The decision in one table
| Off-the-shelf | Custom | |
|---|---|---|
| Speed to first login | Faster | Slower |
| Fit to how you work | You adapt | It adapts |
| Licence shape | Per user / per module, forever | You own the result |
| Change later | Vendor roadmap, or expensive “customisation” | Your backlog |
| Fails when | Your process is the product | You rebuild a solved commodity |
Buy the commodity
Accounting, email, payroll, card payments, generic CRM for a simple pipeline. Building those from scratch is how custom projects get a bad name. Use a product, integrate it, move on.
Build the middle
The job that is specific to your warehouse, your job cards, your dealer network, your billing exceptions — that is where staff time is going today, usually into Excel. That is also where custom software cost is justified, especially if you start with an MVP.
Worked cases, without the slogans
Buy. You need GST-compliant accounts, email, and a way to take cards. Those are solved. Pay the subscription. Integrate. Spend the custom budget on the operational middle.
Build. You job, stock and bill in a way no package author would put on a roadmap this decade. Staff already maintain a shadow system. A focused MVP of that middle, talking to Tally, is cheaper over three years than another round of “customisation” you will not own.
Hybrid, the usual answer. Tally or a SaaS ledger stays. The website stays. You build the portal, the job cards, the branch stock, the thing customers actually feel. That is not indecision. It is refusing to rebuild commodity software.
The cost argument, finished
A product’s year-one invoice is licences plus a partner. Year two is licences again, plus the next customisation. Custom’s year one is higher if you build a lot; year two is hosting and change you control. Neither wins on a spreadsheet with one column. Put five years and the staff hours you already spend on workarounds in the same sheet. Then decide.
Five years, same sheet
Write three columns: product, custom, hybrid. Year one: licences or build, plus the partner, plus your staff hours in workshops. Years two to five: licences again, or hosting and a change budget you control. Add the hours you already spend on the spreadsheet that exists because the last system almost works. That last line is the one vendors leave off. It is often the line that flips the answer.
A product wins when those five years are mostly subscriptions and you are not paying a consultant to invent a unique process inside someone else’s module. Custom wins when year-two “small changes” on the product already look like a second system. Hybrid wins when Tally or a SaaS ledger stays honest and you only build the operational middle.
Switching cost is part of the buy
Leaving a SaaS means an export, a mapping project, and a month where two systems are true. Leaving a custom system you own is a handover of source, hosting and backups — still work, but you are not waiting on a vendor’s exit fee or a data format they redesigned last year. Ask the product vendor what a full export looks like before you sign. If the answer is “CSV of some tables”, price the cleanup.
A one-page decision you can take into the room
- Is this job common, and can we change how we work? Buy.
- Is this job why customers stay, or why staff already keep a shadow system? Build a slice.
- Are we about to rebuild accounts, email or payments? Stop. Those are products.
- Is the quote to “customise the package” already approaching a system we would own? Build.
- Can we name the first live slice in one sentence? If not, buy a discovery, not a platform.
What “80 percent” actually means
It means staff can complete the job in the product without a shadow spreadsheet, and the remaining 20 percent is irritation, not a second process. If the missing 20 percent is how you invoice, or how you allocate stock across branches, you do not have an 80 percent fit. You have a product that does the brochure and a workbook that does the business. That is a build — or a different product — not a pep talk about change management.
Change management is real when the product is close and the habit is the problem. It is a slogan when the data model cannot express the job. Ask someone who types all day, not someone who signed the licence. The full table and failure modes stay on this page; the ERP-shaped version is custom versus SAP versus Odoo.
A dealer portal, decided in public
Dealers need to see their orders, raise a claim, and download a current price list. No reputable product does your claim rules. A product does the catalogue and maybe the login. Hybrid: product or your existing site for the brochure; a small custom portal for claims and the price list that must not leak between dealers; Tally or the ledger stays. That is three sentences. It is also a three-to-four-month first slice if the data is dirty, not a “website plugin”.
If a vendor quotes the plugin, ask what happens when two dealers share a login “just this once”, and who owns the price file. Those answers decide build versus buy more honestly than a feature matrix.
Do the same walk-through for job cards, branch stock, or retainers. If the product can show the job in a sandbox and the person who types says they can live with it, buy. If they describe the workbook they would keep, you are past 80 percent — whatever the brochure claimed. Write the first-slice sentence the same day. Money for that slice is on custom software cost.
ERP-shaped decisions have their own comparison: custom versus SAP versus Odoo. How we estimate a build is in how much custom software costs.
A service firm that bills retainers, parts and call-outs
No mid-market CRM does that invoice the way the customer already expects. A product will give you a project, a timesheet, or a simple invoice. Staff will keep the “real” bill in Excel. That is not an 80 percent fit. It is a product plus a second process. Price a first slice: create the job, add parts and time, raise the bill the customer already recognises, post to Tally. The brochure site and email stay bought.
If a salesperson says “our service module is flexible”, ask them to raise last month’s actual invoice in a sandbox, with the same exceptions. If they cannot, you have your answer. Flexibility in a slide is not a data model. Write the first-slice sentence before you leave the room. If you cannot, you are not choosing software yet. You are still choosing the job. Come back when the sentence exists. The table on this page is then easy to apply.