Questioning the ROI of Automation

As a software engineer, few things give me more joy than kicking off an automation script and watching it accomplish 4 weeks of manual work in a matter of seconds. The value is seemingly self-evident, having “saved” nearly 160 hours of manual effort. But is that an honest assessment? Maybe it took me 2 weeks (80 hours) to develop and test the script for all the different scenarios. Before that were the hours of meetings and communications with me, a business stakeholder, and a business analyst (12 person hours) to understand the process to be automated. Following the execution of my automation, QA is going to take a couple days (16 hours) to validate the results and report on any issues. And, barring perfection, there will be the inevitable re-runs (?? hours) to correct bugs in the initial attempt. Adding it all up, did my automation really result in ROI?

Whenever we’re staring at a task that looks like a good deal of manual effort, the allure of automation is understandable. Before we go down that path, however, we must consider if that effort would be positive ROI, or if development resources would be better deployed elsewhere. Even if a tedious manual process might take 160 hours and automation is estimated to take 80, we might be comparing 160 hours of less allocated (and possibly lower cost) team members against the opportunity cost of tying up development resources for 80 hours. The choice that will maximize ROI is not cut and dry.

Below are some guidelines to help you calculate whether automation is “worth it” when a task can otherwise be handled manually.

Automation with Questionable ROI

• One-time jobs that take ~50% or more time to automate – the added cost of development, risk of misestimation, and likelihood of re-runs due to edge cases and missed requirements can easily lead to a breakeven or negative ROI.

• Highly repeated, short manual tasks – repeatedly doing a 5-minute set of manual steps may feel wasteful, but if it takes even only a single day (8 hours) of effort to automate, the breakeven on time savings comes only after executing the process 96 times. If it takes a full week to automate (40 hours) the breakeven is 480 times doing the process manually. That’s a lot of 5 minutes-es that need to be saved!

• Consolidation to remediate a manual but minor copy and paste – the inner perfectionist programmer in me can barely stand to admit this one, but sometimes it’s okay to live in an imperfect technology world when the worst-case scenario is that someone is going to have to spend 10 minutes updating 5 pages of a website instead of 1.

• Asset migration across DAM platforms for < 10k assets – mindlessly downloading a couple thousand files from one system and uploading into another may feel like an obvious candidate for automation, but when it can be done by any available team member time with low risk of error it’s often cheaper and less headache to do it manually.

• Content migration across CMS platforms – this is a tricky one, as automated content migration can often provide leverage, especially in cases of highly repetitive content structures (e.g. articles). I’ve written an entire article on the nuances of CMS content migration automation in the past and suffice it to say that the answer is almost never “all” of the content.

Automation that Lends Well to ROI

• Manual efforts that are very large – the cost of automating 100 instances of a task generally costs the same as for automating 100,000 instances of the same task. A relatively fixed cost of automation paired with scaling returns naturally increases ROI as the size of a task increases.

• Efforts that are often repeated – the cost of automation is usually a one-time cost, so the more that automation is reused (e.g. a nightly process) the higher the ROI climbs. If the process to automate is perpetual, so is the ROI.

• Tasks that are highly error-prone – if automation of a 5-minute release process can potentially save hours of system rollbacks and restores, that time savings needs to be accounted for in your ROI calculations.

• Situations where inconsistency could lead to costly repercussions – having a marketing message fall temporarily out of sync across the site might result be a risk worth living with, but having the price for your flagship product fall out of sync across the site could lead to having to honor obsolete pricing after a significant cost increase. The ROI on automating common data must be taken case by case.

Be smart and data driven when you’re considering automation. If there isn’t a clear and obvious ROI before embarking on automation, chances are that plowing through the task manually, unglamorous as it may be, is likely your more prudent option.