What is the Leading Indicators Analysis?

Note: Leading Indicators Analysis is in development and may change.

The Leading Indicators Analysis helps you discover metrics that consistently move ahead of your target KPI — so you can spot shifts earlier and act sooner. It’s an early warning system: instead of waiting for a lagging KPI (like revenue) to react, you identify upstream signals (like activation or usage) that tend to change first.

Because it runs inside a Board, Segflow can focus on metrics that are already connected in your strategy graph — so results are more contextual than “random correlation hunting.”

This is not causation and it’s not a guarantee. It’s decision-support: use it to find reliable leading signals you can monitor, respond to, and validate over time.

What Value It Gives

• Earlier visibility – See KPI shifts coming before they show up in headline outcomes.
• Faster learning – Validate whether a bet is working weeks earlier by watching the right upstream metric.
• Proactive planning – Anticipate demand, churn, or pipeline changes and respond before outcomes degrade.
• Clearer focus – Reduce noise by highlighting the few signals most worth watching.
• Stronger alignment – Give teams a shared monitoring layer (“if this moves, we act”) instead of post-hoc explanations.

Common Use Cases

• Early bet validation – Launch an onboarding change and watch leading activation signals instead of waiting for revenue.
• Churn risk detection – Find early indicators that tend to precede retention drops.
• Revenue planning – Identify funnel or usage metrics that lead revenue and by how long.
• Operational monitoring – Spot reliability or support load issues before NPS or renewals move.
• Weekly business review – Build a “leading dashboard” that helps you manage forward, not backward.

How Leading Indicators Are Found

Segflow tests candidate metrics against your target KPI at multiple time lags, then ranks the metrics that look most predictive.

Under the hood (methodology)

1. Prepare aligned time series
   Segflow assumes your target and candidate metrics are aligned to the same cadence (daily/weekly/monthly) and time window.

2. Test lead times (cross-correlation)
   Segflow measures correlation across lags up to a maximum lead time (default: 12 periods).
   A lead time of 3 means: indicator[t−3] is associated with KPI[t].

3. Pick the best lag
   For each candidate, Segflow selects the lag with the strongest correlation that clears minimum thresholds (default: |correlation| ≥ 0.3 and statistically significant).

4. Validate predictive power (Granger causality)
   Segflow tests whether the candidate’s past values help predict the KPI beyond what the KPI’s own past predicts. This helps reduce “spurious” indicators that only share a trend.

5. Rank by a composite score
   Segflow calculates predictive power (0–1) using:

   • correlation strength (how strong the relationship is)
   • correlation p-value (how confident the relationship is)
   • Granger p-value (how confident the predictive test is)

6. Attach reliability and data-quality guardrails

   • reliable = true only when both tests meet significance thresholds
   • warnings highlight small samples, near-threshold correlations, or insufficient data for Granger testing

Interpreting the indicator table

Each indicator row has two primary attributes: “how far ahead” and “how confident.”

• Lead time = how far ahead the indicator tends to move (in your Board’s cadence)
• Predictive power = overall strength + confidence (higher is better)

Every indicator also includes:

• Correlation: how strongly the indicator and KPI move together at that lead time
• Direction: whether the relationship is positive or negative
• p-value: whether the correlation is statistically distinguishable from noise
• Granger p-value: whether the indicator adds predictive power beyond the KPI’s own history
• Reliable: true only when both p-values clear the significance threshold
• Data quality warnings: reasons to treat a result as directional (small sample, limited Granger support, near-threshold correlation)

Example: Getting an Early Signal for Revenue (Illustrative)

Scenario: Revenue is your north star KPI, but it reacts slowly. You want a 1–3 week early signal to manage proactively.

How you’d use Leading Indicators:

1. Set the target KPI to weekly revenue.
2. Add connected candidate metrics from your Board (e.g., trials started, activation rate, weekly active users, upgrade rate, support tickets).
3. Run the analysis and focus on indicators marked reliable, with clear lead times.

What you might do with the results:

• If activation rate leads revenue by 2 weeks, you make activation your early warning signal and create a playbook (what to check when it drops).
• If support tickets lead churn by 1 week, you treat spikes as a risk signal and prioritize reliability or support capacity before outcomes degrade.

What You Provide

Required Inputs

• Target KPI – The outcome you want to anticipate.
• Candidate metrics – A broad set of connected metrics on your Board (funnel, product usage, channels, support, quality).
• Historical time series – Enough history to test time-lagged relationships (15+ points minimum recommended; 50+ is much better for stable lead times).

Optional Inputs (Advanced)

• Max lead time – How far back to test for leading relationships (default: 12 periods).
• Minimum correlation + significance thresholds – How strict the analysis should be about what counts as “real.”

What You Get Back

Core Outputs

• Ranked leading indicators – Metrics sorted by predictive power.
• Lead time – How many periods the indicator leads the KPI.
• Predictive power (0–1) – Composite strength + confidence score for ranking.
• Direction – Whether indicator up tends to precede KPI up (positive) or down (negative).
• Reliability signal – Whether the indicator passes both correlation and Granger significance thresholds.
• Data quality warnings – Flags for small samples, near-threshold correlations, or insufficient Granger support.

Optional Outputs

• All lag results – The set of tested lead times and their correlations (useful when you want a different lead time than the single “best” one).
• Summary stats – How many candidates were tested, how many are reliable, and typical predictive power.

How to Interpret Results

• Start with reliable indicators – They passed both correlation and Granger significance thresholds.
• Use lead time as your reaction window – A 3‑week lead time is only useful if your team can respond within that window.
• Treat direction as meaning – Negative indicators can be the best early warning (e.g., rising support tickets precede falling retention).
• Use predictive power for ranking, not perfection – It’s comparative; it helps you choose what to watch first.
• Heed data quality warnings – Small samples and near-threshold correlations are often unstable; re-run as you collect more history.

Key considerations

• Correlation ≠ causation: Even with Granger validation, these are statistical associations.
• Lead times can drift: Seasonality, pricing changes, and product shifts can change what leads what.
• Re-validate periodically: A good indicator can decay as your system evolves.
• Don’t confuse “early signal” with “root cause”: Use Driver Analysis and experiments to find causality; use Leading Indicators to act sooner.

Best Practices

• Start with candidates that have a plausible causal story (Board connections help).
• Prefer stable, well-defined metrics with consistent cadence.
• Use indicators to drive playbooks (what you do when the indicator moves), not just dashboards.
• Re-run regularly and retire stale indicators.
• Pair with Bet Impact and Driver Analysis to turn early signals into the right bets.

Summary

The Leading Indicators Analysis turns your Board into an early warning system. By surfacing credible signals that tend to move before your KPI, it helps you act sooner, validate bets earlier, and manage with foresight rather than hindsight.