Part 2: Defining the Right Metrics: Leading vs. Lagging KPIs
Series: Data as a North Star – Part 2 of 5
In live-service games, strong metrics help teams prioritize, evaluate impact, and stay aligned. But poor metrics—or unclear definitions—can derail learning and lead to false conclusions.
That’s why choosing the right metrics (and the right way to read them) is as important as the feature itself. And above all, you need specificity. Many teams get this wrong by choosing broad KPIs too early or skipping the process of mapping metrics to the smallest behavior they’re actually trying to shift.
Go Granular: Metrics Begin With Behavior
Metrics are not just abstract numbers — they are reflections of player behavior. If you want to influence outcomes, you need to start by defining the smallest, most specific action you expect to change.
Instead of asking, “Will this improve retention?”, ask:
- Will players open more chests?
- Will they complete the tutorial faster?
- Will they replay this battle node more often?
The smaller and more behaviorally grounded your metric is, the easier it is to validate your hypothesis and generate actionable insights.
🎯 Pro tip: Start by identifying the one action that, if it changed, would show your feature is working.
What is Metric Mapping?
Metric mapping is the process of identifying what success looks like, how you’ll measure it, and what could go wrong — before launch. It connects your hypothesis to real behavior.
A good metric map includes:
- Primary Metric: The smallest, specific behavior you're trying to shift.
- Guardrails: Metrics that ensure you’re not causing unintended harm (e.g., crash rate, churn).
- Evaluation Window: Timeframe over which results should be evaluated (e.g., D1–D3).
- Expected Direction & Magnitude: Are you expecting a lift or a drop? By how much?
✅ Example:Primary: % of players completing all 3 event battlesGuardrail: Average session length (ensure sessions aren’t bloated)Window: D2–D4Expectation: +20% vs. control
Understanding Metrics, Dimensions, and Cuts
Metrics only become useful when paired with the right context. That’s where dimensions and cuts come in.
| Term | Definition | Example |
|---|---|---|
| Metric | A quantitative measurement | D1 retention, ARPDAU |
| Dimension | A lens that segments the data | Country, platform, player cohort |
| Cut | A specific view of a metric using dimensions | D1 retention for Tier 1 Android users |
| Aggregation | Time-based grouping applied to the metric | Daily, weekly, cumulative, cohort |
🎯 Example: "% of players opening second chest by D3 in Tier 1 iOS" is a meaningful, actionable cut that directly reflects engagement.
Snapshot vs. Cumulative vs. Cohorted Views
| Type | Definition | Best For |
|---|---|---|
| Daily | Snapshot of behavior on a specific day | Short-term spikes, early feedback |
| Weekly | Aggregated behavior over 7 days | Mid-range trend analysis |
| Monthly | Aggregated behavior over a 30-day cycle | Long-term shifts or seasonality |
| Cumulative | Running total over time | Measuring full impact (e.g. revenue) |
| Cohorted | Groups players by start time | Tracking retention or funnel behavior |
🧠 Tip: Daily spikes can be misleading. Use weekly, monthly, and cohort views to understand longer-term patterns and behavior.
Leading vs. Lagging Metrics
One of the most common mistakes in game teams is relying only on lagging metrics — like revenue or retention — which only move after behavior changes.
| Type | Definition | Example |
|---|---|---|
| Leading KPI | Early signals tied to specific behaviors | Tutorial completion, retry attempts, chest open rate |
| Lagging KPI | High-level outcomes influenced by many behaviors | D7 retention, ARPDAU, churn |
🎯 Use case: If your goal is to improve D7 retention, leading metrics like mid-session exits or event re-engagement are better early signals.
Great product teams identify the right leading indicators that predict movement in their lagging KPIs—and they root those indicators in granular, measurable player actions.
How Do You Know If You Chose the Right Metric?
Ask yourself:
- Does this metric reflect the specific behavior I’m trying to influence?
- Will it move before my broader KPIs (retention, revenue) change?
- Is it sensitive enough to show impact with small sample sizes?
- Can I explain the logic to a stakeholder in one sentence?
If you answer “no” to any of those, your metric may be too broad—or disconnected from your hypothesis.
Mini Case Study: When the Metric Was Wrong
🎮 Scenario: A team launched a new PvE event with a hypothesis that it would improve engagement by encouraging players to play more battles during off-peak hours.Original Metric: D1 retentionResult: No significant change, feature deemed unsuccessful.
🔍 What went wrong?
D1 retention didn’t directly reflect the behavior the team hoped to influence—participation in event battles. When they drilled into metrics like "average battles played in event per user" and "event participation rate," they saw a measurable lift. That increase also contributed to longer session lengths and higher D3 retention.
✅ Lesson: Always connect your metric to the core player behavior you're trying to shift. Lagging metrics like retention may mask early wins if you’re not granular enough in how you define success.
Final Thought
Metrics should help you learn, not just report. When you define them with granular behavior, layer in smart dimensions, and apply them in context, they become powerful tools for building games that grow and evolve with your players.
If Part 1 was about forming the right hypothesis, this post helps you measure the right outcome. Coming up next in Part 3: How to use A/B testing to validate those outcomes in the wild.