How to estimate ROI before you build an automation

Every operations leader we meet has a backlog of automation ideas. Almost none of them have been costed. That is the gap this guide is trying to close.

You do not need a spreadsheet or a finance partner. You need four numbers and ten minutes.

The four numbers

1. People touching it — how many people perform this task today?
2. Frequency — how many times per week does each person do it?
3. Time per instance — minutes (be honest, not optimistic).
4. Fully loaded cost — hourly cost per person, including benefits and overhead.

The formula

Monthly cost of the manual workflow:

people × frequency × minutes × 4.33 / 60 × hourly_rate

The 4.33 turns weekly into monthly. Multiply by twelve for annual.

Now apply a realistic recovery rate. We use 0.8 by default. Even great automations rarely reclaim 100% of manual time, because edge cases, handoffs, and supervision do not disappear; they get smaller.

An example

A sales ops team manually enriches inbound leads. Four reps do it, each handling fifteen leads a week, eight minutes per lead. Loaded cost: $55/hour.

4 × 15 × 8 × 4.33 / 60 × $55  =  ~$1,905 / month
                                  ~$22,860 / year

Recovery at 80% is about $18,300 a year. That number is now the cost ceiling for the build. If a vendor quotes you $40K to build it, the payback is roughly two years and one month — assuming nothing else changes. That is the call you can now make in your head instead of guessing.

Pitfalls we see constantly

Overcounting time

When you ask someone how long a task takes, you usually get a worst-case answer. The right way to measure is to time three real instances and take the median. Self-reported time is almost always 30–50% high.

Undercounting the supervision tail

An automation that runs in the background still needs review, especially in the first ninety days. Budget one to two hours per week of human oversight even after launch. Subtract it from your savings.

Ignoring the value of not hiring

Time-savings ROI is the floor, not the ceiling. The real prize is often avoided headcount — automating the work that would otherwise force you to hire another junior coordinator next quarter. That number is usually larger than the cost savings, and it should be in the model when it applies.

Counting hours that won't be reclaimed

If the saved time gets absorbed into Slack and meetings, you did not save money — you just changed how the same hours get spent. Be explicit about what the reclaimed capacity gets spent on, or treat the savings at half value.

What the model is good for

Stack-ranking. Run the same five-minute calculation on every idea in your backlog. Sort by annual savings. The top three are your roadmap. Everything below the line gets parked until evidence changes.

What it is not good for is making the case for strategic automations whose value is qualitative — error reduction, compliance, customer experience, knowledge capture. For those, ROI math is a sanity check, not the deciding vote.

Apply this to your own backlog

Pick the three workflows that consume the most attention right now. Run them through the formula. You will be surprised how often the right next move is to drop the one that feels most painful and build the boring one with a much bigger number behind it.