Rasp logo

Rasp

Calculating ROI When You Can't Really Calculate ROI

•By Rasp Team

"What's the ROI of this feature?"

Every PM has been asked this question.

Here's the honest answer most of the time: I have no idea.

How do you calculate ROI when:

  • You don't know if anyone will use the feature
  • You can't directly measure revenue impact
  • The benefit is retention (which compounds over years)
  • It's strategic positioning, not immediate revenue
  • The feature enables other features down the line

The dirty secret of product management: Most ROI calculations are sophisticated guesses dressed up as analysis.

But you still need to make decisions.

Here's how to think about ROI when you can't really calculate it — and how to make better prioritization decisions anyway.

Why Traditional ROI Doesn't Work for Product

The formula seems simple:

ROI = (Benefit - Cost) / Cost × 100%

Easy, right?

Wrong. Because in product:

Problem 1: You Don't Know the Benefit

Example: Building better onboarding

Theoretical benefit: Higher activation rate → more retained users → more revenue

Actual benefit: ???

You can model it. But you're guessing:

  • Will activation actually improve?
  • By how much?
  • Will improved activation drive retention?
  • Will retention drive revenue?
  • Over what time period?

Each assumption multiplies the uncertainty.

Problem 2: You Don't Know the Full Cost

Engineering says "2 weeks."

Actual cost includes:

  • 2 weeks of development
  • Design time upfront
  • QA time
  • Bug fixes post-launch
  • Documentation
  • Support training
  • Ongoing maintenance
  • Opportunity cost (what you're NOT building)

Reality: That "2 week" feature costs 4 weeks and blocks other work.

Problem 3: Benefits Are Often Indirect

Direct benefit: Easy to measure (feature drives paid conversions)

Indirect benefit: Impossible to isolate

Examples:

  • Better UX → higher NPS → more word-of-mouth → easier sales
  • SSO → unlocks enterprise segment → enables expansion
  • Performance improvements → better experience → retention

The problem: These chains have 3-5 links. Each link has uncertainty.

Problem 4: Time Horizon Matters

6-month ROI vs. 3-year ROI are completely different calculations.

Example: Platform rebuild

  • Year 1: Negative ROI (pure cost, no benefit)
  • Year 2: Still negative (migrating, fixing bugs)
  • Year 3+: Massively positive (faster shipping, better product)

Question: What's the "ROI"? Depends entirely on your time horizon.

The Honest ROI Framework

Since perfect ROI calculations are impossible, use these approaches instead.

Approach 1: The Confidence-Adjusted Estimate

Calculate ROI, but add confidence levels.

Template:

Feature: [Name]

Expected Benefit:
- Best case: [Optimistic estimate] (10% probability)
- Likely case: [Realistic estimate] (70% probability)  
- Worst case: [Pessimistic estimate] (20% probability)

Expected Cost:
- Best case: [If everything goes smoothly] (20% probability)
- Likely case: [Realistic timeline] (60% probability)
- Worst case: [If things go wrong] (20% probability)

Weighted Expected ROI:
[(Benefit_best × 0.1) + (Benefit_likely × 0.7) + (Benefit_worst × 0.2)] / 
[(Cost_best × 0.2) + (Cost_likely × 0.6) + (Cost_worst × 0.2)]

Why this works: Acknowledges uncertainty. Forces realistic thinking.

Example:

Feature: SSO integration

Benefit:

  • Best case: Unlocks 10 enterprise deals ($500k ARR) — 10% chance
  • Likely case: Unlocks 3 deals ($150k ARR) — 70% chance
  • Worst case: Enables 1 deal ($50k ARR) — 20% chance

Weighted benefit: (500k × 0.1) + (150k × 0.7) + (50k × 0.2) = $165k

Cost:

  • Best case: 4 weeks ($40k engineering cost)
  • Likely case: 6 weeks ($60k)
  • Worst case: 10 weeks ($100k)

Weighted cost: (40k × 0.2) + (60k × 0.6) + (100k × 0.2) = $64k

ROI: ($165k - $64k) / $64k = 158%

Note: Still an estimate, but more honest than single-point calculation.

Approach 2: The "Table Stakes vs. Differentiator" Framework

Some features don't need positive ROI to be worth building.

Table stakes features:

  • Cost to NOT have: Lost deals, churn, reputation damage
  • ROI calculation: How much do we lose without it?

Differentiator features:

  • Value creation: New revenue, expansion, market position
  • ROI calculation: How much do we gain with it?

Example:

Table stakes: Security compliance (SOC 2)

  • Doesn't drive revenue directly
  • But without it: Can't sell to enterprise (lose $2M+ opportunity)
  • ROI: Infinite (prevents massive loss)

Differentiator: AI-powered insights

  • Could drive new revenue
  • But not required to keep current customers
  • ROI: Compare to other growth investments

Decision: Prioritize table stakes first (prevent loss), then differentiators (create gain).

Approach 3: The "Payback Period" Method

Instead of calculating ROI, ask: "How fast do we make our money back?"

Formula:

Payback Period = Investment Cost / (Monthly Benefit)

Example:

Feature: Automated onboarding

Cost: 8 weeks ($80k)

Benefit:

  • Saves 20 hours/week of manual onboarding
  • Manual onboarding costs: $50/hour
  • Savings: $1,000/week = $4,000/month

Payback: $80k / $4k = 20 months

Decision framework:

  • <6 months = Great investment
  • 6-12 months = Good investment
  • 12-24 months = Acceptable if strategic

  • 24 months = Questionable

Why this works: Easier to estimate than multi-year ROI. Focuses on cash flow.

Approach 4: The "Opportunity Cost" Lens

ROI isn't just "Is this profitable?" It's "Is this MORE profitable than alternatives?"

Template:

Option A: [Feature 1]
- Estimated benefit: [X]
- Cost: [Y]
- Rough ROI: [X/Y]

Option B: [Feature 2]  
- Estimated benefit: [A]
- Cost: [B]
- Rough ROI: [A/B]

Option C: [Do nothing]
- Benefit: $0
- Cost: $0
- ROI: 0%

Winner: [Option with highest ROI]

Example:

Option A: Build marketplace

  • Benefit: $500k ARR (maybe)
  • Cost: 6 months ($300k)
  • ROI: 67% year 1

Option B: Improve core product

  • Benefit: Reduce churn 10% = $200k retained ARR
  • Cost: 2 months ($100k)
  • ROI: 100% year 1 (and recurring)

Option C: Do nothing

  • Benefit: $0
  • Cost: $0
  • ROI: 0%

Decision: Option B (higher ROI, less risk, faster payback)

Why this works: Comparison matters more than absolute numbers.

Estimating the Unknowable Benefit

When you can't measure benefit directly, use proxies.

Proxy 1: Customer Lifetime Value (LTV) Impact

If feature improves retention:

Benefit = (Retention Improvement %) × (Total Customers) × (Avg LTV)

Example:

  • Feature improves D30 retention from 60% to 65% (+5%)
  • 1,000 customers
  • Avg LTV: $5,000
  • Benefit: 0.05 × 1,000 × $5,000 = $250k

Note: This is still an estimate (will retention actually improve?), but it's grounded in real metrics.

Proxy 2: Time Savings Valuation

If feature saves user time:

Benefit = (Time Saved per User) × (# Users) × (Value per Hour)

Example:

  • Feature saves 2 hours/week per user
  • 500 active users
  • Users value time at $100/hour (based on salary)
  • Annual benefit: 2hrs × 500 × $100 × 52 = $5.2M in user value

Note: This measures user benefit, not your revenue. But it helps justify investment.

Proxy 3: Conversion Rate Impact

If feature drives conversions:

Benefit = (Traffic) × (Conversion Lift %) × (Avg Deal Size)

Example:

  • 1,000 trial users/month
  • Feature improves trial-to-paid from 15% to 18% (+3%)
  • Avg deal: $2,000
  • Monthly benefit: 1,000 × 0.03 × $2,000 = $60k/month

Proxy 4: Churn Prevention

If feature prevents churn:

Benefit = (Churn Reduction %) × (At-Risk Customers) × (Avg ARR)

Example:

  • Feature prevents 20% of enterprise churn
  • 50 at-risk enterprise customers
  • Avg ARR: $50k
  • Benefit: 0.20 × 50 × $50k = $500k

When You Literally Can't Estimate

Sometimes there's no reasonable way to estimate.

Examples:

  • Brand new market — no data exists
  • Innovation bet — no comparable
  • Strategic moat — benefits are long-term and fuzzy

What to do: Use other decision frameworks.

Framework 1: Strategic Importance

Ask:

  • Does this support our core strategy?
  • Is this table stakes for our target market?
  • Does this create defensibility?

If yes to 2+: Build it even without clear ROI.

Framework 2: Learning Value

Ask:

  • Will this teach us something important?
  • Is the learning worth the cost?
  • Can we test cheaply first?

If yes: Treat it as research expense, not product investment.

Framework 3: Option Value

Ask:

  • Does this unlock future opportunities?
  • Is it a prerequisite for bigger bets?
  • Does it create strategic options?

If yes: Value the optionality, not just immediate benefit.

Example: Building a platform/API

  • Immediate ROI: Unclear
  • Option value: Enables partners, integrations, ecosystem
  • Decision: Worth it for the options it creates

The "Five Lenses" Decision Framework

When ROI is unclear, evaluate across multiple dimensions:

Lens 1: Revenue Impact

  • Direct revenue increase?
  • Enables expansion?
  • Prevents churn?

Score: 1-5

Lens 2: Strategic Alignment

  • Supports core strategy?
  • Moves us toward vision?
  • Builds competitive moat?

Score: 1-5

Lens 3: User Value

  • Solves real pain?
  • High-frequency use case?
  • Differentiated value?

Score: 1-5

Lens 4: Execution Confidence

  • Clear requirements?
  • Team has expertise?
  • Low technical risk?

Score: 1-5

Lens 5: Opportunity Cost

  • What are we NOT doing?
  • Is this best use of resources?
  • Better alternatives exist?

Score: 1-5 (inverted — low score = high opportunity cost)

Total Score: Sum / 25 × 100 = Percentage

Decision framework:

  • 80%+ = Definitely build
  • 60-80% = Probably build
  • 40-60% = Maybe (depends on capacity)
  • <40% = Probably skip

Example:

Feature: Dark mode

  • Revenue: 1/5 (no direct revenue)
  • Strategic: 2/5 (nice-to-have, not core)
  • User value: 4/5 (requested often)
  • Execution: 5/5 (easy to build)
  • Opportunity cost: 4/5 (small effort)

Total: 16/25 = 64% → Probably build (low cost, high user value)

Common ROI Mistakes

Mistake 1: Optimistic Bias

What it looks like: Only using best-case scenarios.

Example: "This will drive 50% more conversions!" (No it won't)

Fix: Force yourself to include worst-case in calculations.

Mistake 2: Ignoring Compounding

What it looks like: Only calculating first-year benefit.

Example: Retention improvement compounds over years. Year 1 benefit is just the start.

Fix: Model multi-year impact for retention/growth features.

Mistake 3: Forgetting Opportunity Cost

What it looks like: "This has positive ROI, so we should build it."

Example: ROI of 10% is good — unless you have options with 50% ROI.

Fix: Always compare against alternatives.

Mistake 4: Sunk Cost Fallacy

What it looks like: "We've already spent 3 months, we have to finish."

Example: Project is failing but you keep investing because you've already invested.

Fix: Evaluate based on future cost/benefit, not past investment.

Mistake 5: Confusing Activity with Outcome

What it looks like: Measuring feature usage, not business impact.

Example: "Feature has 40% adoption!" (But does it drive retention/revenue?)

Fix: Trace from activity → outcome → business metric.

Real-World Example: Breaking Down a Fuzzy ROI

Scenario:

Should we build an integration with Salesforce?

Traditional approach:

"Let's calculate ROI."

  • Benefit: ???
  • Cost: 8 weeks
  • ROI: Can't calculate, stuck

Better approach:

1. Confidence-Adjusted Estimate:

Benefit:

  • Best case: Unlocks 15 enterprise deals ($750k ARR)
  • Likely: 5 deals ($250k ARR)
  • Worst: 1 deal ($50k ARR)
  • Weighted: $287k

Cost:

  • Best: 6 weeks ($60k)
  • Likely: 8 weeks ($80k)
  • Worst: 12 weeks ($120k)
  • Weighted: $88k

ROI: ($287k - $88k) / $88k = 226%

2. Payback Period:

$88k cost / ($287k ÷ 12 months) = 3.7 months payback

3. Strategic Lens:

  • Revenue: 5/5 (enables enterprise)
  • Strategic: 5/5 (table stakes for enterprise)
  • User value: 4/5 (requested by target segment)
  • Execution: 3/5 (complex integration)
  • Opportunity cost: 3/5 (delays other work)

Score: 20/25 = 80%

4. Opportunity Cost:

Alternative: Improve core product for SMB

  • Benefit: $150k reduced churn
  • Cost: 8 weeks ($80k)
  • ROI: 88%

Comparison: Salesforce has higher ROI AND strategic importance.

Decision: Build Salesforce integration.

Why this worked: Multiple perspectives revealed same conclusion despite fuzzy numbers.

When to Skip the ROI Calculation Entirely

Sometimes it's not worth the effort.

Skip ROI when:

1. It's a Bug Fix

Just fix it. Don't calculate ROI on fixing broken things.

2. It's Compliance/Security

Not optional. Build it regardless of ROI.

3. It's Tiny

Decision cost > Implementation cost.

If it takes 2 hours to build and 3 hours to calculate ROI, just build it.

4. It's Strategic Imperative

Executive decision already made.

Don't waste time justifying what's already decided.

5. It's a Learning Experiment

Value is in learning, not ROI.

Run experiment, gather data, then decide on full build.

Final Thought

ROI calculations in product are mostly fiction.

You're predicting:

  • User behavior (unknown)
  • Market conditions (changing)
  • Implementation complexity (underestimated)
  • Second-order effects (invisible)

But you still need to make decisions.

So use frameworks that:

  • Acknowledge uncertainty
  • Compare alternatives
  • Balance multiple factors
  • Bias toward action (imperfect decision > no decision)

The best PMs don't have perfect ROI calculations.

They have better judgment about where to bet.

Build that judgment.

The math will follow.