The Upskilling ROI Problem: How Do You Prove This to Your CFO?

General
1

Following the excellent discussions from @vp_eng_keisha, @eng_director_luis, and @product_david about upskilling vs hiring, I want to tackle the hardest part: How do you actually prove ROI to finance teams?

Everyone agrees upskilling “sounds good.” But CFOs don’t fund things that sound good. They fund things with measurable returns.

The Measurement Challenge

The problem with upskilling ROI is that benefits show up 12-24 months out, but CFOs want proof in 3-6 months.

How do you justify an investment when:

  • The velocity dip is immediate and visible
  • The retention benefit is delayed and hard to attribute
  • The cost savings are “avoided costs” (invisible)
  • The cultural impact is qualitative, not quantitative

This is the conversation I have with our CFO every quarter. Here’s how I approach it.

The Incomplete Metric: Cost-per-Hire vs Cost-to-Upskill

Most people start here:

  • Average cost-per-hire: $5,200 (recruiting, onboarding, ramp time)
  • Average cost-to-upskill: $3,800 (courses, mentorship time, reduced velocity)

Conclusion: Upskilling is cheaper.

But that’s incomplete. It only measures direct costs, not total impact.

My Framework: Total Cost of Talent (TCT)

Instead of Cost-per-Hire, I measure Total Cost of Talent, which includes:

1. Acquisition Cost

  • Hiring: Recruiting fees + interview time + offer negotiation
  • Upskilling: Course fees + platform subscriptions

2. Ramp Cost (Lost Productivity)

  • Hiring: 3-6 months @ 0-50% productivity = Lost output
  • Upskilling: 6 months @ 80% productivity on existing + 0-50% on new skills = Smaller lost output

3. Retention/Turnover Cost

  • Hiring: If they leave in 2 years, you pay acquisition cost again + knowledge loss
  • Upskilling: If retention improves 25-40%, you avoid repeated acquisition costs

4. Cultural Continuity Value

  • Hiring: New person needs to learn culture, norms, unwritten rules (hard to measure but real)
  • Upskilling: Person already embedded in culture (preserves institutional knowledge)

The Formula

TCT = Acquisition Cost + Ramp Cost + (Turnover Cost × Turnover Rate) - Cultural Continuity Value

This is messier than “cost per hire” but more accurate.

Real Example: Our Company (120 Engineers)

Let’s compare hiring 10 engineers vs upskilling 10 existing engineers:

Option A: Hire 10 Engineers

  • Acquisition: ,000 (,200 × 10)
  • Ramp cost (4 months @ 50% avg): ~,000 in lost productivity
  • Turnover (30% leave in 2 years): ,600 replacement cost
  • Total: ,600

Option B: Upskill 10 Engineers

  • Acquisition: ,000 (courses, platforms)
  • Ramp cost (6 months @ 80% productivity on existing work): ~,000 in reduced velocity
  • Retention improvement (retention goes from 70% to 90%): SAVINGS of ,200 in avoided turnover
  • Total: ,800

Net ROI: ,800 savings over 2 years

The Problem: Proving “What If” Scenarios

The hardest part is proving counterfactuals:

  • How do you prove that upskilling CAUSED higher retention vs other factors?
  • How do you prove that the upskilled engineer would have left if you hadn’t invested?

You can’t. You can only:

  1. Track cohorts: Compare retention of upskilled vs non-upskilled engineers
  2. Survey retention reasons: Ask people why they stayed (many will cite growth opportunities)
  3. Benchmark externally: Compare your retention to industry averages

What I Track Quarterly (Leading Indicators)

Because I can’t wait 2 years for retention data, I track leading indicators:

Q1-Q2 (Early Indicators)

  • % of engineers with active development plans
  • Hours per week protected for learning (target: 4 hours)
  • Engagement scores (“I have growth opportunities here”)

Q3-Q4 (Mid Indicators)

  • Internal mobility rate (engineers changing roles internally)
  • Promotion velocity (time from hire to first promotion)
  • Skills coverage (% of team with critical skills)

Year 2+ (Lagging Indicators)

  • Retention rate (upskilled vs non-upskilled cohorts)
  • Recruiting funnel (are people joining BECAUSE of learning culture?)
  • Avoided hiring costs (positions filled internally vs externally)

The Conversation with My CFO

I don’t lead with “upskilling is good for culture.” I lead with:

“Last year, we upskilled 15 engineers at a cost of ,000. If just 3 of them stayed who would have otherwise left, we saved ,600 in recruiting costs. If 5 stayed, we saved ,000. Our retention data shows 8 are still here 18 months later vs historical 4-5. ROI: 3.4x.”

Then I add the cultural piece as a bonus: “AND it signals to the team that we invest in growth, which drives engagement scores up 12%.”

My Question for the Community

How do you measure upskilling ROI in your organizations?

@eng_director_luis - You mentioned k savings. How did you calculate that, and how did you present it to leadership?

@vp_eng_keisha - At a startup, how do you justify upskilling investment when you don’t have 2 years of data?

@product_david - From a product lens, how would you measure the opportunity cost of delayed features due to upskilling velocity dip?

I’m always looking for better frameworks to make this case to finance teams.

2

@cto_michelle This is exactly the framework I needed. The Total Cost of Talent formula makes this so much clearer than just “cost per hire.”

How We Measured the $300k Savings

You asked how I calculated our savings. Here’s the breakdown:

Avoided Hiring Costs

  • Needed 8 engineers, hired only 5 (upskilled 3 positions worth of capacity)
  • 3 avoided hires × $5,200 average recruiting cost = $15,600
  • BUT the bigger cost: 3 avoided hires × 6 months ramp @ $150k salary = $225,000 in avoided low-productivity period

Retention Savings

  • Historical retention: 70% (meaning 30% leave within 2 years)
  • Post-upskilling retention: 95%
  • 12 upskilled engineers × 25% retention improvement × $5,200 replacement cost = $15,600

Increased Velocity (Year 2)

  • After upskilling, engineers could work on ANY part of the stack (no more “waiting for frontend help”)
  • Estimated 15% velocity gain from unblocking = ~$40,000 in additional output

Total: ~$296,200 (I rounded to $300k in my original post)

The Hard Part: What Did It COST?

Your TCT framework makes me realize I didn’t fully account for costs:

  • Courses/training: $12,000
  • Senior engineer mentorship time: 288 hours × $150/hr = $43,200
  • Reduced velocity during 6 months: ~$75,000

Total cost: ~$130,000

Net ROI: $166,200 (not $300k—I was only counting the savings, not the costs)

How I Presented to Leadership

I used a simple slide:

  • Problem: Need 8 more engineers. Market rate to hire = $52k recruiting + 6mo ramp
  • Solution: Upskill 12 internal engineers for adjacent capabilities
  • Cost: $130k (courses + mentorship + velocity dip)
  • Benefit: Fill 3 positions internally + improve retention 70% → 95%
  • ROI: $166k net savings + faster time-to-productivity (already know our domain)

The “already know our domain” part was the clincher. CFO understood that 6 months to learn React < 9 months to learn React + our financial domain.

3

@cto_michelle Your leading indicators framework is brilliant. This is what I need—metrics I can track in Q1-Q2, not wait 2 years for.

My Startup Challenge: No Historical Data

Here’s my problem: We’re Series B, scaling fast. I don’t have 2 years of retention data to prove upskilling works. How do I justify the investment NOW?

What I’m Thinking: Commitment Metrics vs Outcome Metrics

Your leading indicators gave me an idea. I can’t prove retention improvement yet, but I can prove COMMITMENT to upskilling:

Quarter 1-2 Commitment Metrics:

  • % engineers with development plans: Target 100%
  • Protected learning time: 4 hrs/week × 25 engineers = 100 hrs/week total
  • Learning budget utilization: $2k/engineer × 25 = $50k annual budget

If I can show we’re DOING upskilling seriously, that’s a leading indicator of future retention.

Quarter 3-4 Early Results:

  • Internal mobility: Are people moving into new roles internally?
  • Skill coverage: Do we now have 3+ people with critical skills (not just 1)?
  • Engagement scores: “I have growth opportunities here” trend

The Question I’ll Get from My Board

“If you invest $50k in upskilling and people leave anyway, didn’t you just train them for their next job?”

How do you respond to this? Because it IS a risk. @maya_builds is loyal because of upskilling, but is she the exception or the rule?

Framework I’m Building

Based on your TCT formula + my startup constraints:

Year 1 (Prove Commitment):

  • Invest in upskilling infrastructure (learning time, budget, mentorship)
  • Track leading indicators (engagement, internal mobility, skill coverage)
  • Target: 10-20% of team goes through structured upskilling

Year 2 (Prove ROI):

  • Compare retention: upskilled cohort vs non-upskilled cohort
  • Measure avoided hiring costs (positions filled internally)
  • Survey exit interviews: did people leave for growth opportunities elsewhere?

Year 3 (Cultural Advantage):

  • Recruiting pitch: “We invest in your growth” attracts better candidates
  • Retention becomes a recruiting advantage (people join BECAUSE of learning culture)

Does this framework make sense for a startup context?

4

@cto_michelle Your TCT framework is exactly what I need to make the business case. Let me add the product perspective on measuring opportunity cost.

Measuring the Cost of Delayed Features

You asked: “How would you measure the opportunity cost of delayed features due to upskilling velocity dip?”

Here’s how I’d approach it:

Step 1: Quantify the Velocity Dip

  • Historical velocity: 50 story points/sprint
  • Upskilling velocity: 42 story points/sprint (15% dip)
  • Duration: 6 months (12 sprints)
  • Total lost capacity: 96 story points

Step 2: Translate to Features

  • Average feature size: 13 story points
  • Features delayed: 96 ÷ 13 = ~7 features delayed by 3 months

Step 3: Revenue Impact of Delayed Features

This is the hard part. Not all features have equal revenue impact:

  • High-impact feature (key customer ask, prevents churn): $50k-$200k revenue impact if delayed 3 months
  • Medium-impact feature (nice-to-have, competitive parity): $10k-$50k
  • Low-impact feature (internal efficiency, tech debt): $0-$10k

So the opportunity cost depends on WHICH features got delayed.

The Comparison

Option A: Hire (No Velocity Dip):

  • Ship all 7 features on time
  • Revenue impact: Let’s say 2 high, 3 medium, 2 low = $140k-$450k
  • Cost: $52k hiring + $200k ramp = $252k
  • Net: -$112k to +$198k depending on feature value

Option B: Upskill (15% Velocity Dip):

  • Delay 7 features by 3 months
  • Revenue impact: Lost/delayed $140k-$450k over 3 months
  • Cost: $130k (upskilling)
  • Benefit: $166k savings (retention + avoided hiring) + long-term capability
  • Net: +$36k to -$284k depending on feature value

The Insight

If the delayed features are high-revenue impact, hiring wins short-term.
If the delayed features are low-medium revenue impact, upskilling wins long-term.

So the question becomes: Can you sequence your roadmap to upskill during periods when you’re working on lower-revenue features?

What I’d Recommend

  1. Track feature revenue impact (estimated, not perfect)
  2. Plan upskilling during low-impact cycles (tech debt, internal tools, refactoring)
  3. Hire for high-impact, time-sensitive capabilities
  4. Upskill for strategic, long-term capabilities

This way you minimize opportunity cost while still building the learning culture.

5

@cto_michelle I’m coming at this from a totally different angle, but I think there’s a metric missing from your TCT formula:

The Cost of NOT Upskilling (Cultural Degradation)

What I Mean

When companies ONLY hire externally and never upskill, here’s what happens over 2-3 years:

  • Engineers realize growth means leaving
  • People start looking externally at 12-18 months
  • Average tenure drops to 2 years
  • You’re constantly onboarding new people
  • Institutional knowledge is lost
  • Culture becomes transactional

This is hard to measure, but it’s REAL. I’ve seen it at my previous company.

How I’d Measure It (Qualitatively)

Question in exit interviews: “Why are you leaving?”

  • If 40%+ say “no growth opportunities here” → Cultural red flag
  • If 10-20% say “no growth opportunities” → Normal attrition

Question in engagement surveys: “Do you see a career path here?”

  • If <50% say yes → Cultural problem
  • If >80% say yes → Healthy growth culture

The ROI of Preventing Cultural Degradation

Let’s say your average engineer tenure is 2 years (industry average). That means:

  • You replace 50% of your team every 2 years
  • 25 engineers = 12-13 replacements every 2 years
  • 12 × $5,200 recruiting = $62,400

If upskilling increases tenure from 2 years to 3 years:

  • You replace 33% of your team every 2 years
  • 25 engineers = 8 replacements every 2 years
  • 8 × $5,200 recruiting = $41,600

Savings: $20,800 every 2 years just from reduced churn.

Plus the intangible: You keep people who know your product, your customers, your unwritten rules. That institutional knowledge is worth more than we account for.

My Answer to Keisha’s Board Question

“If you invest $50k in upskilling and people leave anyway, didn’t you just train them for their next job?”

My answer: “Yes, some will leave. But the alternative is EVERYONE leaves because we don’t invest in growth. Which is cheaper?”

Some data:

  • Companies with strong learning cultures: 30-50% higher retention (LinkedIn Learning report)
  • Companies that don’t invest in learning: Employees are 12× more likely to leave (Gallup)

You can’t guarantee every upskilled person stays. But you CAN guarantee that if you never invest in growth, people will leave for companies that do.