Making the case

Build vs buy software: custom development or off-the-shelf SaaS?

A practical build vs buy framework: process differentiation, integration depth, five-year licensing math, SaaS customization ceilings and hybrid patterns.

David Lambert Published 6 min read

Build custom software when the process it supports differentiates your business, when off-the-shelf products can’t follow your workflow without constant workarounds, or when the five-year licensing bill outgrows the cost of owning your own system. Buy SaaS when the process is standard - accounting, email, payroll - and a mature product fits with light configuration. Most companies should do both: SaaS for commodity functions, custom software for the workflow that makes them money.

Updated July 2026.

The build vs buy software decision is rarely urgent, which is why it is so often made badly - defaulted rather than decided. A subscription starts as the low-risk option and quietly becomes an architecture. This framework is the one we walk through with clients in Bangkok and abroad before any code is discussed - including the cases where building is clearly the wrong answer.

What question should you actually ask?

Not “can a product do this?” - almost every product can do something close. The better question: is this process one where being different is worth money to us?

  • Commodity processes - accounting, email, payroll, calendars, ticketing in its plain form - are the same in every company by design. Regulators, auditors and employees all benefit from standardization. Building here is re-inventing a wheel that mature products sell cheaply.
  • Differentiating processes - the way you price, quote, schedule production, handle claims, serve customers - are where your operating advantage lives. Forcing them into a generic product’s shape means trading away the difference that wins you business.

Most companies have far fewer differentiating processes than they think - usually one to three. That is precisely why they deserve tailored software: the list is short and the leverage is high.

How do the two options really compare?

CriterionOff-the-shelf SaaSCustom software
Time to first valueDays to weeksMonths for a first release
Upfront costLowSignificant
5-10 year costSubscriptions scale with seats and modules, indefinitelyConcentrated early, then maintenance
Process fitYou adapt to the productThe product adapts to you
IntegrationWhatever the vendor’s API allowsAs deep as your systems require
Competitive edgeSame tool as your competitorsYours alone
OwnershipRented; terms and prices can changeYou own the code and IP
Maintenance dutyVendor’s problemYours (or your partner’s)

Why does the five-year licensing math surprise people?

Because SaaS pricing is designed to be easy to enter and easy to underestimate. The pattern to run before deciding:

  • Count all seats at target scale, not at pilot scale - per-user pricing that looks trivial for 10 users reads differently at 200.
  • Count the tiers and add-ons. The features that made the product fit your demo are often in the higher plan; API access, SSO (single sign-on) and audit logs frequently are.
  • Assume price revisions. Subscription prices historically move in one direction, and switching costs grow with every year of accumulated data.
  • Add the workaround labor - the spreadsheets, re-keying and manual reconciliation that fill the gap between the product and your process.

Put that five-year total next to a real quote for building the system, including maintenance. Sometimes SaaS still wins decisively - then buy it with confidence. But run the math over the horizon you will actually operate the system, not over year one. If most of your cost sits in offshore development anyway, our note on the total cost of software development in Thailand shows how to read the build-side quote.

What is the customization ceiling - and why does it matter more than features?

Every configurable product has a point where the checkboxes run out. Below the ceiling, adapting the tool is cheap. Above it, three expensive things happen:

  1. The process bends to the tool. Teams change how they work - not because it is better, but because the software insists.
  2. The workaround economy appears. Exported spreadsheets, duplicate data entry, “the real numbers are in this other file”.
  3. Integration debt accumulates. Each satellite tool bought to patch a gap adds one more sync problem.

The diagnostic is simple: if you are already inventing workarounds during the trial, you are at the ceiling before rollout. Legacy in-house systems hit an equivalent wall from the other direction - when patching an aging application costs more than replacing it, which is a separate decision we cover in legacy modernization: re-engineer or rewrite.

What does the hybrid pattern look like?

The strongest architectures we see are deliberately mixed:

  • SaaS core for commodity functions. Accounting, HR, email, CRM (customer relationship management) in its standard form - mature products, easy fit, someone else’s maintenance.
  • Custom software for the differentiating workflow. The pricing engine, the claims process, the production scheduler - built to your process exactly, owned by you.
  • Custom glue in between. Integration code that moves data between the SaaS products and your systems automatically, ending the re-keying that otherwise fills the gaps.

This keeps the build surface small - you build only what earns its keep - while the integration layer quietly becomes one of the highest-ROI (return on investment) pieces of software a company can own. It is also, in our experience as a custom software development partner, the most common first project: not a grand replacement, but the glue and the one workflow that matters.

When is building clearly wrong?

An honest framework needs this list:

  • A mature product fits with light configuration. Take the win; spend your build budget elsewhere.
  • The process is regulated into uniformity. Payroll and statutory accounting shouldn’t be creative.
  • You cannot commit to ownership. Custom software needs a maintenance budget and a decision-maker; if neither will exist after launch, it will decay.
  • The requirement is still fog. If nobody can describe the workflow, a configurable product used for a year is cheap requirements research - build the replacement once the shape is clear.
  • You are building to avoid a subscription on principle. Resentment of per-seat pricing is not a business case; the five-year math is.

How to decide in practice

  1. List your processes and mark the one to three that differentiate. Everything else defaults to buy.
  2. For each candidate, run the five-year totals - all seats, all tiers, workaround labor versus build quote plus maintenance.
  3. Check the integration depth required. The more systems must talk, the more the custom-glue pattern pays.
  4. Pilot the smallest useful build. One workflow, a few months, working software - the decision stops being theoretical.

In short

  • Buy for commodity processes; build for the one to three workflows that actually differentiate you.
  • Compare costs over five years, all seats, all tiers, plus workaround labor - not year one.
  • The SaaS customization ceiling shows up in trials as workarounds; it only gets costlier after rollout.
  • The hybrid pattern - SaaS core, custom differentiators, custom glue - beats both extremes for most companies.
  • Building is wrong when a mature product fits, the process is standardized, or nobody will own the result.

AD System Asia has built custom software from Bangkok since 2010 - and told a fair number of prospects that the honest answer to build vs buy software was to buy. If you want that same walk-through on your own case, request a call: bring the process and the SaaS quotes, and we will help you run the math before anyone talks about code.

A few more answers

Frequently asked questions

When should a company build custom software instead of buying SaaS?

Build when the process is a genuine differentiator, when no product covers your workflow without heavy workarounds, when deep integration with existing systems is essential, or when per-seat licensing across many users exceeds the cost of ownership over five years. Buy when the process is standard and a mature product already fits.

Is custom software more expensive than SaaS?

Upfront, almost always. Over five to ten years, often not: SaaS subscriptions scale with users and modules and never end, while custom costs concentrate in years one and two, then fall to maintenance. The honest comparison is total cost over the horizon you expect to run the system, not year one.

What is the customization ceiling of SaaS products?

The point where configuration options run out and the product cannot follow your process any further. Teams then bend the process to fit the tool, buy more add-ons, or maintain spreadsheets alongside it. If you are already hitting workarounds during the trial period, the ceiling will only get more expensive after rollout.

Can I combine SaaS and custom software instead of choosing?

Yes - it is often the strongest pattern. Keep commodity functions on mature SaaS (accounting, email, HR), and build custom software only for the differentiating workflow and the integration layer that connects everything. You get proven products where fit is easy and tailored software only where it earns its cost.

How long does it take to build a custom business application?

A focused first release commonly takes a few months rather than years, if scope is disciplined: one core workflow done properly, then iteration. Timelines balloon when the first version tries to match every feature of an off-the-shelf suite instead of the handful your operation actually uses.

What are the risks of building custom software?

The main ones are underestimating lifetime maintenance, building what a mature product already does well, and depending on a vendor without securing code ownership and documentation. All three are manageable: scope the first release narrowly, buy commodity functions, and contract for IP, handover and documentation from day one.