A good custom software strategy answers nine questions before any code is written: who the users are, what outcome you need, who owns decisions, what the first release includes, whether to build or buy, how you will fund ongoing work, how you will contract, how you will judge progress, and who runs the software after launch. Teams that answer these first get more useful software for the same budget.
This guide is for owners, operations leaders, and IT managers about to hire a development team. We use the same questions when we scope projects for clients.
Who will use the software, and what do they do today?
Start with the people, not the feature list. Write down each user group, the task they do now, the tools they use, and where the work gets stuck. Then talk to a few of them. A requirement that comes from a stakeholder meeting is a guess until a real user confirms it.
The public sector learned this the hard way. The 18F de-risking guide, written for government technology buyers, puts user-centered design first among its modern methods: build around research with real users, not assumptions (18F Derisking Guide).
What business outcome are you paying for?
"A new portal" is an output. "Cut intake time from two days to same day" is an outcome. Pick one or two outcomes you can measure, and write down how you measure them today. That baseline is what lets you judge the project later.
Who on your side will own product decisions?
Every project needs one person on the buyer side who can say yes or no to scope questions within a day. This person does not need to be technical. The 18F guide says it plainly: "A product owner doesn't need to be an expert in technology." They need to know the business and have the authority to decide.
Without that person, the development team either waits or guesses. Both cost money.
What is the smallest first release that proves value?
Define a first release that one user group can use for real work within a few months. Everything else goes on a ranked list for later. A small first release surfaces wrong assumptions early, while they are still cheap to fix.
One of our clients, an Employee Assistance Program provider, started with a single intake bottleneck. They fixed it with Arches workflow automation and then improved scheduling. That is the kind of narrow first step we recommend.
Should you build, buy, or extend what you have?
Custom software makes sense when the process is a real part of how you win work or serve clients, and off-the-shelf tools force you into workarounds. It makes less sense for common back-office tasks where a mature product exists. Many projects end up in the middle: a standard CRM or ERP extended with custom workflows and integrations. Ask your vendor to argue the buy option before they quote the build.
How will you fund the software after launch?
Custom software keeps changing as your business changes. 18F's state software budgeting handbook makes the point that custom software is never finished, and that maintenance should simply mean continuing to modify it for user needs (18F State Software Budgeting Handbook). The same handbook uses a rule of thumb that a team of 5 to 9 developers costs $1 to 2 million per year, depending on local costs. Most private-sector projects are smaller, but the lesson holds: plan an ongoing budget, not a one-time build.
A simple test: if your plan has a launch date and no line for the following year, the plan is not finished.
How should you structure the contract?
Large fixed contracts for a full feature list push risk onto change orders. A safer pattern is to fund a team for a defined period, with a ranked backlog and the right to adjust priorities. The 18F guide calls this performance-based contracting: pay for results, not a prescribed way of working.
Whatever the structure, make sure the contract covers code ownership, access to the repository, hosting accounts in your name, and a clear exit plan. We cover vendor checks in more detail in Remote Development Team: What to Check First.
How will you know the project is on track?
Ask to see working software on a fixed schedule, every two to four weeks. Demos of real features, used by real users, tell you more than status reports. Agree up front on what "done" means for a feature, including testing. Our guide to custom software testing lists what to require.
Who will run the software after launch?
Decide who handles hosting, security updates, backups, monitoring, and user support. If it is the vendor, put response times and costs in writing. If it is your team, plan the handover early, with documentation and access transferred before the final payment.
What are the common mistakes?
- Writing a long requirements document before talking to users.
- Having no single decision owner on the buyer side.
- Funding the build but not the year after launch.
- Judging progress by status reports instead of working software.
- Leaving code, hosting, and data access in the vendor's name.
Frequently asked questions
It is a short plan that defines users, outcomes, decision ownership, first release scope, build or buy choices, funding, contract structure, progress checks, and ongoing support before development starts.
For most small and mid-sized projects, a few weeks is enough. The goal is clear answers to the core questions, not a perfect specification.
No. The product owner needs business knowledge and decision authority. The development team supplies the technical expertise.
Fixed price can work for a small, well-understood scope. For larger or less certain work, funding a team for a set period with a ranked backlog usually manages risk better.
Planning a custom software project?
We help teams answer these questions, scope a first release, and build it, with ownership and support planned from day one.
Explore custom development servicesRelated: Application development services and Does Your Software Vendor Use AI? What to Ask.

