Scenario and Sensitivity Analysis Questions

Approach summary: build a 24-month cohort-based unit-economics model comparing Path A (↑marketing spend 25% for 6 months to defend share) vs Path B (same spend reallocated to retention features → +5ppt 12‑month retention). Compute CAC, LTV, payback, and margin impact under both paths and run sensitivity analysis.

Key inputs (per new-acquisition cohort and existing base):

• CAC (current): $X
• ARPU (monthly): $R
• Gross margin on revenue: m (e.g., 70%)
• Baseline monthly churn: c (or retention r = 1−c)
• Incremental marketing spend: ΔM = 25% * current marketing budget for 6 months
• Retention lift (Path B): +5 percentage points at 12-month mark (model as improved monthly churn schedule)
• Time horizon: 24 months; discount rate: d (optional for NPV)

Model structure / formulas:

• Cohort revenue at month t: Rev_t = ARPU * active_users_t
• active_users_t = cohort_size * Π(1 − churn_i) for i ≤ t
• LTV (gross) = Σ_{t=1..T} Rev_t * m
• LTV:CAC = LTV / (CAC + allocated incremental spend per acquired user)
• Payback months = months until cumulative gross contribution ≥ CAC

Assumptions (explicit):

• Marketing efficiency: Path A yields X% lift in acquisition volume (estimate from past campaigns); assume CAC constant or improves/worsens by Y%
• Retention improvement shape: either immediate +5ppt at month 12 onward or a ramp (e.g., linear 3–12 months); specify which
• No major product/market changes; ARPU stable; margins unchanged except development costs capitalized
• Allocate ΔM to acquisition in Path A or to product dev & rollout in Path B; amortize product build cost over N months to per-user basis

Sensitivity test plan:

• Vary acquisition lift for Path A: low/medium/high (e.g., −10% / baseline / +20%)
• Vary CAC elasticity: CAC increases/decreases with higher spend
• Vary retention lift for Path B: +2ppt, +5ppt, +8ppt
• Test different churn baselines (±25%) and ARPU volatility (±10%)
• Vary time horizon (12/24/36 months) and discount rate (0%–10%)
  Report tornado charts and break-even thresholds (e.g., minimum retention lift for Path B to beat Path A).

Presentation to leadership:

• One-slide executive summary: recommendation, headline LTV:CAC, payback, margin impact, and risk
• Supporting slides: model inputs & assumptions, cohort LTV curves (Path A vs B), sensitivity charts, scenario table (best/likely/worst), and action plan (milestones, measurement plan, guardrails)
• Include clear asks: instrumentation to measure retention lift, A/B test plan, budget reallocation trigger points, and KPIs to monitor weekly/monthly.
• Conclude with decision criteria: prefer Path B if retention lift ≥ Xppt within Y months and implementation risk acceptable; else choose Path A for short-term share defense.

Start by clarifying goals: detect which inputs (and interactions) drive revenue variance, quantify their contributions, and reduce dimensions so stakeholders focus on actionable drivers.

Sensitivity & interaction techniques

• Sobol indices: global variance-based method that decomposes output variance into main effects and interaction (second-order, total-order). Gives precise shares of variance attributed to single parameters and parameter pairs/groups.
• Partial Rank Correlation Coefficient (PRCC): nonparametric method that estimates monotonic relationships while controlling for others; good for screening and robust to non-normality.
• Morris (elementary effects): cheap screening to detect non-linear and interaction-prone inputs when computation is limited.

Dimensionality reduction & model-agnostic explainers

• PCA: identify correlated input combinations; use principal components to group redundant features but interpret components before presenting to stakeholders.
• Regularized models (Lasso/ElasticNet) or tree-based feature importance and SHAP values: identify and quantify driver importance including interactions (SHAP interaction values).

Practical PM steps with limited data

1. Run a screening (Morris or PRCC) on plausible parameter ranges to shortlist ~5–7 important inputs.
2. If simulation/model is cheap: use Latin Hypercube Sampling then compute Sobol indices on a surrogate (Gaussian Process or LightGBM) to estimate interactions without excessive runs.
3. If data scarce: bootstrap existing data, use cross-validation, and prefer rank-based methods (PRCC) and simpler models for stability.
4. Use PCA to collapse highly correlated features, but map components back to original parameters for business interpretation.
5. Communicate results as a ranked driver list, pairwise interaction heatmap, and “what-if” scenarios for top drivers. Recommend concrete experiments or telemetry to collect data where uncertainty/impact is highest.

Outcome: prioritize measurement and product experiments on top drivers, reduce model complexity by grouping or removing low-impact inputs, and plan targeted data collection to refine interaction estimates.

Approach overview: build a discrete-time Monte Carlo model projecting monthly users, revenue, and churn over 36 months, then discount cash flows to compute NPV per run. Run many stochastic scenarios and summarize distributional outcomes.

Choice of distributions (justify and params):

• Adoption (new user acquisition per month): use a lognormal or beta-PERT depending on data. If you have a central estimate + bounds, use beta-PERT (min, mode, max) to capture skew; if multiplicative growth and heavy right tail, use lognormal (µ, σ fitted to market research).
• Churn (monthly retention rate): use a beta distribution (bounded 0–1) parameterized from historical cohorts or expert elicitation (alpha/beta from mean and variance). Beta captures uncertainty and correlation with tenure if modeling cohort churn curves.
• Price / ARPU: use a normal or lognormal depending on skew; if price changes come from planned pricing experiments, model price as a discrete distribution of scenarios (low/expected/high) or a triangular/beta-PERT around the forecast.
• Correlations: model sensible correlations (e.g., higher discounts → higher adoption but lower ARPU, churn may correlate with price). Implement via copula or correlated sampling (Cholesky on transforms).

Simulation details:

• Time step: monthly for granularity of churn and seasonality.
• Iterations: 10,000 runs for stable tail estimates (5,000 minimum; more if tail precision needed).
• For each run: sample parameter time series (respecting correlations), simulate cohorts/new users, apply churn curve, compute monthly revenue, subtract costs (fixed/variable), discount cash flows to present value → NPV.

Primary outputs:

• Distribution of NPV: median, mean, standard deviation.
• Percentile bands: 10th/90th (90% CI), 5th/95th for downside/upside.
• Probability NPV > 0 (project viability).
• Key metric time series: median and percentile bands for MRR/ARR, cumulative users, churn rate.
• Sensitivity / contribution analysis: tornado chart or regression/variance-based (e.g., Sobol) showing which inputs drive NPV variance.
• Scenario summaries: best-case, base, worst-case representative runs.

Presentation to non-technical executives:

• Start with headline: “Median NPV = $X; 90% CI = [$A, $B]; P(NPV>0)=Y%.”
• Use one-page visuals: (1) histogram of NPV with median and CI lines; (2) fan chart of revenue/users over time showing median and 10–90% bands; (3) tornado chart showing top drivers; (4) a simple set of scenarios with recommended actions (e.g., thresholds for go/no-go, pricing experiments).
• Explain intuitively: translate distributions into business outcomes (“There’s a 20% chance we lose money, primarily driven by adoption uncertainty. If adoption meets target, break-even in 18 months.”)
• Recommend next steps tied to levers: run targeted experiments to reduce uncertainty in top drivers (e.g., paid acquisition tests, pricing A/B), and re-run the model with updated priors.

Validation and governance:

• Backtest model where possible against similar product launches.
• Keep inputs documented and retrain/update monthly as new data arrives.
• Deliver an interactive dashboard (filters for assumptions) so execs can test “what-if” quickly.

Assumptions (transparent):

• Current ARR = $50M; monthly recurring revenue = $4.167M.
• Roadmap feature (F) scheduled month 0 drives adoption/expansion: incremental +6% ARR uplift over 12 months after launch (phased), improves gross retention by +2ppt and NPS by +8 points.
• Hiring freeze delays launch by 6 months; all other growth and churn drivers constant.
• Baseline churn = 8% annual (0.67% monthly); expansion MRR contribution modeled as % of existing MRR.
• Discounting ignored for simplicity; results rounded.

Modeling approach:

1. Build a monthly cohort revenue model for 12 months post-launch comparing On-time vs Delayed.
2. Apply uplift to ARR starting month 1 after launch, phased linearly over 12 months (0.5% monthly to reach 6%).
3. Translate retention improvement: reduce monthly churn rate proportional to 2ppt annual (from 0.67% to ~0.5% monthly).
4. Map NPS to churn elasticity: assume each +10 NPS points → -10% relative churn; so +8 NPS → -8% churn reduction.
5. Run scenarios: Base (on-time), Delay (6-month shift), Mitigated (delay + mitigation actions recover 75% of lost uplift).

Quantified outcomes (12-month view):

• Base (on-time): cumulative ARR after 12 months ≈ $50M + 6% uplift = $53.0M (+$3.0M incremental ARR).
• Delay: launch at month 6 → only 6 months of uplift in 12-month horizon (average half effect) → incremental ≈ $1.5M. Net lost ARR = $1.5M over year.
• Retention: Base reduces annual churn by 2ppt → from 8% to 6% → saves ~$1.0M in churned ARR. Delay removes that benefit for 6 months → ~ $0.5M additional churned ARR.
• NPS: Base +8 → measurable CSAT/expansion gains; delay defers NPS gains by 6 months harming referrals and sales efficiency (qualitative impact; model as 50% of expansion loss).
• Combined financial impact (ARR lost + extra churn) ≈ $2.0M in year 1; longer-term compounding beyond year 1.

Mitigation recommendations (with estimated recovery):

1. Prioritize MVP scope of F for a minimal-engineer launch in 3 months using cross-functional “skeleton crew” — recovers ~50% of uplift (saves ~$0.75M).
2. Reallocate contractors or temporary agency hires for critical delivery — recovers additional 25% ($0.38M).
3. Accelerate go-to-market: targeted campaigns and customer pilots pre-launch to protect NPS/expansion — reduces churn impact by 30% ($0.15M).
4. Offer retention incentives to at-risk cohorts (discounts, concierge onboarding) — prevents ~$0.4M churn.
   Net under Mitigated scenario: recover ~75–90% of lost ARR; residual hit ≈ $0.3–0.5M.

Recommended next steps:

• Run the spreadsheet model with live MRR cohorts and sensitivity to NPS→churn elasticity.
• Present trade-offs to leadership: hire freeze cost vs. revenue risk; request approval for targeted contractors or scope reduction.
• Implement short-term retention programs and a customer communication plan to preserve NPS while launch is delayed.

Framework: build a 3-year P&L under three scenarios using monthly cohorts -> revenue, CAC, churn, ARPU, and costs (COGS, platform fees, marketing, R&D). Present headline assumptions, sensitivity ranges and decision triggers for product pivots.

Key baseline assumptions (year 0 = today):

• MAU start: 200k, organic growth 5%/mo, CAC $3, ARPU $1.50/mo, gross margin 70%.
• App Store commission: 30% on paid/subscription revenue, 15% for subscriptions after 1 year (per current rules).
• Churn 4%/mo, ARR retention improvements +1%/yr from product work.

Scenario A — Baseline (current fees/rules):

• Driver: steady discoverability, stable conversion (2% conversion to paid).
• 3yr outcome: revenue CAGR ~28%, operating margin positive by year 2.
• Sensitivity: ±20% ARPU or CAC changes swing profit by ±10–18%.
• Trigger for pivot: 15%+ sustained ARPU decline or CAC increase >25% — trigger price/packaging changes and push conversion-focused roadmap (free-to-paid funnel optimizations, trial tweaks).

Scenario B — Increased commissions (e.g., 40%):

• Key change: take-home revenue per transaction drops 10–15%.
• 3yr outcome: revenue growth slower, break-even delayed 6–9 months; gross margin compressed.
• Sensitivity: commission elasticity — if conversion and retention drop >5% due to price pass-through, NPVs fall sharply.
• Triggers: if commission hike reduces EBIT margin below target (e.g., <10%) or CAC payback >12 months -> initiate alternative monetization (ad monetization, lower price + upsell, web-checkout for subscriptions where allowed) and cost rationalization (deprioritize non-core features).

Scenario C — Open third‑party stores (competition + fee arbitrage):

• Assumption: 30% of new installs move to third‑party stores with 0–10% fee; discovery fragmentation increases UA costs by 10%.
• 3yr outcome: potential revenue uplift if we redirect payments off-store (improved take rate), but marketing complexity raises CAC; net benefit depends on ability to implement off‑store flows.
• Sensitivity: adoption of third‑party stores (10–40%) and off-platform conversion rate (50–90%) determine upside.
• Triggers: if off‑store take rate improves net margin by >8% and off-platform conversion >60% -> invest in off-store checkout, build web account linking, and change App Store-only features. Conversely, if fragmentation increases CAC >30% with <5% fee savings -> focus on in-app retention and premium feature expansion.

Decision framework (common):

• Leading KPIs to monitor weekly: ARPU, CAC, CAC payback months, gross margin on in-app revenue, paid conversion, churn.
• Monthly deep-dive triggers: 10% movement in any KPI vs forecast -> product prioritization meeting; 25% movement -> execute pivot playbook (pricing, checkout engineering, marketing channel reallocation).
• Tactical playbook items: implement web checkout, experiment with in-app promotions, negotiate platform programs, accelerate high-LTV features (educational content, enterprise tier).

This plan balances short-term mitigations (pricing, ads) with medium-term bets (off‑platform flows, product-led retention) tied to measurable triggers.