Data and Analytics

The Narqix view of data

A report is only useful if it changes a decision. Every dashboard you build should answer: where is the bottleneck, and is it getting better or worse? Lead with the answer, then show the numbers that prove it.

Definition of done

• The report or dashboard answers the questions listed in the Scope Sheet, and each question has a named owner who will use it
• Every metric has a written definition: formula, source, filters, time zone, refresh schedule
• Numbers reconcile to the source system within an agreed tolerance, with the check documented
• Refresh runs on schedule and alerts on failure
• A short guide explains how to read it and what to do when a number looks wrong
• The client accepted it in writing

Quality checklist

Data foundation

• Sources listed with owners and access method
• Tracking plan for events and conversions, agreed before implementation
• Raw data kept untouched; cleaning done in a separate, documented step
• Duplicates, test records and internal traffic excluded, with the rule written down
• Time zone and currency stated and consistent

Metrics

• One definition per metric across every report (a "lead" means the same thing everywhere)
• Metrics read in pairs where one can mislead alone (cost per lead with lead quality, volume with conversion rate)
• Every number shown with a comparison: previous period, same period last year, or target
• Small samples flagged; no conclusions drawn from a handful of records

Dashboard design

• Top row: the 3 to 5 numbers that matter most
• Clear titles that state the insight, not just the metric name
• Simple charts: lines for trends, bars for comparisons; no 3D, no unnecessary pie charts
• Filters for date, channel and owner where useful
• Last refresh time visible
• Loads quickly

Analysis

• Question, method, data used and limits stated up front
• Findings first, in plain words, then evidence
• Correlation not presented as cause
• Recommendations with expected impact and confidence level
• Work reproducible: queries and steps saved so someone else gets the same result

Pre-launch checklist

• Totals checked against the source system for at least two different periods
• Edge cases checked: zero values, missing data, a day with no activity
• Permissions set: each viewer sees only what they should
• Refresh schedule and failure alert tested
• Metric definitions published with the dashboard
• Walkthrough recorded for the client team

Reporting KPIs

KPI	Notes
Reports delivered on schedule	100 percent
Reconciliation difference to source	Within the agreed tolerance, usually zero for counts and money
Data freshness	Refresh succeeded on schedule
Data issues found by the client before you	Zero; you should find them first
Dashboard use	Who opens it and how often, where the tool shows it
Decisions or changes driven by analysis	Listed in the monthly review

Access and security rules

• Read-only access to source systems unless the Scope Sheet requires write access.
• Data warehouses, analytics properties and dashboards are owned by the client.
• Work with the minimum personal data needed. Use aggregated or anonymized data where you can.
• Never download customer-level data to personal devices or personal storage. Exports go to client-owned storage and are deleted from your device when the task ends.
• Share dashboards with named people, never with "anyone with the link" when they contain customer or financial data.
• Queries, credentials and connection details stay in the client's systems and the vault.

Common mistakes

Open the list

• Building a dashboard with 40 charts that answers nothing.
• Metric definitions that differ between reports.
• Not reconciling to the source, then losing trust the first time a number is wrong.
• Mixing time zones, so daily numbers never match.
• Reporting averages that hide what is really happening (one big deal skews the month).
• Presenting a pattern as proof of cause.
• Manual reports that depend on you, with no documentation.

Red flags to escalate

Red flag	Level	What to do
Customer personal or financial data exposed through a shared link or wrong permissions	P1	Remove access, alert Delivery Lead within 15 minutes
Numbers already shared with the client turn out to be wrong	P2	Tell the client the same day with the corrected figures and the cause
Tracking broken, so data for a period is missing	P2	Fix, note the gap on the dashboard, inform client
Request to change definitions or hide data to make results look better	P2	Decline, inform Delivery Lead
Source systems disagree and no one owns the truth	P3	Raise in the weekly report with a recommendation

---

© Narqix Academy