Guide

Custom software or an off-the-shelf tool: how to choose

19 min read

By . Published . Last updated . Reviewed by Mavrin Labs on . How this guide was researched and checked.

A card reading "Custom software or an off-the-shelf tool: how to choose" beside the Mavrin Labs logo
The card that shows when this guide is shared or linked.

Custom software is an application built for one business, while an off-the-shelf tool is built for many and configured at the edges. Start with the least complex option that fits: an existing tool wins when your process is ordinary, and connecting it to the systems you already run fixes most fit problems.

A custom build earns its place when the way you work is itself part of your advantage, and when you can name the result the build has to reach.

The guide is written for owners and operations leads deciding whether to buy, connect or build. It covers the criteria that decide the choice, the usual traps and how to scope a first version. Choosing a particular product and agreeing the investment are left to the conversation that follows.

What is custom software, and how does it differ from an off-the-shelf tool?

Custom software is software built for one business, shaped around how that business works and owned by it, while off-the-shelf software is built for many businesses, configured at the edges and owned by its supplier. The difference that matters is who decides what the software does next.

The decision right runs through everything else. With a tool, a change you need waits for the supplier to agree it is worth building, and a change you did not want can arrive anyway. With a custom build, you decide, and you also carry the upkeep.

Neither position is better in the abstract. The two suit different businesses, and the same business at different times.

ExistingtoolThesupplierdecidesUpkeepincludedQuickestto startConnectedtoolYou ownthe joinsJoins towatchQuick withconnectorsCustombuildYou decideUpkeep isyoursLongest tostart
The three paths side by side: who decides changes, whose upkeep it is and how soon it is useful. The comparison table below adds the other tests.
Off-the-shelf tool
Built for a common case, configured rather than changed, upkeep included, and the supplier sets the direction.
Custom software
Built for your steps and definitions, changed when you decide, upkeep yours, and the direction is yours to set.
The middle path
An existing tool configured and joined to the systems you already run, so the fit comes from the connection rather than the software.

Why it matters: software now runs most firms' work

Software now sits close to how most firms work. In the Census Bureau's report on its 2023 annual business survey, most businesses described specialized software (58.9%) and cloud-based technology (59.0%) as very important to their processes or methods (Census Bureau report on technology use, checked Saturday, October 10, 2026).

That dependence is why the choice deserves a written reason. A tool picked without a reason becomes the system the whole business works around, and replacing it later means moving every record it holds. The seven tests below give the reason a shape.

How do the three paths compare?

The three paths compare most clearly on seven criteria. They are fit to how you work, who decides changes, reach of your records, pace of change, security responsibility, upkeep and time to a first useful version. The table below sets each path against those criteria and adds a best-for row, with no money column, because the investment follows the scope.

How do the three paths compare?
What decides it Existing tool Connected tool Custom build
Fit to how you work close enough for common work fits where joining was the gap shaped to your own steps
Who decides changes the supplier the supplier, and you for the join you
Reach of your records what the tool allows reachable across the joined systems yours throughout
Pace of change the supplier's pace moderate the pace you choose
Security duty mostly the supplier shared by both sides you, with the builder
Upkeep asked of you least the joins watched most, and visible
Time to first use quickest quick where connectors exist longest, as the design is the work
Best for ordinary work done ordinarily a fit problem that is a joining problem underneath work that is part of your advantage

Read the rows as trade-offs rather than as a score. Each path is strong somewhere, and the last row is the short version of the whole table. A business that cannot absorb upkeep should weight that row above the others.

Does the software fit how you work?

Fit is the first criterion because a poor fit shows up every working day, and the table gives each path's fit. What the table cannot show is the sign to look for: a second record kept beside the tool, because the tool will not hold something the work needs.

Fit is also what buyers say they were after. In the Census Bureau's report on its 2023 annual business survey, improving the quality or reliability of processes or methods was the most common reason given for adopting cloud-based technology (51.8%) (Census Bureau technology adoption figures, checked Thursday, October 15, 2026).

That is a process answer, which is why fit is judged against your steps rather than against a feature list.

Who decides what the software does next, and how fast?

The supplier decides what changes in an existing tool, and at its own pace, so a request waits until it suits the supplier's plans. With a connected tool the supplier still owns the tool, but you own the joins, and each join can be adjusted on its own.

Pace matters most where your rules shift often. A supplier's pace may be faster or slower than you need, and a change you did not ask for can arrive anyway. The other side of deciding every change on a custom build is carrying the work of making each one.

Can you reach your own records?

Reach means whether you can see, export and reuse what the software holds: an existing tool allows what its supplier allows, while a custom build keeps them yours. Reach matters for years, since the IRS says to keep most business records three years and some up to seven (IRS guide on keeping records, checked Monday, November 16, 2026).

On a custom build, records stay yours only while you hold the accounts it runs on.

Test reach before you commit rather than after. Ask for an export of a few sample records, open it in a spreadsheet, and check that the history and attachments came out with the names. An export that leaves the history behind gives you reach on paper only.

Who answers for security?

The supplier answers for most of an existing tool's security, and you and your builder answer for a custom build's, as the table shows. Whichever path you take, ask the maker how the software is built and kept secure; the CISA and NIST references later in this guide give the questions.

Those questions are published, and there are 6 areas of them: general practice, authentication, eliminating classes of vulnerability, evidence of intrusions, supply chain and vulnerability disclosure (CISA questions for software buyers, checked Thursday, October 15, 2026).

How much upkeep falls to you?

Upkeep is the criterion that most often rules a build in or out. The table gives the order, from least on an existing tool to most on a build. What it cannot show is that a build's upkeep is visible work, so somebody has to own it before the build starts.

Part of that upkeep is reading a public list. CISA's catalog of vulnerabilities known to be exploited held 1,745 entries on the day it was read for this guide, and the agency tells organizations to use it as an input to deciding what they patch first (the CISA catalog of known exploited vulnerabilities, checked Thursday, October 15, 2026).

On an existing tool the supplier does that work; on a build, somebody of yours does.

How soon is the software useful?

Software is useful soonest when it is an existing tool and latest when it is a custom build, because on a build the design is the work. That is the reason to keep a build's first version small, and to ask for a date when the first real record goes through.

Ask any supplier or builder what has to be true before the first real record goes through: which data has to be moved, who sets the software up and who checks the result. That answer is a better guide to time to first use than a demo.

How do you rate the comparison for your own case?

The comparison turns into a decision once you rate it against one real task. The answer is the path that wins the rows that matter most to you in practice. Use the same five steps each time, so two people who rate the same task reach answers they can compare.

  1. Name the task. Pick one piece of work, not the whole business.
  2. Mark the deciding rows. Choose the rows that would decide it.
  3. Check upkeep first. If nobody will own it, rule out a build.
  4. Compare across them. Take the path that wins most deciding rows.
  5. Write down the reason. Record which rows decided it, for the review.

A tie between two paths usually means the task is still too broad. Narrow it to one team or one kind of record and rate it again.

If you are still choosing which task to rate, the six-question project finder asks where your team's week goes and points to the kind of project to look at first, using fixed rules and no contact details.

When is an existing tool the better fit?

An existing tool is the better fit when your process is ordinary and the work is not where you compete. It also fits when the tool's assumptions match yours closely enough that nobody needs a spreadsheet beside it. Ordinary work often meets that test.

  • Somebody in your trade would recognize your process without explanation.
  • The work is necessary but not where you win business.
  • The tool's own way of doing it is acceptable, not merely tolerable.
  • Staff do the whole job in the tool, with no side record.
  • You can get your data out whenever you want it.

Look hard at the last two. The usual sign that a tool does not fit is not complaint. The sign is a spreadsheet, kept quietly beside the tool, holding the information the tool will not hold.

When does custom software earn its place?

Custom software earns its place when how you work is part of why customers choose you. It also earns it when no tool handles the shape of your work without a workaround, or when the workarounds already take more attention than a build would.

  • How you work is part of the offer, not an internal habit.
  • Staff maintain a parallel record because the tool cannot hold something essential.
  • The work spans several systems and nothing joins them properly.
  • Rules change often, so waiting on a supplier is a real constraint.
  • You need records to stay reachable and exportable on your own terms.

One caution applies. Each of those statements is also what somebody says when they are attached to an inefficient process. The test is whether a customer would notice the difference.

A rule for the hours side: add up the hours a month your staff spend on workarounds, such as the side spreadsheet, the retyping and the checking. Set them against the hours a month the builder estimates, in writing, for upkeep once the build is live.

Hours alone justify a build only when the workaround hours are clearly the larger figure and the work will run for years. When the two are close, the hours do not decide it, and the other tests in the table do. The hour-value tool turns a weekly count into hours a year.

What is the middle path between buying and building?

The middle path is keeping an existing tool and joining it to the other systems you run, so information moves between them without retyping. The fit then comes from the connection rather than from new software, and staff keep the tools they already know.

The middle path suits a business whose tools each do their own job well but do not yet pass information to one another.

  • The middle path in practice

    Picture an enquiry arriving on a website, landing in the system that holds customer records, appearing on somebody's list, and producing an invoice from the same details later. No part of that needs new software. The work needs three existing systems to agree about one customer, and the systems integration checklist sets out what that takes.

    The system holding customer records is often a CRM (customer relationship management system), the program that keeps every client and prospect with their calls, emails and deals.

    Joining it to the rest is what a CRM integration is, and Mavrin Labs takes that on through systems integration, the work a CRM integration consultant does.

  • Securing each join

    Each join is also a point to secure, because it usually runs through an API (application programming interface), the published doorway one program offers so another can read or send its data.

    The OWASP project on API security keeps a list of the ten most serious risks where systems meet, revised in 2023 (OWASP list of API security risks, checked Thursday, October 8, 2026).

    The last of those ten, API10:2023, is the unsafe consumption of APIs: what arrives from a joined system needs checking as carefully as what a stranger types into a form (the 2023 OWASP API security risks, checked Thursday, October 15, 2026). That is the main piece of work the middle path adds.

How do you define the first useful version?

Define the first useful version as the smallest thing that proves the result on real work. Name the one outcome it has to reach, the one group who will use it, and the one measure that will tell you whether it worked. Everything else waits until the first version has earned a second.

One outcomeA result, not a featureOne group, one kind of recordThe main path onlyExceptions go to a person for nowWrites to the systems youalready runBaseline and review dateRecorded before it goes live
The scope of a first useful version, from the result it must reach to the baseline it is judged against. Exceptions go to a person until the first version has earned a second.
  • One outcome, stated as a result rather than a feature.
  • One group of users, and one kind of record or request.
  • The main path only, with exceptions routed to a person for now.
  • The first version writes to the systems that already hold your data.
  • A baseline recorded before it goes live, and a review date agreed.

The baseline is the item most often skipped, and skipping it means nobody can say afterwards whether the build worked. The measurable results method sets out how the baseline and the review date are agreed.

What are the common mistakes in this decision?

The common mistakes in a build-or-buy decision come from judging too early or too broadly. They are deciding from a demo, buying a platform to fix a joining problem, building everything at once, and treating upkeep as finished. Each has a plain correction, set out below, and each is easier to avoid than to undo.

Deciding from the demo
A demo shows the main path at its best. Ask instead to see the awkward case you have, with your own kind of record.
Buying a platform to solve a joining problem
When the complaint is retyping, the fix is usually a connection between what you already run, not a new system for everyone to learn.
Building the whole thing first
A first version that covers every case cannot be judged. Build the smallest version that proves the result, then extend it.
Treating upkeep as finished work
Software you own needs somebody to own it. Decide who watches it and who decides on changes before the build, not after.

Three of the four come from deciding before the comparison has been tested against one real task. Time the task first, and the demo, the platform and the scope each have something to be measured against.

What shapes the investment in each path?

Three things shape the investment in any of these paths. They are how clearly the outcome can be stated, how many systems have to be reached, and how much of the awkward case the first version has to handle. The path itself matters less than those three.

A scope written after the first step is therefore worth more than a quote offered before it.

Mavrin Labs writes the scope once the task, the baseline and the target are clear, and the investment is agreed against that scope. The page on how an engagement is staged sets out each stage, and the five steps are Assess, Design, Build, Launch and Support.

How was this comparison put together?

Mavrin Labs built the comparison in this guide from the questions that change the answer in a build-or-buy decision, not from any product's feature list. Each criterion is one a buyer can check against a real task, and no product is scored or ranked. The table has no money column, because the investment follows the scope.

Mavrin Labs builds custom software and connects existing tools, so it has a stake in two of the three paths and chose criteria that work whichever path wins. The security row draws on the CISA and NIST guidance cited in this guide, and the other rows describe how each path works.

If the table points at a build or a join, custom software describes the service, and you can start a conversation about the task you have in mind.

How has guidance on buying software changed?

Guidance on buying software has moved from comparing features to asking whether the supplier builds securely. CISA first published its secure-by-design principles in April 2023 and updated them on October 25, 2023, asking makers to own customer security outcomes, be transparent and lead from the top (CISA Secure by Design guidance, checked Thursday, October 8, 2026).

The same agency's Secure by Demand guide turns those principles into questions an organization buying software can put to the manufacturer. The questions work whether you are choosing a tool or commissioning a build (CISA Secure by Demand guide, checked Thursday, October 8, 2026).

Sources behind this guide

The secure software development framework published by NIST, version 1.1 in its publication SP 800-218, groups its practices into four sets: preparing the organization, protecting the software, producing well-secured software and responding to vulnerabilities. Those four sets give a buyer a structure to hold any builder to (NIST secure software development framework, checked Thursday, October 8, 2026).

The two CISA documents above cover the buying side, and the agency's catalog of vulnerabilities known to be exploited covers the upkeep side. Where those sources say nothing, the guide relies on the criteria in the comparison table.

Each claim and its source

The table below pairs each claim the guide makes from outside evidence with that evidence and its publisher.

Each claim and its source
Claim Evidence Source
Most businesses see software as central to how they work Most businesses described specialized software and cloud-based technology as very important to their processes or methods, the figures given in the section on why it matters (2023 survey report, checked Saturday, October 10, 2026) Census Bureau
Better processes are the usual reason firms adopt software Improving the quality or reliability of processes or methods was the most common reason for adopting specialized software (49.8%) between 2020 and 2022 (technology adoption figures, checked Saturday, October 10, 2026) Census Bureau
A buyer can hold a maker to secure practices Secure-by-design principles ask makers to own customer security outcomes, be transparent and lead from the top CISA Secure by Design guidance
A builder's practices can be checked against a framework The secure software development framework groups its practices into four sets, from preparing the organization to responding to vulnerabilities NIST secure software development framework
The questions to ask a maker are already written down Secure by Demand groups them into areas covering authentication, intrusions, supply chain and disclosure CISA questions for software buyers
Upkeep includes watching a public list of exploited flaws The agency tells organizations to use its catalog as an input to deciding what to patch first the CISA catalog of known exploited vulnerabilities
Information from a joined system needs checking The last of the current API risks is the unsafe consumption of APIs the 2023 OWASP API security risks

Key takeaways

  • An existing tool wins when your process is ordinary.
  • Most fit problems are joining problems, so connect before you build.
  • A build earns its place when how you work is part of your advantage.
  • Rate the comparison against one real task, and check upkeep first.
  • Scope a first version to one outcome, one group and one measure.

If the choice ahead is wider than build or buy, the framework for all four paths covers them. If the fit problem turns out to be a joining problem, read the checklist for connecting systems. If the work you are considering varies case by case, the guide to agents for varied work is the relevant one.

A decision made this way gives hours back to the people doing the work, without replacing systems they already rely on.

Frequently asked questions

How do I know whether my process is unusual?

Test it by describing your process to somebody in the same trade. If they recognize it, an existing tool probably handles it; if they do not, write down where it differs and why that matters to your customers. A difference that matters is worth building around. A difference that is only habit is worth dropping.

Is a custom build riskier than buying a tool?

A custom build carries different risks rather than larger ones. Building puts the risk in scope and upkeep, which you can see, while buying puts it in the supplier: their roadmap, their terms and the features they may withdraw. The question to settle is which of the two your business is better placed to absorb.

What is the smallest version worth building?

The smallest version worth building is the one that proves the result on real work. It handles the main path for one team or one kind of record, writes to the systems that already hold your data, and leaves the exceptions to a person for now. The first version exists to answer a question, and nothing more.

Who owns custom software once it is built?

You own custom software built for you, which means you hold the accounts it runs on, you can read what it does, and another builder could take it over. Agree that in writing before the build. Ownership that depends on goodwill is not ownership, and it matters most once the relationship with the builder has ended.

What happens when my chosen tool is discontinued?

A discontinued tool forces a migration on somebody else's timetable, which is the main argument for keeping your own records reachable. Export your data regularly, know which system owns each record, and avoid arrangements where the only copy of something important lives inside one supplier. That discipline takes a little attention each month and saves a crisis you cannot schedule.

Can I start with a tool and build later?

Starting with a tool and building later is often the right sequence. Use the tool to learn what the work needs, keep your records in a form you can export, and avoid shaping the business around one supplier. The year spent with an existing tool is not wasted, because it produces the specification a later build needs.

The Mavrin Labs mark

Who writes these guides

Client work at Mavrin Labs is delivered by a team in mixed roles, led by our founder. AI tools help with research and drafting, and a person at Mavrin Labs checks each fact against its named source before a guide is published.

Enquiry form

Start with one workflow

Tell Mavrin Labs what is stuck and which systems it touches. A person at the firm reads every enquiry and replies by email.

Start a conversation

Fields marked with an asterisk are the ones a reply needs.

So a reply can address you properly.

The address the reply goes to.

Optional. Give one only if you would rather talk than write; the reply goes by email either way.

Optional, and useful when it says what the work has to fit around.

The result you want, and when you would want it working. Two or three sentences are plenty.

The task as it actually happens, roughly how often it runs, and which systems it touches. Two links at most, please.

Optional.

What you send goes to the Mavrin Labs inbox and is used to answer you, and a phone number, if you give one, is used for the same thing and nothing else.

A person reads it and replies by email.

If the form does not send, write to info@mavrinlabs.com instead; it reaches the same inbox.

Which hours would you take back first?

Name the one workflow that eats the most of your week. Mavrin Labs reads it, and the reply comes from the people who would build the fix.

Start a conversation