employee Behavior
How to Write a Market Research Brief for an Accurate Quote


