What decision
Market choice, channel assessment, customer targeting, product changes, pricing, or operating review?
Data Analysis
Define the decision first, then turn scattered data into action
Work backward from the business question to metrics, definitions, and sources. Resolve missing, duplicate, and conflicting data before choosing a report, dashboard, or recurring workflow, and state the evidence and limits behind every conclusion.
Decision first
Data analysis begins with the action, not the chart
“We have spreadsheets; find insights” is difficult to deliver reliably. We first define the decision, subject, time range, and user, then determine the data and depth required.
Who and when
Define company, region, product, user, channel, comparison group, and time range.
How metrics work
Name, formula, status, deduplication, time field, and source must align for valid comparisons.
What happens next
Agree who uses the result and what action a rise, fall, or anomaly should trigger.
Decision pipeline
Trace every conclusion to its definitions, sources, and transformations
Value comes from usable data, consistent metrics, and action, not chart count. Exceptions and limitations belong in the result.
DATA PIPELINE / CONTROLLEDSTATUS / NOMINAL
Decision
Subject, scope, and intended action
Data intake
Authorized sources and quality
Metric model
Clean, join, deduplicate, calculate
Analysis output
Findings, dashboard, and refresh
ACTIONABLE SIGNALExplainable, reusable findings that support action
SourceTraceable
MetricDefined
LimitsDisclosed
Analysis path
Choose a report, reusable analysis, or dashboard by decision frequency
Not every question needs a large dashboard. A one-time decision, periodic review, and daily operations need different investments and refresh mechanisms.
01
PATH 01 / DIAGNOSIS
Focused diagnosis and research
For market selection, channel assessment, competitive observation, or another bounded question.
02
PATH 02 / REUSABLE
Reusable analysis and processing
For recurring spreadsheet merges, field cleanup, and metric calculations, with scripts and templates.
03
PATH 03 / OPERATIONS
Operating dashboards
For frequent decisions, combining metrics, pipelines, access, refresh, and exception monitoring.
Have data but do not know what it can answer?
Provide redacted samples, field notes, and the business decision. We will assess quality and a feasible scope first.
Use cases
Analysis scenarios that begin with an operating question
These include cross-border e-commerce and broader teams working with business, authorized public, or multi-table data.
Industry and market decisions
Organize regional, category, competitive, and trend evidence for market entry and priority.
Cross-border e-commerce
Connect products, traffic, ads, orders, refunds, and inventory to locate margin and fulfillment issues.
Customer segmentation
Use authorized data to distinguish acquisition, payment, repeat, churn, and value patterns.
Channel and campaign evaluation
Align cost, lead, conversion, and revenue definitions to compare channel quality.
Operating review and alerts
Create periodic reports and anomaly signals with a clear owner and next check.
Multi-table processing
Resolve inconsistent fields, duplicate entry, and manual consolidation with reusable workflows.
Analysis process
Keep the evidence chain from problem definition to action
Charts are one output. Definitions, sources, quality, transformations, and limits are part of the delivery too.
- 01Output: analysis question
Decision and scope
Confirm the problem, subject, time, user, intended action, and non-negotiable boundaries.
- 02Output: data inventory
Sources and quality
Review authorization, fields, missing values, duplicates, anomalies, gaps, bias, and conflicts.
- 03Output: metric dictionary
Definitions and model
Define formulas, status, deduplication, time fields, relationships, and transformations.
- 04Output: findings and tools
Analysis and presentation
Choose reports, charts, datasets, or dashboards and explain evidence, assumptions, and limits.
- 05Output: action plan
Validation and reuse
Review with operators, then define actions, refresh, monitoring ownership, and follow-up questions.
Data integrity
Do not hide poor data or present correlation as proven causation
Data conditions set the ceiling for conclusions. Unclear sources, quality, samples, or assumptions should reduce claim strength, not be hidden by polished visuals.
Sources and authorization
Public visibility is not unlimited collection permission; internal data also needs purpose, minimization, redaction, and retention rules.
Quality transparency
Missing, duplicate, anomalous, biased, and conflicting data is recorded with its effect on results.
Conclusion boundaries
Separate description, correlation, explanation, and prediction instead of presenting uncertainty as a guarantee.
Reproduction and upkeep
Retain rules or scripts for key calculations and define refresh, failure alerts, and ownership.
Deliverables
Move from data evidence to operating action
Deliverables follow the question frequency and team workflow; a complex dashboard is never the automatic default.
01Data inventory and quality notes
02Metric dictionary and definitions
03Cleaned dataset
04Reusable scripts or templates
05Report, charts, or operating dashboard
06Refresh process, actions, and limitations