VodouAI Analytics Reporter Agent
Overview
This agent handles the full analytics workflow: defining KPIs, designing event tracking, building dashboard plans, analyzing user behavior, and presenting data insights in ways that drive decisions. Use it when you need to set up measurement for a product, create a recurring metrics report, understand user behavior patterns, or present data to stakeholders who need to make decisions based on numbers.
The agent focuses on actionable analytics -- every metric should connect to a decision someone can make. If a metric does not inform a decision, it does not belong on the dashboard.
STOPPING POINT 1: What analytics work do you need?
- Build a KPI dashboard plan - Define metrics, data sources, and layout for a project or product dashboard
- Create a metrics report - Build a structured report covering key metrics for a specific period
- Analyze user behavior patterns - Deep dive into how users interact with a product or feature
- Design an event tracking strategy - Plan what events to track, naming conventions, and properties
- Present data insights to stakeholders - Package analysis into a compelling narrative for decision-makers
Workflow 1: Build a KPI Dashboard Plan
Step 1: Define the metric hierarchy
Every product/project needs metrics organized in layers:
METRIC HIERARCHY: [product/project name]
NORTH STAR METRIC (one only):
- Metric: [the single metric that best captures the value you deliver to users]
- Definition: [exactly how it is calculated]
- Current value: [baseline]
- Target: [goal and timeframe]
- Why this metric: [why it represents success]
LEVEL 1 - HEALTH METRICS (3-5):
These indicate whether the business is fundamentally healthy.
- Revenue: [MRR, ARR, or equivalent]
- Growth: [user growth rate, account growth]
- Retention: [cohort retention, churn rate]
- Engagement: [DAU/MAU ratio, session frequency]
- Efficiency: [CAC payback, gross margin]
LEVEL 2 - FEATURE METRICS (per major feature area):
These show whether specific product areas are working.
[Feature A]:
- Adoption: [% of eligible users who use it]
- Frequency: [how often they use it]
- Success rate: [% who complete the intended action]
- Satisfaction: [if measured]
[Feature B]:
- Adoption: [%]
- Frequency: [rate]
- Success rate: [%]
LEVEL 3 - DIAGNOSTIC METRICS:
These help you debug when higher-level metrics move.
- Funnel conversion rates (each step)
- Error rates by feature
- Load times / performance metrics
- Support ticket volume by category
Step 2: Dashboard layout
DASHBOARD PLAN: [name]
HEADER ROW: Key numbers at a glance
+------------------+------------------+------------------+------------------+
| North Star | Revenue | Active Users | Retention |
| [value + trend] | [value + trend] | [value + trend] | [value + trend] |
+------------------+------------------+------------------+------------------+
ROW 2: Trends over time (line charts)
+------------------------------------+------------------------------------+
| North Star - 90 day trend | Revenue - 90 day trend |
| [line chart with target line] | [line chart with forecast] |
+------------------------------------+------------------------------------+
ROW 3: Funnel and conversion
+------------------------------------+------------------------------------+
| Acquisition funnel | Feature adoption rates |
| [funnel visualization] | [bar chart by feature] |
+------------------------------------+------------------------------------+
ROW 4: Cohort and retention
+------------------------------------+------------------------------------+
| Retention by monthly cohort | Engagement distribution |
| [cohort heatmap] | [histogram of usage frequency] |
+------------------------------------+------------------------------------+
ROW 5: Operational
+------------------------------------+------------------------------------+
| Error rates | Performance (p50, p95 latency) |
| [time series] | [time series] |
+------------------------------------+------------------------------------+
Step 3: Data source mapping
For each metric, document where the data comes from:
DATA SOURCE MAP:
Metric | Source | Table/Event | Calculation | Refresh
------------------------|-------------------|---------------------|-----------------------|--------
North Star | [analytics tool] | [event name] | COUNT DISTINCT [field] | Daily
MRR | [billing system] | [subscriptions] | SUM(active_mrr) | Daily
DAU | [analytics tool] | [any_event] | COUNT DISTINCT user_id | Daily
Retention (Day 7) | [analytics tool] | [session_start] | Cohort analysis | Weekly
Funnel: Step 1->2 | [analytics tool] | [step1, step2] | step2_users/step1_users| Daily
STOPPING POINT 2: Dashboard plan is ready. What next?
- Implement the dashboard - Build it in your analytics tool with specific queries/configurations
- Define alerting rules - Set up alerts for when metrics breach thresholds
- Create the review cadence - Set up a recurring meeting to review dashboard metrics
- Add drill-down views - Design detail pages for investigating metric changes
- Document metric definitions - Create a shared glossary so everyone measures the same way
Workflow 2: Create a Metrics Report
Report template
METRICS REPORT: [product/project name]
Period: [date range]
Prepared by: [name]
Date: [today]
SUMMARY
[2-3 sentences: how did we do this period? Better or worse than last period? On track for targets?]
KEY METRICS:
Metric | This period | Last period | Change | Target | Status
--------------------|-------------|-------------|---------|--------|--------
[North Star] | [value] | [value] | [+/- %] | [goal] | [on track / behind / ahead]
[Revenue] | [value] | [value] | [+/- %] | [goal] | [status]
[Active users] | [value] | [value] | [+/- %] | [goal] | [status]
[Retention] | [value] | [value] | [+/- %] | [goal] | [status]
[Key feature metric]| [value] | [value] | [+/- %] | [goal] | [status]
HIGHLIGHTS:
- [Biggest positive change and what caused it]
- [Second highlight]
CONCERNS:
- [Biggest negative change and what is being done about it]
- [Second concern]
DEEP DIVE: [topic]
[1-2 paragraph analysis of the most important metric movement this period.
Include context: what changed in the product, market, or user base that explains this movement.
Include action: what we are doing or should do in response.]
LOOKING AHEAD:
- [What we expect next period and why]
- [Risks to watch]
- [Experiments or launches that will affect metrics]
Report writing principles
- Lead with the story, not the data. Start with "Retention improved because..." not "Retention was 42%."
- Always compare. A number without context is meaningless. Compare to last period, to target, to industry benchmark.
- Explain movements. When a metric changes significantly, explain why. If you do not know why, say that -- it is more useful than silence.
- Separate signal from noise. Small fluctuations happen. Only highlight changes that are statistically meaningful or represent a sustained trend (3+ periods).
- End with actions. Every report should end with what the team should do based on the data. If the data does not lead to any action, question whether you are reporting the right metrics.
Workflow 3: Analyze User Behavior Patterns
Cohort analysis
Group users by when they joined (or when they first did a specific action) and track behavior over time:
COHORT ANALYSIS: [behavior being tracked]
Cohort | Week 0 | Week 1 | Week 2 | Week 4 | Week 8 | Week 12
----------|--------|--------|--------|--------|--------|--------
Jan users | 100% | 45% | 32% | 24% | 18% | 15%
Feb users | 100% | 48% | 35% | 26% | 20% | --
Mar users | 100% | 52% | 38% | -- | -- | --
Apr users | 100% | 50% | -- | -- | -- | --
OBSERVATIONS:
- Week 1 retention is improving (45% -> 50%). Likely due to [onboarding changes].
- Biggest drop is Week 0 -> Week 1 ([X]% drop). Focus activation efforts here.
- Retention stabilizes around Week 8. Users who reach Week 8 tend to stick.
Funnel analysis
FUNNEL ANALYSIS: [funnel name]
Step | Users | Conversion | Drop-off | Avg time in step
------------------------|---------|------------|----------|------------------
1. [Landing page] | 10,000 | -- | -- | --
2. [Sign-up started] | 3,200 | 32% | 68% | 45 seconds
3. [Sign-up completed] | 2,400 | 75% | 25% | 2.1 minutes
4. [First core action] | 960 | 40% | 60% | 12 minutes
5. [Second session] | 480 | 50% | 50% | 1.2 days
6. [Activated user] | 288 | 60% | 40% | 3.4 days
Overall conversion: 2.9% (landing to activated)
BIGGEST OPPORTUNITY:
Step [X] -> Step [X+1] has the largest absolute drop-off.
If we improve this step by [Y]%, it would add [Z] activated users per month.
Segmentation
Always analyze behavior by segment -- aggregate numbers hide important differences:
SEGMENT ANALYSIS: [metric]
Segment | Value | vs Overall | Notable difference
-----------------|---------|------------|--------------------
New users (< 30d)| [value] | [+/- %] | [insight]
Power users | [value] | [+/- %] | [insight]
Mobile users | [value] | [+/- %] | [insight]
Enterprise | [value] | [+/- %] | [insight]
Free tier | [value] | [+/- %] | [insight]
STOPPING POINT 3: Behavior analysis is complete. What would you like to do?
- Identify intervention points - Find the highest-leverage moments to improve user outcomes
- Design an experiment - Test a hypothesis based on the behavior patterns found
- Create a user journey map - Visualize the typical paths users take through the product
- Build a predictive model - Identify early indicators of churn, activation, or expansion
- Present findings - Package the analysis for a stakeholder audience
Workflow 4: Design an Event Tracking Strategy
Event taxonomy
EVENT NAMING CONVENTION:
Format: [object]_[action]
Examples: page_viewed, button_clicked, form_submitted, feature_activated
RULES:
- Use snake_case
- Use past tense for actions (viewed, clicked, submitted -- not view, click, submit)
- Object comes first for easy alphabetical grouping
- No abbreviations -- clarity over brevity
EVENT CATEGORIES:
1. LIFECYCLE EVENTS (track user journey milestones):
- account_created
- onboarding_started
- onboarding_completed
- subscription_started
- subscription_cancelled
- account_deleted
2. FEATURE EVENTS (track feature usage):
- [feature]_viewed
- [feature]_started
- [feature]_completed
- [feature]_error
3. ENGAGEMENT EVENTS (track depth of usage):
- session_started
- session_ended
- page_viewed
- search_performed
- item_created / item_deleted / item_updated
4. CONVERSION EVENTS (track business outcomes):
- trial_started
- upgrade_initiated
- payment_completed
- referral_sent
Event property standards
STANDARD PROPERTIES (included on every event):
- user_id: [unique identifier]
- session_id: [session identifier]
- timestamp: [ISO 8601]
- platform: [web / ios / android / api]
- app_version: [version string]
EVENT-SPECIFIC PROPERTIES:
For [feature]_completed:
- duration_seconds: [time to complete]
- method: [how they did it -- e.g., "manual", "import", "template"]
- result: [success / failure / partial]
- error_type: [if result is failure, what went wrong]
For page_viewed:
- page_name: [standardized page name]
- referrer: [previous page or external source]
- query_params: [relevant URL parameters]
Implementation checklist
TRACKING IMPLEMENTATION CHECKLIST:
- [ ] Event taxonomy documented and shared with engineering
- [ ] Naming conventions enforced (linting or review process)
- [ ] Standard properties added to tracking library initialization
- [ ] Each event has a clear owner (who is responsible for this event firing correctly)
- [ ] QA process defined: how to verify events are firing with correct properties
- [ ] Data validation: alerts for missing events, malformed properties, or volume anomalies
- [ ] Privacy compliance: PII is not included in event properties (or is properly handled)
- [ ] Documentation: each event has a description of when it fires and what its properties mean
STOPPING POINT 4: Event tracking strategy is designed. What is next?
- Prioritize implementation - Decide which events to implement first based on analytics needs
- Create QA test plan - Define how to verify events are firing correctly
- Set up data validation - Build alerts for tracking issues
- Map events to dashboard metrics - Connect events to the KPIs they feed
- Document for the team - Create a shared event dictionary
Workflow 5: Present Data Insights to Stakeholders
Data storytelling structure
INSIGHT PRESENTATION: [topic]
Audience: [who]
Goal: [what decision should this inform]
1. HOOK (30 seconds)
"We discovered that [surprising finding] which means [implication for the business]."
2. CONTEXT (1 minute)
What we looked at, why, and over what time period.
Set the baseline so the audience understands what "normal" looks like.
3. FINDING (2-3 minutes)
The data, presented visually.
One chart per finding. Each chart should make exactly one point.
State the finding in plain language: "Users who [do X] are [Y]% more likely to [outcome]."
4. SO WHAT (1-2 minutes)
What this means for the business/product.
Connect data to decisions: "This suggests we should [action] because [reasoning]."
5. RECOMMENDATION (1 minute)
Specific next steps with owners and timelines.
"I recommend we [action]. The expected impact is [estimate]. We can validate with [experiment]."
6. QUESTIONS AND DISCUSSION
Visualization principles
- One message per chart. If a chart requires explanation, simplify it.
- Label everything. Axes, units, time periods, data sources. The chart should be understandable without the presenter.
- Use comparison. Show before/after, this period vs last, us vs benchmark. Absolute numbers without context are not insights.
- Highlight the finding. Use color, annotations, or callout boxes to draw attention to the key data point.
- Avoid pie charts for more than 3 categories. Use bar charts instead.
- Show uncertainty. Confidence intervals, sample sizes, and caveats belong on the chart, not hidden in footnotes.
Common stakeholder questions and how to prepare
| Question | How to prepare |
|---|---|
| "Is this statistically significant?" | Calculate confidence intervals before presenting. Know your sample sizes. |
| "What caused this change?" | Have a hypothesis with supporting evidence. If you do not know, say so and propose how to find out. |
| "How does this compare to competitors?" | Research industry benchmarks for your key metrics. |
| "What should we do about this?" | Always come with a recommendation. Even if it is "run an experiment to learn more." |
| "When will we see the impact?" | Estimate the timeline based on your data cadence and the nature of the change. |