What is the Growth Loop Analysis?

The Growth Loop Analysis helps you understand whether your system compounds — and why. Instead of treating growth as disconnected activities, it quantifies the feedback loop between connected metrics so you can see where momentum accelerates, stalls, or decays over time.

This is not a promise of future growth. It’s a diagnostic lens that shows where energy compounds and where it leaks — so you can strengthen the loops that power sustainable growth.

What Value It Gives

• System visibility – See how metrics connect and influence each other through feedback, not just isolated cause-and-effect guesses.
• Leverage identification – Find which metrics have the highest leverage inside the loop so you can strengthen compounding, not just improve a local metric.
• Momentum diagnosis – Understand whether your growth is self-reinforcing or whether it depends on constant external input to keep moving.
• Bottleneck detection – Identify weak links where energy leaks out of the system and where the loop fails to “return value.”
• Better sequencing – Decide what to fix first to make the loop healthy before you scale spend, headcount, or velocity elsewhere.

Common Use Cases

• Growth feels linear – You’re investing effort but momentum doesn’t compound, and you want to know whether the system is capable of self-reinforcement.
• A loop isn’t closing – Referrals, retention, or reactivation aren’t feeding back into acquisition, so the loop never returns value.
• You need to pick where to push – Identify the best lever to strengthen the entire loop, not just one metric in isolation.
• You want to spot “leaks” – Find the weakest link that limits compounding (e.g., activation is strong but retention breaks the cycle).
• Teams disagree on the growth engine – Align on a shared view of what reinforces vs. what dampens growth, using the same loop definition.
• You’re scaling acquisition – Check whether downstream metrics can sustain the loop before you pour more into the top of funnel.

How Growth Loops Are Analyzed

Segflow treats your loop like a dynamic system: if one metric gets a “shock” (a temporary lift), how much of that lift comes back around the loop — and does it amplify or decay as it cycles.

• Direct effects capture immediate relationships between metrics.
• Indirect effects capture ripple effects that travel through the rest of the loop.
• A time-series model estimates how today’s values influence future values.
• The result is Loop Gain (G):
  • G > 1: the loop amplifies (compounding)
  • G ≈ 1: the loop sustains
  • G < 1: the loop decays

When there isn’t enough data to fit the full time-series model, Segflow returns a more conservative, correlation-based estimate and marks the analysis status accordingly.

Under the hood (methodology)

1. Standardise the time series
   Segflow aligns your loop metrics and normalises them so different scales don’t distort relationships.

2. Model how influence flows through the loop
   Segflow builds a transition view of “what tends to lead what” using lag‑1 relationships, and preserves whether each connection is reinforcing (+) or dampening (−).

3. Estimate loop gain with a time-series model
   When data supports it, Segflow fits a Vector Autoregression (VAR) model and computes loop gain from the system’s amplification factor (spectral radius).

4. Fall back safely when data is limited
   If there isn’t enough history for VAR, Segflow uses a correlation-based approximation and labels the result as fallback so you treat it as directional.

5. Quantify uncertainty
   When VAR succeeds, Segflow estimates uncertainty via bootstrap simulation and reports a confidence interval for loop gain.

6. Surface “where to act” signals

   • Sensitivity highlights which metrics give the biggest system-wide leverage.
   • Steady state highlights which metrics are most central in the loop’s long-run dynamics.
   • Lagged effects highlight delayed relationships when the model can detect them.

Interpreting the loop report

Think of the Growth Loop report as having two primary attributes: “how strong” and “how reliable.”

• Loop Gain = how strong the loop is (compounding vs. decaying)
• Status + uncertainty = how reliable the estimate is (full model vs. fallback, and how wide the interval is)

Then use sensitivity, steady state, and (optional) lagged effects to decide where to intervene and what timing to expect.

Example: Fixing a Referral Loop That Doesn’t Compound (Illustrative)

Scenario: You built a referral feature, but growth still feels linear. You suspect the loop isn’t actually closing.

How you’d use Growth Loop Analysis:

1. Make sure the loop is represented on your Board (e.g., New Users → Activation → Referrals → New Users).
2. Run Growth Loop Analysis from any metric in the loop.
3. Review loop gain, then use direct/indirect effects and sensitivity to identify where energy leaks.

What you might learn:

• Loop Gain is below 1 → the loop decays, so the system can’t sustain compounding on its own yet.
• Sensitivity is highest for Activation → improving early product value has the biggest system-wide leverage.
• Direct effects show a weak Referrals → New Users link → the loop “breaks” at the return step.

What you do next:

• Prioritise bets that strengthen the weakest link (e.g., improve referral incentives, sharing surfaces, or activation-to-referral prompts).
• Use Bet Impact Analysis to compare fixes by effort vs expected lift.
• Re-run Growth Loop Analysis after shipping to see if loop gain moved toward (or above) 1.

What You Provide

Required Inputs

• Loop metrics – Typically 3–5 metrics that define a real feedback loop (e.g., New Users → Activation → Engagement → Referrals → New Users).
• Loop connections – The loop must close (a cycle) on your Board so Segflow can detect it.
• Historical data – Clean, consistent time series for every metric in the loop (same cadence and enough history to detect timing).

Optional Inputs (Advanced)

• Lag depth – If you want to control how far back the analysis looks for delayed effects.

What You Get Back

Core Outputs

• Loop Gain – The headline signal for whether your loop compounds (> 1), sustains (≈ 1), or decays (< 1) — a quick read on whether momentum is self-reinforcing.
• Convergence Time – How long it takes for the loop to “settle” after changes, which helps set expectations for when interventions should show through the system.
• Steady State – Normalised importance weights (0–1) showing where influence concentrates over the long run (centrality), not a forecast of metric values.
• Sensitivity Index – Normalised leverage weights (0–1) showing which metrics give you the biggest system-wide effect when you improve them.

Relationship Outputs

• Direct Effects – Immediate influence between connected metrics in the loop.
• Indirect Effects – Ripple effects that travel through multiple links before returning.
• Feedback Signs – Whether each connection is reinforcing (+1) or dampening (-1).
• Transition Matrix – The full set of probabilistic relationships between all metrics in the loop.

Reliability Outputs

• Analysis status – success, reduced (less history → lower lag order), fallback (limited data), or failed.
• Uncertainty bands – Confidence intervals around loop gain (available when the time-series model fits successfully).
• Lagged effects – Time-delayed influences that pass significance checks (only available when VAR succeeds).

How to Interpret Results

• Loop Gain > 1 means compounding – Each cycle amplifies the previous one. Small inputs create disproportionate outputs over time.
• Loop Gain < 1 means decay – Energy leaks out faster than it returns. Growth requires constant external input to sustain.
• Loop Gain ≈ 1 means stability – The system maintains current levels but doesn’t naturally accelerate or decline.
• High sensitivity = high leverage – Metrics with high sensitivity scores are the best places to intervene.
• Steady state shows centrality – Where influence concentrates if current dynamics persist; it highlights which metrics dominate the loop’s long-run behaviour.
• Wide uncertainty bands mean caution – If intervals span both sides of 1.0, the loop direction is uncertain.
• Status matters – “Fallback” means limited data for dynamic modelling; treat results as directional.
• Convergence time is about stability – If convergence is very large or infinite, the system may not settle under current dynamics (either compounding or too noisy to estimate cleanly).
• Direct vs indirect effects tell you where change spreads – Strong indirect effects mean improvements can ripple through the loop; weak indirect effects often signal the loop isn’t really connected.

Limitations to note:

• Results illustrate patterns, not precise predictions.
• Seasonal effects or one-off campaigns can temporarily distort results.
• Artificial or incomplete loops will show weak or misleading signals.
• All relationships are correlation-based, not causal.

Best Practices

• Start with 3–4 core metrics that capture your system’s feedback structure. Add complexity only after validating basic relationships.
• Ensure all metrics are consistently tracked and free from major data breaks.
• Make sure the loop actually closes — a one-way chain isn’t a loop.
• Revisit your loop structure as you learn which connections matter most.
• Treat loop gain, convergence time, and sensitivity as directional signals, not exact predictions.
• Use this analysis to identify where to push, then validate with experiments.

Summary

The Growth Loop Analysis reveals the invisible feedback mechanics that determine whether your efforts compound or dissipate. Use it to understand momentum, identify leverage points, and strengthen the loops that power sustainable growth. Once you see where the loop leaks or where leverage is highest, you can prioritise targeted bets to fix the weakest link — and re-run the analysis to see whether the loop actually got stronger.