What we build, once we know what's wrong.
- You already know something is wrong. Naming it is the hard part, and it is why choosing from a list is a poor way to start: what you can see is rarely what is costing you. What follows is the whole range, in three groups, each one open about when it is the wrong thing to buy. None of it is a recommendation until we understand how the business actually runs. Then the diagnosis picks, not you.
A range, not a menu.
Sixteen things sit under the three groups below. Every one of them is something we have built and can build again, and not one of them is a recommendation yet. What decides the work is never the list. It is how your business actually runs: where the enquiry lands, who answers it and how fast, why the quote takes three days, what only one person can do.
The diagnosis picks, not the list
Nothing on this page is sold on its own. We look at how the work runs first, then say what is wrong, what to fix first, and in what order. Anything built after that is built against a written answer.
We say no in writing
Nine of these sixteen are listed further down with the reason not to buy them. The fastest way to waste a budget is to buy the thing you can see instead of the thing that is costing you.
One studio, start to handover
Over 100 projects across more than 70 clients since 2023. Strategy, design, build and the running of it afterwards, done senior throughout and not handed down the corridor.
This is where the expensive problems live. Software is what you build when the work itself is the problem: the quote that takes three days, the process that only one person can complete, the spreadsheet that has quietly become load bearing. It is also the category with the most ways to waste money, which is why none of it starts before a diagnosis.
The three we are asked for most sit first. The rest of the family is underneath, and every one of those lines opens: what moves the cost and, where it applies, when not to buy it at all.
Custom software
Software built around how one business already works, rather than a product you reshape the business to fit.
AI tools
Internal tools that use a model to do one narrow job inside your business.
Web apps
A system that lives in a browser and does real work, not a website with a login bolted onto it.
A product is a different business from a project, and the part that gets underestimated is not the build. It is the buyer, the support path, the billing, and who owns the thing on a Tuesday six months after launch. Start small enough to reach real users in weeks, because the first version exists to find out whether anyone wants it, not to be complete. If the idea cannot be said in one sentence, it is not a product yet. It is a list of features hoping to become one.
The honest starting position is that most businesses asking for an app want a good mobile site and would be better served by one. A real app earns itself when you need the camera, location, offline use somewhere with no signal, or notifications people actually act on. Field teams, drivers and site staff are the common case. What moves the cost is how many platforms you need and how the app behaves with no connection, plus app store review, which is a schedule nobody controls.
Useful when the same twenty questions arrive every week and the answers already exist in writing somewhere. It replaces a person retyping a reply they have sent a hundred times. It does not replace a person making a judgement, and anchoring it to your own documents is the difference between the two. The version we will not build is the one that hides the route to a human, because that is where most of them start costing more than they save.
The symptom is always the same. The office cannot say where a job stands without phoning someone, and the paperwork comes back three days later, if it comes back. It replaces a whiteboard, a chat group and a stack of job sheets, and the first thing it usually reveals is how much scheduling was being held in one person’s memory. Routing and constraints can be assisted by a model, but most of that work is ordinary engineering rather than AI. The constraint that decides whether this succeeds is not the software: if the technicians will not use it standing on a site with one hand free, it does not exist.
This is the one item on this page worth buying before anything gets built. A half day with the people who actually do the work usually surfaces two or three tasks worth automating and several that are not, which is a far cheaper way to learn that than commissioning software and finding out afterwards. It is consultancy rather than a course, so the value depends entirely on how specific your examples are. Arrive without real work in hand and you will get a demonstration that everyone forgets within a fortnight.
We turn down more software than we take.
Most of what arrives described as a software problem is a process nobody has written down, or a decision one person is holding in their head. Building around either one makes it permanent and expensive. Separating the two is what the first conversation is for, and it costs a great deal less than finding out after the build.
These are the things a business is judged on before anyone speaks to it. Most of what goes wrong here is neglect rather than taste: the site was built once, handed over, and never touched again. A great deal of what we found in our own audit was not ugly design at all. It was things that had quietly stopped working.
From our own review of 54 Singapore B2B sites, August 2026.
- 19 of 54 published no project record, portfolio or case studies at all.
- 12 of 54 were running an outdated or end of life CMS or plugin.
- 7 of 54 had a contact page or navigation link returning an error.
Websites
The public site, built for the thing it is actually used for. For most B2B firms that is a verification step rather than a sales channel, and the record of work is the page a buyer came looking for.
Infrastructure support
Hosting, updates, backups and monitoring after launch. The item most often cut, and 5 of the 54 sites we reviewed failed their certificate outright. Quoted once we have looked at what you have, never before.
Brand identity
The name, mark, typography, colour and the rules for using them, plus the language the business uses to describe itself.
Most companies do not have a design problem here so much as a consistency problem, where four people are building slides from four different templates and none of them match. We take this on inside a larger engagement rather than as a standing service. A design job with no strategy behind it is work a lot of people can do well, and we would be a slow way to buy it. If you have an identity you are happy with and you need hands, hire hands. If the material keeps drifting, the fix is the system behind it, not the next deck.
Six decisions, and who they were for.
None of these started as a brief for a website. Each one started with a question about who was reading and what they were deciding. The case studies say what we decided and why.
Towa Marine
Marine air conditioning and refrigeration. We opened the site on availability and reach rather than company history, because a superintendent with a vessel alongside is deciding whether a yard can take the job now.
Hong Hock Global
Water infrastructure. Built as a prequalification document rather than a brochure, with certifications and live contract scope stated early and emergency response given its own route.
THK Engineering
M and E contracting. Five disciplines presented as one continuous capability, because the only question on a shortlist is whether half the package ends up sublet to someone nobody has vetted.
U.S. Dairy Export Council
Regional hub for Southeast Asia. Two readers, a formulator researching ingredients and a partner booking a workshop, kept on visibly separate paths instead of sharing one journey.
Kinetics Sports Alliance
Badminton academy. Organised by level rather than by programme name, so a parent and a competitive player self-select in one step. The identity was built in the same engagement.
Actualytix
Reinsurance analytics. We decided against translating the vocabulary, because over-explaining to actuaries would have proved the opposite of what the site was there to establish.
Film, photography, social and paid work all do the same job. They point attention at a business. That makes them the last thing we do, because attention aimed at a business that has not been fixed converts worse than no attention at all: the visit confirms the doubt instead of removing it.
It earns its place when what you do is genuinely hard to describe in writing, which covers most industrial and technical work. One well-made film of the actual work will outlast a year of stock footage and captions. The limit is access. If we cannot get into the yard, onto the site or next to the machine, the result is a montage of material you already have, and that is not worth what it costs. What moves the cost is shoot days and locations, far more than the edit.
A surprising number of website problems that look like design are picture problems. There are no real photographs of the business, so the page fills up with stock, screenshots and whatever the last supplier had. We direct the shoot and brief the photographer rather than claiming a studio can be everything at once. Be warned that this is the line item most likely to delay a project, because it needs a day when the site is presentable, the people are available and the work is actually happening. It is also the one most often skipped, which is a large part of why so many B2B sites look like each other.
The useful version is narrow: publish the work, show the people, and keep the page from looking abandoned when a buyer checks it the evening after a meeting. Nobody in your market is deciding on the strength of a carousel, but a company page that stopped three years ago does real damage at exactly the moment you are being verified. We take this on inside a longer engagement rather than as a standalone retainer, because a channel with nothing behind it becomes a posting treadmill, and treadmills are how agencies bill without moving anything. If volume is the goal rather than the work being visible, we are the wrong studio for it.
Paid works when the offer is already proven and the constraint is volume. It does not fix an offer nobody wants, and it will establish that faster and more expensively than you would like. Attention aimed at a business that has not been fixed converts worse than no attention at all, because the visit confirms the doubt instead of removing it.
Which is why we would rather spend the first month making sure an enquiry has somewhere to land than spend it buying clicks.
Nine things we will talk you out of.
Not modesty. Every one of these is work we are equipped to sell you and would rather not, because the thing you can see is rarely the thing costing you money. If one of them is what you had in mind, it is worth reading before you brief anyone, us included.
- Custom software. If your process is genuinely standard, buy the standard product.
- AI tools. Most problems do not need AI, and we will say so before you spend anything.
- AI products. If the idea cannot be said in one sentence, it is not a product yet.
- Web apps. Everything you build here becomes something you have to operate.
- Mobile apps. Most firms asking for an app want a good mobile site.
- Chatbots. We will not build the one that hides the route to a human.
- Field service systems. If technicians will not use it one handed on site, it does not exist.
- Brand identity. Postpone it if enquiries are still landing in an inbox nobody owns.
- Graphic design. If you have an identity you are happy with and you need hands, hire hands.