Note · Methodology

How I estimate payback in SGD

The arithmetic behind every SGD figure on my homepage, and why the honest ones are labelled as estimates, not stated as measurements.

Every project card on my homepage carries a yearly SGD figure. None of them are made up, but most of them are estimates, and I think the difference deserves a note of its own.

Measured vs estimated

Some numbers on this site are counts, not guesses. 135 combined tests guard the Retail Shift Scheduler. Shipyard IMS handles roughly 82,000 inventory allocations in a browser tab. Daily Market Reports has posted 180+ market briefs across more than 90 consecutive days. These are things I can point at directly: a test runner's summary line, a row count in the data model, an archive of messages. I state them flat, with no label attached, because there's nothing to estimate.

Money is a different category. Turning "this tool saves time" into "this is worth SGD X a year" needs an assumption about what an hour of that time is worth, and an assumption about how much of the old workflow the new tool actually replaces. Both are judgment calls, even reasonable ones. So every SGD figure on the homepage and in the case studies carries the label est., and I try to keep the arithmetic behind it visible enough that you can check my working.

The arithmetic

The rate basis I use throughout is SGD 35/hr, over roughly 50 working weeks a year: 52 weeks minus a couple for leave and public holidays. It's a flat, conservative number, not tuned per project to make any single figure look better.

Take the Retail Shift Scheduler. Before the tool, a store manager spent an afternoon most weeks wrangling the roster by hand, matching coverage rules, chasing leave requests, checking the numbers twice. Call that roughly 3 hours a week. After the tool, the same job takes minutes: the rules enforce themselves and payroll export falls out the other end. The arithmetic is 3 hours x SGD 35/hr x 50 weeks = SGD 5,250 a year, per store. A chain running the tool across five stores clears SGD 26k+ a year, because the saving is per store and compounds with headcount, not with cleverness.

What I don't count

I only fold in avoided licence fees where a real quote existed to check against. Shipyard IMS is the clearest example: est. SGD 15k-30k/yr all-in combines the hours a costing clerk gets back, an ERP licence the shipyard didn't have to buy at a market rate I could actually verify, and the costing errors the tool catches before they reach a client invoice. That range is wide on purpose. I'd rather give a bracket I can defend than a single number I can't.

What doesn't go in the number: speculative "productivity gains", the soft stuff like morale or fewer arguments about the roster. Those might be real, but I have no way to price them defensibly, so they stay out of the arithmetic entirely instead of getting smuggled in as padding.

Why bother with any of this

The audience for these numbers is an SME owner or an operations manager, someone who can and should redo the arithmetic themselves, on a napkin, against their own payroll. If they multiply their own hourly cost by the hours they actually lose to the chore and the number doesn't survive that napkin, that's useful information for both of us: it means I shouldn't build the tool. A defensible SGD 5,250 per store beats an impressive-sounding number nobody can check.