As a technical sales manager, I built Kansai Colourmatch mainly for myself, with the aim of eventually making it available to our sales and customer service teams so they would not need to wait for me. In the paint industry, we need to understand different colour standards and determine whether there is a close match in the Kansai range. There are tens of thousands of colours, and every standard has its own code-naming convention. Some codes give an indication of the colour, but others are difficult to interpret or remember. Some colour standards are also only available as physical colour charts and cannot be found online.
I would often receive enquiries from customers or sales asking me to check a special colour, either from a competitor brand or an international colour standard. I would reply with the closest equivalent colour in the Kansai range, after which the salesperson could do the necessary calculations and prepare the quotation. The faster a business responds to a lead, the higher the chance of landing the sale.
Most of the time it’s not our colour
The standard a customer brings is usually somebody else’s: a competitor’s paint code, a Pantone number, or a British Standard or RAL colour code. So the real question behind “can you make this” is which Kansai colour comes closest, and how close it actually is.
That’s what the details panel for each colour swatch answers. Search or select any colour in the database and it returns the nearest Kansai match with a ΔE figure, calculated with CIEDE2000. There are 13,725 records across popular paint brands, Munsell, Pantone, Kansai, BS 4800 and RAL. You can search by code, name, hex, RGB or Lab values.
Browsing colour records using searchable swatches and catalogue filters.
A BS 4800 colour record with its technical values and nearest Kansai matches.
Getting the colours in
Most of the colours came from colour data that was publicly available online. For each colour, I structured the information into values such as RGB, Lab, HEX and CMYK. I also recorded where the colour data came from, such as a website or a scan of a physical colour chart. The difficult part was adding colours that were only available on paper.
For those, I messaged my Hermes Agent, Optimush, who helped me build this app. The last one I sent was a scan of the Kanpe Epoxy Paint brochure, with the message “please process this pdf and add the colours to the database. this is the colour brochure for Kanpe Epoxy Paint”. No format, no column mapping, and the file was a JPG rather than a PDF. Hermes inspected the swatch grid, wrote an extraction script for that specific layout, and imported the result. I checked the output by hand — there weren’t many colours, so that was quick.
From a scanned brochure sent to Optimush to 29 colour records imported into Colourmatch.
Optimush can convert a scanned colour chart into colour swatches that can then be added to the database. The conversion takes minutes, compared with days if done manually.
I can also point Optimush at a website and have it scrape the colours. This way, I can obtain thousands of colour records, which I could never have done easily without Hermes Agent.
Stack and hosting
Next.js, TypeScript and Prisma, running on Cloudflare Workers with a D1 database.
The MVP took about a day to build, while the total build took around a week. Most of the remaining time was spent polishing the app and adding features.
Building custom software with Hermes
Hermes Agent can help anyone build custom software for their own needs. Because the agent built the software with me, I can continue using it like a teammate to add features and work with the app.
Optimush, my Hermes agent, can already query it and hand back the five closest Kansai alternatives with swatches, codes and ΔE scores.
I knew the paint problem. Optimush knew how to build the software. Colourmatch is what happened when those two kinds of knowledge met.