How to actually measure workflow automation ROI (in rand, not hand-waves)

Every automation pitch you've heard has the same line: "this will save you time and money." Your CFO has heard it too, and is unimpressed. Here's the version that works — a formula you can run yourself, with assumptions that will survive a finance review.

The honest formula

For any workflow you're considering automating:

Annual savings (R) =
  hours_per_week
  × hourly_cost_inclusive_of_overhead
  × 4.33 weeks/month
  × 12 months
  × automation_yield
  − ongoing_run_cost

Three of those variables get fudged by most consultancies. Let's pin them down.

Hourly cost inclusive of overhead

Not the salary. The fully-loaded cost: salary + UIF + COIDA + medical aid contribution + leave provision + workspace + IT + management overhead. For an SA SME, the rule of thumb is salary × 1.4–1.7. A R25,000/month admin person costs roughly R210/hour fully-loaded, not R145.

Automation yield

The percentage of those hours that the automation actually replaces. We default to 65% in our ROI calculator because:

• Some edge cases will always need a human.
• The team will spend time monitoring and improving the automation.
• Reality is messier than the demo.

If a vendor quotes you 95% yield, ask them what happens at 70%.

Ongoing run cost

This is the trap. You'll pay for: hosting (R150–R900/month for n8n + Postgres on Hetzner), API calls (R100–R5,000/month depending on volume), monitoring (R0–R600/month), and a maintenance retainer (R0–R12,000/month). For a typical SA SME workflow, run cost lands at R1,500–R6,000/month.

A worked example

Three-person ops team, R220/hour fully-loaded, 14 hours/week each on repetitive work, 65% automation yield, R3,500/month run cost:

3 × 14 × 220 × 4.33 × 12 × 0.65 − (3500 × 12)
= 312,840 − 42,000
= R270,840 annual saving

If the build cost was R85,000, payback period is roughly 3.8 months. After year one, the ROI is 219%. Those are conservative SA numbers, not US case-study fantasy. Try the calculator with your own variables.

What CFOs actually ask

When we present this in finance reviews, three questions come back every time:

1. "What if the team just does the automated work in less time anyway?" — Real risk. Mitigation: pair the build with a redeployment plan. Either the headcount-equivalent cost reduces, or the team takes on revenue-generating work that justifies it.
2. "What's the cost of the automation breaking?" — Add a downtime estimate to your formula. We typically model 99.5% uptime, which is roughly 3.6 hours/month of human fallback. At R220/hr × 3.6 hrs = R790/month buffer.
3. "What happens if the vendor disappears?" — Self-hosted, exported workflows, written runbooks. Lock-in is a finance risk, not just an engineering one.

The category most teams undercount

Hard rand savings (hours × cost) get all the attention, but two categories are often larger:

Cash unlocked. An invoice follow-up automation that reduces days-sales-outstanding from 52 to 31 unlocks roughly R1.4m of cash for a R15m revenue business. That's not a saving, but it's real money in the bank account today instead of in 21 days.

Risk avoided. An IoT alert that catches a freezer failure at 2am saves R220,000 of stock that no formula would have predicted. Build it for the median month, but pay back in the catastrophic month.

If you'd like a rand-by-rand audit of where automation would and wouldn't pay off in your specific business, see our two-week Automation Audit. Fixed price. Written 90-day plan.