How to Choose a Custom Software Development Company

group of people using laptop computer

A capable custom software development company can name a system it still supports after five years, name the engineers who will write your code, and show its change-control process in writing.

Capable custom software development companies can do three things on a first call: show you a system they still support years after launch, name the people who will write your code, and explain exactly how the estimate changes when scope changes. Risky ones sell capability in the abstract, quote a single number, and staff the work after you sign.

This piece assumes you've made the build decision. It is purely about shortlisting and scoring vendors.

Below are 12 questions, each with what a strong answer sounds like and what a red-flag answer sounds like. It also settles the three objections that stall most Australian procurement decisions: local versus offshore delivery, Australian data residency, and time-zone collaboration.

The 12 Questions at a Glance

These are the questions to ask a software development company, worded the way you would ask them on a call. Copy them straight into your shortlisting document.

  1. How long have you been building software, and what is the oldest system you still support?

  2. What have you built in our domain, and what did you get wrong the first time?

  3. Who exactly writes the code, and where do they sit?

  4. What happens in discovery before you quote?

  5. How do you estimate, and what happens when scope changes?

  6. Why this technology stack for our problem?

  7. Who owns the code, the IP and the repositories?

  8. How do you handle security, and where will our data physically live?

  9. What does your testing and QA process actually include?

  10. What do we receive at handover?

  11. What does support look like after launch, and what are the response times?

  12. Fixed price or time and materials, and why for our project?

Track record, domain fit and who builds it (questions 1 to 4)

These four questions establish whether a vendor has actually done this before, in a domain like yours, with named people you can hold accountable rather than a generic capability statement.

1. How long have you been building software, and what is the oldest system you still support?

Longevity is measurable. Ask for a system still in production after five or more years, and the name of the person who has maintained it.

Strong answer: a named system, its launch year, the framework versions it has been carried through, and the engineer who did the upgrades. Twenty years of continuous delivery, in Sentia's case since 2006, mostly proves one thing: the team has survived several technology shifts without abandoning clients mid-stack.

Red flag: “we have delivered over 200 projects” with no system they still support. Project counts and directory star ratings measure sales throughput, not reliability.

2. What have you built in our domain, and what did you get wrong the first time?

Domain fit is about failure modes, not logos. A vendor who has built in your space knows which integration breaks and which compliance step blocks release.

Strong answer: a specific build, the scale it operates at, and a candid account of a wrong assumption and the correction. Sentia's own reference points are multi-year relationships with organisations like Airtasker, Shippit, RANGEme, eclipx and Boardroom, plus web applications serving 15M+ users and 5M+ app downloads. Ask any vendor for the relationship, not the logo.

Red flag: case study snippets with no numbers, no dates and nothing that went wrong.

3. Who exactly writes the code, and where do they sit?

You are buying named people, not a capability statement. Get first names, seniority, hours of overlap with AEST, and employment status.

Strong answer: a clear split between local, nearshore and offshore, disclosed upfront. Sentia's is a Sydney-based senior team with nearshore engineering in Zagreb, and we say so before anyone asks.

Red flag: senior architects on the pitch and unnamed juniors on delivery. Also treat undisclosed subcontracting as disqualifying, not as a detail to sort out later.

4. What happens in discovery before you quote?

A quote produced without discovery is a guess wearing a number.

Strong answer: a paid or scoped discovery producing prioritised scope, named assumptions, a defined first release, and an explicit MVP (minimum viable product, the smallest release that delivers real use) boundary listing what is deliberately excluded.

Sentia treats discovery as a paid, scoped step rather than a free guess; our guide on writing a scoping document sets out what it should include.

Red flag: a fixed price within 48 hours off a two-page brief, with no assumptions document.

The single takeaway: verify track record through long-term retention and post-launch support history, not case study counts.

Estimates, stack, IP and security (questions 5 to 8)

These four cover the mechanics that determine whether a disagreement six months into the build is even possible, how estimates hold up, what the code is built on, who owns it, and where your data actually sits.

5. How do you estimate, and what happens when scope changes?

Scope will change. The question is whether you document the change process before it happens.

Strong answer: estimates broken down by feature with ranges and stated assumptions, plus change control in plain terms: written change requests, re-estimation before work starts, and a decision log both parties can read.

Red flag: one fixed number, no assumptions listed, and change handled “collaboratively”. That means the argument happens later, under time pressure.

6. Why this technology stack for our problem?

Stack choice is a maintenance decision, not a preference.

Strong answer: the choice justified against your hiring pool, long-term maintenance cost and integration needs. Ruby on Rails, for example, is a defensible choice where speed of iteration and a deep Australian hiring pool matter. Sentia has 20 years of Rails experience, including published open-source gems. API (application programming interface) development and integration need a stated reason too: which systems, which direction, which failure handling.

Red flag: novelty as the rationale, or a stack no one else in the market hires for.

7. Who owns the code, the IP and the repositories?

Strong answer: repository access from day one under your organisation's account, intellectual property assigned on payment of each invoice, and a written list of third-party libraries and licences with any commercial obligations flagged.

Red flag: withholding repository access until final payment, or transferring IP only at project completion. That is leverage, and it converts a delivery dispute into a hostage situation.

8. How do you handle security, and where will our data physically live?

Name the region, not the provider. “AWS” is not an answer.

Strong answer: the region where personal data is stored and processed, who has production access, whether subcontractors touch production data, and how the vendor meets the Australian Privacy Principles under the Privacy Act 1988. Cross-border disclosure carries specific obligations under APP 8.

Red flag: vagueness about where data sits. Data stored or processed outside Australia may be reachable under foreign jurisdiction access laws.

The single takeaway: assumptions, change control and data location belong in writing before contract, not in a kickoff deck.

QA, handover, support and commercial model (questions 9 to 12)

The last four questions decide what you are left holding once the build ships, how thoroughly it was tested, what you receive at handover, how support behaves after launch, and how the work gets priced.

9. What does your testing and QA process actually include?

Strong answer: automated test coverage at unit and integration level, a CI/CD (continuous integration and continuous delivery, the automated pipeline that builds, tests and releases code) pipeline that blocks failing builds, a staging environment matching production, and a named person who signs off releases.

Red flag: manual testing performed by the developer who wrote the feature, with no pipeline and no sign-off step.

10. What do we receive at handover?

The real test: could you move vendors next quarter without stopping the business?

Strong answer: architecture notes, environment setup instructions another engineer can follow, a deployment runbook, API documentation, credential transfer through a password manager, and a recorded walkthrough session.

Red flag: “the code is self-documenting”, or documentation offered as a paid extra after go-live.

11. What does support look like after launch, and what are the response times?

Strong answer: named clients supported for more than five years, how support is staffed during leave and incidents, and an SLA that states response windows in AEST business hours plus what happens outside them. Sentia's long-running relationships date back to 2006, which is why our support answer is specific.

Ask for two referees, one current client and one whose engagement has ended, and ask each what happened when something broke. A referee whose project has finished will tell you more about support than any live client will.

Red flag: a support retainer with no response times, or an SLA that measures acknowledgement rather than resolution.

12. Fixed price or time and materials, and why for our project?

Strong answer: a recommendation tied to how certain your scope is, and the reasoning behind it.

Factor

Fixed price

Time and materials

Scope certainty needed

High, documented upfront

Low to moderate, evolving

Cost of change

High, via change requests

Lower, reprioritised in backlog

Who holds risk

Vendor, priced into buffer

Buyer, managed via cadence

Best fit

Defined integration or fixed-scope release

Discovery, MVP, ongoing product work

Ask for evidence of professional indemnity and public liability cover. Australian procurement teams routinely require certificates of currency, and a vendor who has them ready has already been through real due diligence.

Red flag: fixed price offered for genuinely undefined work. Any AUD figures at this stage should be ranges with assumptions attached, never quotes.

The single takeaway: handover documentation and SLA specifics are the two areas vendors improvise, so ask for them in writing.

What changes between local, nearshore and offshore delivery

The delivery model matters less than knowing which model you are buying and who is accountable in Australian hours.

Delivery factor

Local (Sydney/Australia)

Nearshore

Offshore

Overlap with AEST

Full day

Partial, usually mornings or evenings

Minimal to none

Communication overhead

Low, decisions same day

Moderate, needs written clarity

High, async clarification cycles

Data residency exposure

Australian, simplest under APP 8

Depends on hosting region

Higher, foreign access laws may apply

Who fronts up in an outage

Named local lead

Local lead, engineers on call

Often unclear

Overlapping hours are the practical variable. Decisions made in a stand-up cost minutes; the same decision handled asynchronously costs a day and produces rework. We have written before on choosing between offshore and local talent, and the same logic applies to vetting a vendor.

Look for a disclosed model. Sentia's is custom software development sydney led by a senior Sydney team with nearshore engineering in Zagreb, with production data hosted in Australia and no undisclosed subcontractors. See how we approach offshore versus local talent and where your data lives.

Offshore delivery is not disqualifying. Undisclosed offshore staffing is.

How to score the answers and shortlist with confidence

Score each of the 12 answers 0, 1 or 2. Zero for no answer, one for a generic answer, two for a specific and evidenced answer with names, dates or documents attached.

Then weight. Double the score on questions 1, 8 and 11, because track record, data handling and post-launch support carry the most risk in a multi-year build.

Here is that weighting shown as a scorecard: three colour-coded groups of questions, with the three double-weighted ones spiking outward from the centre.

A flat zero anywhere on one of those three spikes carries more weight than a low score anywhere else on the wheel, which is exactly what the disqualifiers below are built around.

Four answers that should end the conversation

  • No named delivery team

  • No written change-control process

  • No clarity on where data is stored and processed

  • No repository access until final payment

Run all 12 questions with every shortlisted Sydney software development agency, in the same order, and compare the specifics rather than the pitch decks. The gap between vendors usually shows up in questions 4, 10 and 11.

If you want a benchmark before you start, read how Sentia approaches custom software development and use it to calibrate what a specific answer sounds like.

Frequently Asked Questions

How do I choose a software development company if I have never bought custom software before?
Start with discovery, not the build. Ask three vendors the same 12 questions, then compare only the specifics: named people, named systems still in support, written assumptions. Choosing well mostly means refusing to accept abstract capability claims.
What should I ask about data residency?
Ask which region personal data is stored and processed in, whether subcontractors have production access, and how the vendor meets Australian Privacy Principles obligations under the Privacy Act 1988. Data held offshore may be subject to foreign jurisdiction access laws, so get the hosting region in the contract.
What is a reasonable engagement to start with in Sydney?
A discovery and scoping engagement, not a full build. As at June 2025, discovery work with custom software developers in Sydney commonly ranged from roughly $20,000 to $50,000 AUD depending on integration count and stakeholder numbers. That is an indicative range, not a quote.
Is offshore development riskier than local?
Only when it is undisclosed. Offshore, nearshore and local models all deliver working software for buyers of bespoke software development in Australia. The risk sits in unnamed teams, no AEST overlap for incident response, and unclear data residency. Disclosed models with named accountability manage all three.
Should we hire in-house developers instead of an agency?
For a first custom build, an agency gives you a named team, discovery discipline and post-launch support without the hiring lead time. In-house tends to suit ongoing product work once the system and its direction are settled.

Ready to shortlist with confidence

Building the shortlist is the first move, not the build. When you are ready, start a no-obligation scoping conversation with Sentia and we will answer all 12 of these questions about your project before you commit to anything.