Both answers are right for somebody. Here is the test we use on the first call, and we send people to a subscription when that is what the test says.
Five questions
1. How many people, and is that number going up?
Per-user pricing is cheap at five seats and expensive at fifty. Under eight people with no hiring plan, a subscription is almost always the right call. Above fifteen with hiring, run the arithmetic in the three-year post.
2. Has the process stopped changing?
A build encodes how you work. If the way a lead becomes an order changed twice this year, building it now means paying for changes later. Subscribe until it settles, then build what settled.
3. Do your customers need to log in?
This is the question that decides most cases. A retailer app, a parent app, a client portal: no subscription CRM gives your customers a screen with your name on it that reads from your own data. Once customers need a login, you are building something anyway, and the only question is whether the back office it reads from is yours too.
4. How many tools have to agree with each other?
Two tools with one integration is fine. Five tools with an automation plan holding them together is a system you are already maintaining, just without owning it. Count the integrations. At three or more, a single data model is cheaper to run and easier to explain to a new hire.
5. What happens in year four?
With subscriptions, year four costs what year three cost, plus the price increase. With a build, year four is hosting and whatever changes you choose to make. If you expect to be in the same business in four years, that difference is the whole argument.
Where SaaS wins outright
- Accounting and statutory filing. Regulations change monthly and the vendors track them. Do not build this.
- Email, documents, calendars, video calls. Commodity, and the network effect matters.
- Anything you will use for less than a year.
- Any team that cannot name one person who owns the system internally. Software you own needs an owner.
Where a build wins outright
- Trade-specific workflows a generic tool bends around: karigar issue and receive by weight, batch and expiry picking, connector payout grids, fee ledgers with concessions.
- Customer-facing apps that read from the same data the office uses.
- Reporting that the subscription tier does not include and never will at your size.
- Businesses where the software is part of what they sell, and a competitor on the same subscription looks identical.
The honest middle
Most businesses we talk to land in between: keep the accounting subscription, keep email, build the back office and the customer app on one engine, and point the website at both. That is what the four layers on the home page are, and the bands there are the same whether you take one layer or all four.
Your numbers
Pick your layers, see the band.
The scope builder uses the same bands as this post and shows a timeline and price for your combination in a minute.