How do you calculate and communicate the return on investment for automation projects, and how do you prioritize which toil to eliminate first?

Question

Accepted Answer

Calculating automation ROI and prioritizing toil elimination effectively is critical for a senior SRE because engineering time is finite and not all toil is equally worth eliminating. A systematic approach ensures the team delivers maximum impact with limited resources.

The basic ROI calculation compares the ongoing cost of manual work against the one-time cost of automation. For ongoing cost, multiply the frequency of the task by the time per occurrence by the loaded cost of an engineer's time. A task performed 20 times per month taking 30 minutes each time costs 10 hours monthly. Over a year, that is 120 engineer-hours, which at typical loaded costs of 150 dollars per hour represents 18,000 dollars annually. If automation requires 80 hours of engineering time, roughly two weeks of focused work, the payback period is under five months with ongoing savings thereafter.

However, the simple calculation misses important factors. I enhance it with what I call the full-spectrum ROI that accounts for five dimensions beyond direct time savings. First, reliability improvement: manual processes introduce human error. If a manual task has a 5% error rate and each error causes a 30-minute incident, the risk-adjusted cost is significantly higher than the time alone. Second, speed improvement: automated processes often complete faster than manual ones, reducing mean time to resolution for operational tasks. Third, knowledge risk: if only one or two people know how to perform a manual task, that is a bus factor risk. Automation captures the knowledge in code. Fourth, scalability: manual toil scales linearly with growth while automation has near-zero marginal cost. A task that takes 10 hours monthly today might take 40 hours monthly in a year as the service grows. Fifth, team morale: toil reduction improves job satisfaction and retention, which has substantial but hard-to-quantify value.

For prioritization, I use a scoring matrix that evaluates each toil item across four criteria. Impact score measures time saved per month multiplied by the number of people affected. Effort score estimates the engineering time required for automation. Risk score captures the consequence of human error in the manual process. Growth score projects how the toil will scale with anticipated service growth. I combine these into a prioritization score and maintain a ranked backlog.

Quick wins, items with high impact and low effort, go first regardless of the scoring model. Eliminating a high-frequency task with a simple script in a day or two builds momentum and demonstrates value. These quick wins also build organizational support for larger automation investments.

I communicate ROI to leadership using language they understand: cost savings, risk reduction, and capacity creation. Rather than saying we reduced toil by 200 hours per quarter, I frame it as we freed up the equivalent of a full-time engineer who can now work on reliability improvements that reduce outage risk. I maintain a running dashboard showing cumulative time saved by automation projects, which creates a compelling visual story of SRE engineering impact.

One important caveat: I do not automate everything. If a task occurs once a quarter and takes 15 minutes, automating it may take 40 hours and the payback period extends to decades. I document these tasks in runbooks instead, which provides knowledge capture without engineering investment. The art of toil prioritization is knowing when automation is the right tool and when good documentation suffices.