Build, buy, or hybrid: a decision framework that holds
Custom software is the right call less often than founders think. Here is a simple way to decide where to buy off the shelf, where to build, and how to stitch the two together without regret.
Vinerals TechnologiesWorkshop notes
Every few months a founder tells us they want to build custom software. Usually the reason is real. Sometimes a demo went badly, or a competitor mentioned their in-house tool at a conference, and “we need our own system” became the plan overnight. The first useful move is to slow down and ask what the software is actually for.
Most of the time, some of the work should live in tools you rent, and a small slice deserves something you own. The trick is knowing which slice. This framework will not hand you a single number that ends the debate. It will give you better questions, and a clearer picture of what you are signing up for over five years.
Start by naming what the software actually does
Before build or buy, write down the job in one plain sentence. Skip labels like “a CRM.” Try something like: “We need to track which restaurants have reordered in the last 60 days and nudge the ones who have not.”
Once the job is on paper, most of them sort into two piles. There is work that thousands of companies do in nearly the same way: payroll, email, accounting, scheduling a call, storing files. Then there is the work that is specific to how your business wins: the way you price a job, route a delivery, qualify a lead, or run the process your customers actually pay you for.
Keep that sentence handy. You will use it in every section below.
When renting off the shelf is the right answer
SaaS wins more often than founders expect, and that is fine. If a category has ten mature products and a crowd of happy users, that problem is largely solved. Someone has already spent years on the edge cases you have not thought of yet.
Rent it when:
The work is common across companies.
Bookkeeping does not become a competitive edge because you wrote it yourself.
Your process fits the tool around 80 percent or more.
If you can adapt your habits to the software without real pain, do that.
You are under about 20 people.
At that size a custom build rarely pays for itself before your needs change again.
You need it working in weeks, and being wrong is cheap.
A month-to-month subscription you can cancel is a small bet.
The quiet benefit of SaaS is that maintenance is somebody else’s job. Security patches, uptime, a support line at 2 a.m.: you pay a monthly fee and stop thinking about it. For most of what a growing company touches, that is exactly the deal you want.
Where SaaS starts to hurt is subtler. Per-seat pricing compounds as you hire. A tool that costs a little at eight people costs real money at sixty. You bolt on connectors to make three products talk to each other, and those connectors become their own fragile system that breaks on a Tuesday. When the vendor ships a feature you did not ask for, or retires one you relied on, you adjust. That does not make SaaS the wrong choice. It means you should watch the meter.
When custom software earns its keep
Custom becomes worth it when the process is the thing customers pay you for, and no product on the market runs that process the way you do.
Picture a logistics company whose margin lives in how they sequence pickups. An off-the-shelf routing tool gets them most of the way, and quietly caps their advantage at “as good as everyone else who bought the same tool.” The last stretch, the part that is genuinely theirs, is where a custom build can pay back. You are paying to keep the thing that makes you money out of a template.
Build when:
The process is your actual advantage.
Matching a competitor’s tooling would flatten what makes you different.
You have tried to buy it.
The good options force you to work in ways that break how you serve customers.
Data ownership, residency, or compliance leaves little room to negotiate.
The rented product cannot meet the constraint without painful workarounds.
The five-year cost of the SaaS stack for this job keeps climbing.
As a rough signal, that often means you have crossed about CAD $80,000 to $100,000 a year and the curve is still up.
Two honest cautions. First, custom keeps costing after launch. Plan on annual maintenance somewhere around 15 to 25 percent of the build cost, plus hosting and the internal time to decide what gets built next. Software you own needs tending. Second, “we could build it” is almost always true. The real question is whether building it is the best use of the money and attention you have this year.
The hybrid pattern most SMEs should use
Here is where most companies land once the emotion drains out of the decision. You buy the commodity layer. You build the thin, specific piece that is yours. Then you connect them cleanly.
In practice that looks like keeping accounting, email, payroll, and calendars in mature products you would never want to maintain. Around them, you build a small system that owns the one workflow that defines you, and you let it read from and write to the rented tools through their APIs.
Keep the custom surface small on purpose.
The more of it you build, the more of it you maintain. Build the part that is genuinely yours and rent the rest without guilt.
Own your data model.
Even when the tools are rented, the record of who your customers are and what they did should live somewhere you control and can export.
Treat the connectors as real software.
The integration layer is where hybrid systems rot. Give it the same care as the rest.
Write down why each piece is build or buy.
In two years someone will ask. The reasoning is worth more than the decision.
Done well, the custom part is the smallest thing that captures your advantage, and everything else is a subscription you can reconsider each year.
A five-year cost sketch, without pretending it is precise
People often compare the monthly SaaS bill to the upfront build price and conclude custom is absurd. Look at both over five years, with every real cost included. The numbers below are directional. Your real spread depends on scale, complexity, and how much your process bends.
A fair five-year SaaS total is more than the sticker. Count the subscription with a yearly increase of roughly 5 to 8 percent, the adjacent tools you buy to fill gaps, the connectors wiring them together, onboarding and training, and the eventual cost of migrating off when you leave. For a mid-sized team, the visible license is often a third to a half of what the stack actually costs once you add it all up.
A fair five-year custom total is also more than the build. Count the initial build, then annual maintenance in that 15 to 25 percent range, hosting, and the internal time to steer it.
To make it concrete with round CAD figures, imagine a team of about 50 people running one core workflow.
SaaS path.
A core platform plus the tools and connectors around it might land somewhere near CAD $90,000 to $140,000 per year once everything is counted. Over five years, call it roughly CAD $450,000 to $700,000, and it grows with headcount.
Custom path.
A CAD $150,000 build amortizes to about CAD $30,000 a year, plus CAD $20,000 to $40,000 a year in maintenance and hosting. Over five years, roughly CAD $250,000 to $350,000, and it does not scale up every time you hire.
Do not treat those as promises. Treat them as a shape. SaaS is usually cheaper in years one through three, and for commodity work at moderate scale it often stays cheaper. Custom tends to look better over five years at higher seat counts, or where the process is genuinely your advantage and the SaaS bill keeps climbing. Below about 20 staff, the build rarely amortizes before your needs move. The interesting range is 50 seats and up, where the gap at year five is often much smaller than it looked at year one.
Build the spreadsheet before you build the software. If a rough five-year model does not clearly favour custom, that is usually your answer, and it is a good one.
A short checklist you can run in an afternoon
Run each candidate system through these. If you cannot answer yes to at least the first three for a build, rent it.
1. Is this process one of the few things customers actually pay us for?
If it is background work, buy it.
2. Did we honestly try to buy it?
Do the good options force us to work in ways that hurt how we serve customers?
3. Is our five-year SaaS cost for this job real money and still climbing?
As a rough signal: past about CAD $80,000 to $100,000 a year.
4. Can we fund the ongoing attention a custom system needs after launch?
If not, rent it and revisit later.
5. Have we drawn the smallest possible custom surface?
Confirm the rest can stay rented and connected.
Most companies that run this end up hybrid, with a short list of build items and a longer list of things they were relieved not to build.
We build software by hand for small and mid-sized companies. A fair amount of our first conversations end with us telling someone to buy the SaaS product and save their money for the piece that actually matters. That keeps the work honest, and it is usually the right call.
If you are staring at this decision and want someone senior to sketch the five-year picture with you, without a sales pitch attached, we are happy to sit down and work through it. Bring the one-sentence description of the job. We will start there.
