"What's the ROI going to be?" is nearly impossible to answer honestly after a project has already started. It has to be answered, and measured against, before.
The question that actually matters
Not "what will this save us," but "what are we measuring today, on day one, before anything changes?" Response time to a new lead. Conversion rate from inquiry to booked. No-show rate. Hours per week spent on a specific manual task. Revenue from a specific channel. Whatever is relevant to the initiative, written down, dated, and frozen before a single prototype gets built.
Why memory doesn't work as a baseline
Ask any team how things were "before," and you'll get an answer shaped by recency, by whoever's telling the story, and usually by whoever benefits from the comparison looking favorable. Memory is not a baseline. A number written down and dated, before the work started, is.
The discipline, in practice
At kickoff, we write down the specific metrics that define success for that engagement, with the date attached. At a fixed interval, typically 90 days, we measure the same metrics again, the same way, using the same definitions. If a metric improved, we can show exactly how much and trace it to what changed. If it didn't, that's the answer, communicated the same way a positive result would be.
Nowhere to hide, including for us
This discipline cuts both ways. It's also how we hold ourselves accountable: if we can't point to a dated baseline and a dated remeasurement, we don't get to claim the credit either. An ROI claim without a frozen baseline is a story. Ask for the baseline before you ask for the projection, from anyone proposing to build you anything.