All decision guides

For teams weighing a bespoke build against a platform

Custom software has no ceiling. The question is what you are paying the ceiling for.

Build custom when the software is the product. Choose a platform when the differentiation is in your operating model and the backlog is data, screens, permissions and workflows.

Type
Comparison
Updated
2026-08-21
Read
4 min
Sources
1

What the build is actually for

Build custom when the software itself is the product you sell, or when the architecture is genuinely novel. Choose a platform when the differentiation lives in your operating model and the backlog is mostly data, screens, permissions, workflows and integrations — because that backlog is commodity engineering you would be paying to write, secure, deploy and then maintain.

A project price, an ownership commitment

A custom build is priced as a project and lived as an ownership commitment. The first release is the smallest part of it. Afterwards, every change needs a developer, and the permission model, audit trail, deployment pipeline, monitoring and backups that a platform includes are all yours to build and keep working.

Signals to look for

None of these is decisive on its own. Three or more together usually is.

  • The backlog is forms, data and permissions

    Read the last twenty tickets. If most are a field, a screen, a rule or an integration, you are about to pay engineering rates for commodity work.

  • Operations cannot wait for the release cycle

    The business changes something monthly and the software changes quarterly. The gap fills with spreadsheets, which is how you end up with two systems.

  • The estimate covers the first version only

    Security, testing, deployment, monitoring, backups and the next two years of change are separate lines — or missing from the quote entirely.

  • You already know the requirements will move

    That is exactly the case a fixed-scope build handles worst, and exactly the case that gets estimated as though it were stable.

  • Nobody has decided who owns it after launch

    The build has a team. The system, after handover, frequently does not — and an unowned system stops changing within a year.

What you are actually buying

Custom development is not the worse option. It is a different purchase, with a different thing included and a different thing left to you.

What you are deciding Custom development You buy exactly your specification Bondi You buy the commodity layer
What arrives Exactly what you specified, and nothing you did not A working system from a described operation, reviewable before it is created
Time to a working version A discovery and build cycle The AI Builder produces a reviewable plan for data, screens, permissions and workflows
Who changes it in month nine A developer, in a release Whoever owns the operation, in Development, published after review (Pro)
Permissions, audit, deploy, backup Built and maintained by you Part of the platform: role, screen and action permissions on Standard; record and field permissions plus Development → Production publishing on Pro; audit history for 90 or 365 days
Data ownership Yours by construction Yours by construction — dedicated PostgreSQL, import and export, full SQL export on Pro
The ceiling None, and you pay for it whether or not you reach it The platform's model, with APIs, integrations and controlled extension points past it
Where it stops When the backlog is forms and rules rather than novel behavior When the software itself is the differentiated product

Both sides describe a purchase, not a quality judgment. Plenty of operations are better served by custom development.

How to evaluate it

Five steps, in order. Each one produces something you can put in front of other people rather than an opinion you have to defend.

  1. 01

    Split the backlog into unique and commodity

    The unique part is your business rules. The commodity part is everything a platform already includes. Price them separately and the decision usually makes itself.

  2. 02

    Price ownership, not the first release

    Three years of change, dependency updates, on-call, and the person who understands the codebase well enough to be the one who changes it.

  3. 03

    Write down the verification requirements

    Audit trail, access review, session handling, logging, recovery, rollback. OWASP's Application Security Verification Standard is a usable checklist for what a custom build has to cover itself and what a platform should already answer.

  4. 04

    Name the person who changes it after launch

    If the answer is "the agency", that is a subscription too — just an unpredictable one with a longer response time.

  5. 05

    Keep the genuinely novel part

    If something really is unique engineering, a platform with APIs and extension points can host the operation around it rather than replacing it.

Where this leaves Bondi

Where Bondi fits

Bondi supplies the production platform — database, screens, permissions, workflows, APIs, agents, environments — and the AI Builder that turns a described operation into a reviewable plan. Your team, or an implementation partner, concentrates on the operating model and adoption. Where you genuinely need to go outside the model, there are APIs, integrations and controlled extension points.

  • A dedicated PostgreSQL database per environment, with data import and export
  • REST, GraphQL and webhooks on every plan, plus a full SQL export of the database on Pro
  • Approvals, version history and restore on Pro, rather than built by you
See what that looks like for customer onboarding

The other half of the answer

When Bondi is the wrong answer

  • The software itself is the product you sell
  • You need an architecture Bondi does not have — your own hardware, a non-relational core, or devices in the loop
  • The interaction design is the differentiator and has to be pixel-specific

If one of these describes you, the honest recommendation is somewhere else. Nothing on this page is improved by pretending otherwise.

Common questions

Is Bondi no-code?

Not exactly. It removes most of the repetitive application work, and it still exposes REST and GraphQL surfaces, webhooks, integrations, a full SQL export of the database on Pro, and controlled technical extension points. Teams with engineers use those. Teams without them mostly do not need to.

Can we export our data?

Yes. Bondi runs on a dedicated PostgreSQL database per environment and supports data import and export, and Pro adds an owner-level export of that whole database as SQL. What Bondi does not do today is hand you a live Postgres connection string — BI read replicas are an Enterprise arrangement. Raise it before you buy if a live BI connection is a requirement.

What if we outgrow it?

The data sits in a PostgreSQL database you own and can export, which is the part that makes leaving possible at all. That is a deliberate answer to a fair question rather than a reason not to ask it.

Can our agency build it in Bondi?

Yes. An implementation partner can build the system and hand over something the client can keep changing, rather than a codebase they cannot. Scoped implementation work is quoted separately from the subscription.

Sources

Claims about third-party products are limited to what these documents state on the date shown. Where a document is silent, this guide describes the product's design center rather than asserting a limit.

Answer it with your own operation in it.

Describe how your operation actually works and the AI Builder turns it into a reviewable plan for data, screens, permissions and workflows — before anything runs.

Start the 30-day trial Or book a fit call

30 days free · no card · Pro capabilities on Standard quotas, up to 3 users