Business Strategy

By Fossilite Team

Published

10 November 2026

Read time

16 min read

Software Requirements Document: How To Write One Before You Hire A Developer

FOSSILITE GUIDESoftwareRequirementsDocumentHow to write one beforeyou hire a developerREQUIREMENTS, VERSION 1.0R1Coordinator creates a job withfour required fieldsMUSTR4Job closes only after a photoand a noteMUSTR6Customer gets an email withthe photoSHOULDR8Weekly report as a spreadsheetfileCOULDIllustrativeFOSSILITE GUIDESoftwareRequirementsDocumentHow to write one before you hire adeveloperREQUIREMENTS, VERSION 1.0R1Coordinator creates a job withfour required fieldsMUSTR4Job closes only after a photoand a noteMUSTR6Customer gets an email with thephotoSHOULDR8Weekly report as a spreadsheetfileCOULDIllustrative

You describe the project on a call, three firms send quotes, and the numbers are far apart. Or the software arrives months later and does what was said, not what was meant. Both problems usually have the same cause: nobody wrote down, in terms anyone could check, what the software had to do.

A software requirements document is a written list of what new software must do and how well it must do it, agreed before anyone builds it. It is also called a software requirements specification, or SRS. It lets every vendor price the same work, and it gives you a clear way to check the finished result.

Getting this wrong is expensive at any scale. In a survey of more than 2,000 project practitioners and business analysts, PMI found (opens in a new tab) that 5.1% of every dollar spent on projects and programs was wasted because of poor requirements management. The 2014 survey covered all kinds of projects, not only software, but the pattern is familiar to anyone who has paid for a rebuild.

This guide is for business owners and managers who are about to buy custom software and have never written a specification. It explains what to include, how to write requirements that can be tested, and how to set priorities. It ends with a free template and a filled-in example you can copy.

What Is A Software Requirements Document?

A software requirements document describes what the software must do, who will use it, the rules it must follow and how you will check that it works. It does not say how to build it, which belongs to the developer's design. Its job is to make your request precise enough to price, build and test.

You will see it under several names: software requirements specification, software requirement specification document, SRS or SRD. They mean the same thing. The formal reference is the international standard ISO/IEC/IEEE 29148 (opens in a new tab), last confirmed in 2024 and due to be revised. It is a detailed engineering standard, and a small business does not need its full outline.

The document sits early in a software project. Figure 1 shows where it fits and why it comes back at the end.

IN A SOFTWARE PROJECTWhere the requirements document fits1Problem andgoalsWhat is wrongand how you willmeasure success2RequirementsdocumentWhat thesoftware mustdo, and how well3RFP andquotesEvery vendorprices the samework4AgreementThe signedversion becomesthe scope5BuildChanges arenumbered, pricedand approved6AcceptanceEach requirementis checkedbefore you signoffChecked against the same requirementsFossilite | Conceptual diagram, not measured data.IN A SOFTWARE PROJECTWhere the requirementsdocument fits1Problem and goalsWhat is wrong and how you willmeasure success2Requirements documentWhat the software must do, and howwell3RFP and quotesEvery vendor prices the same work4AgreementThe signed version becomes the scope5BuildChanges are numbered, priced andapproved6AcceptanceEach requirement is checked beforeyou sign offChecked against the same requirementsFossilite | Conceptual diagram, not measured data.
Fig. 1: The requirements you write before quotes are the same ones you check against at the end.

Several other documents get confused with it. Table 1 shows what each one answers. In a small business, the business case and the software requirements usually live in one document, and that combined version is what this guide covers.

DocumentWhat it answersWho usually writes it
Business requirements document (BRD)Why the business needs the project and what it should achieveThe owner or project sponsor
Software requirements documentWhat the software must do, and how wellThe owner, often with an analyst or the vendor
Request for proposal (RFP)Who should build it, how, and at what priceThe buyer. See our software RFP template
Technical design documentHow the software will be builtThe development team

Table 1: Four documents that are easy to mix up.

When You Need One And How Detailed It Should Be

You need a software requirements document whenever someone else will build, price or test your software. Size it to the risk. One to three pages suits a small tool or a first release, while regulated data, several departments or a fixed-price contract call for more detail and a formal sign-off.

These situations call for the fuller version:

  • A fixed price. If the vendor carries the risk of overruns, they will price every gap in your description as risk.

  • Several departments. When sales, operations and finance all use the system, written requirements settle disagreements before they become rework.

  • Sensitive or regulated data. Customer records, payments and health information need written rules about who can see and change what.

  • Replacing an old system. You need to list what the old system does today, including the things nobody remembers until they stop working.

For a first release, keep it short and focused on one workflow. Our guide to what growing businesses build first can help you choose that starting point. A clear document also makes the schedule easier to predict, and our custom software development timeline guide explains what else sets the pace.

Who Writes It: You, The Vendor, Or Both

The business owner should write the business side: the problem, the users, the rules, the exceptions and the priorities. A developer or business analyst can then sharpen the wording and add technical detail. You read every line before sign-off, because the document defines what you are paying for.

A business analyst is someone whose job is to turn how a business works into instructions a development team can build from. Some software firms offer a paid discovery phase in which an analyst writes the document with you. That can work well, as long as you stay involved and the finished document belongs to you, so you can take it to another firm if you need to.

If you have no technical staff, the choice between building a team and hiring a firm affects who does this work. Our comparison of in-house and outsourced development covers that decision.

How a firm handles requirements also tells you a lot about it. One that asks about your exceptions before quoting is usually more reliable than one that quotes from a five-minute call. Our guide to choosing a custom software development partner lists other things to check.

What To Include In A Software Requirements Document

A useful software requirements document has ten parts: purpose and scope, users and roles, how the work happens today, functional requirements, non-functional requirements, data and integrations, business rules, priorities, acceptance checks, and what is out of scope. Add a short list of assumptions and open questions so nothing important stays unsaid.

Table 2 lists each part with an example line. The template later in this guide turns them into wording you can copy.

PartWhat to writeExample line
Purpose and scopeWhat the software is for, how you will measure success, and what this version covers"Replace paper job sheets so every job is tracked from booking to completion."
Users and rolesEach type of user, how many, and what they need to do"4 office staff create jobs. 24 technicians update them on their phones."
How the work happens todayThe current steps, including the awkward exceptions"If a part is missing, the technician leaves and the job is rebooked by phone."
Functional requirementsWhat the software must do, one numbered line each"R4: A technician can mark a job complete only after adding a photo and a note."
Non-functional requirementsHow well it must work: speed, availability, security, devices"The job list loads within 3 seconds on a phone with a 4G connection."
Data and integrationsWhat information it holds and which systems it connects to"Customer names and addresses are entered by the office. Invoicing stays in the current accounting software."
Business rulesCalculations, approvals and limits that must always apply"Jobs over the agreed price need office approval before work starts."
PrioritiesMust have, Should have, Could have or Won't have this time"R6: Should have."
Acceptance checksHow you will confirm each important requirement works"Given a job with no photo, when the technician taps Complete, then the job stays open."
Out of scopeWhat this version will deliberately not do"No customer portal in this release."

Table 2: The ten parts of a software requirements document, with illustrative example lines.

If the new software replaces an old system, list the data that has to move across and who owns it today. Our guide to legacy software modernization covers the questions that come up when old and new systems overlap. Budget, deadlines and vendor questions belong in your request for proposal, not here.

Functional And Non-Functional Requirements

Functional requirements say what the software does: a technician marks a job complete, the system emails the customer. Non-functional requirements say how well it must do it, covering speed, security, availability and who can see what. A system can have every feature you asked for and still fail if it is slow or unsafe.

Functional (what it does)Non-functional (how well)
The coordinator assigns a job to a technicianThe technician's phone shows the job within 1 minute
A technician uploads a photo of the finished workPhotos are stored for at least 2 years
The office sees all open jobsOnly office staff can see customer phone numbers
The app works on technicians' phonesIt keeps working with no signal and sends updates when the connection returns

Table 3: Functional and non-functional requirements side by side. The numbers are illustrative.

Owners often skip the non-functional side because it sounds technical. It does not need to be. Describe what you would notice as a user, and put a number on it. "Fast" means nothing to a developer, but "loads within 3 seconds on a phone" can be tested. If you do not know the right number, write your best guess and mark it as an open question for the vendor.

How To Write Software Requirements That Can Be Tested

Write each requirement as one numbered sentence that names who does what. Replace words like fast, easy and secure with something you can see or count, include the awkward cases, and pair every important requirement with an acceptance check written before building starts.

Six rules cover most of the problems:

  1. One requirement per line, each with a number. R1, R2 and so on let everyone refer to the same thing in meetings, quotes and change requests.

  2. Name the person. Write "a technician can", not "the system allows". It makes clear who the requirement serves.

  3. Use numbers instead of adjectives. Fast, easy, user-friendly and secure mean different things to different people.

  4. Write down the exceptions. What happens with no signal, a cancelled job or a missing photo? Exceptions are where most of the building time goes.

  5. Avoid escape words. "If possible", "etc.", "and/or" and "as needed" leave room for two different readings.

  6. Describe the result, not the screen. Leave layout to the designer unless it truly matters to the work.

Figure 2 shows how those rules come together in a single requirement.

ONE TESTABLE REQUIREMENTAnatomy of a good requirementR4 A technician can mark a job complete only after adding at least onephoto and a short note.MUST1IDR4A number everyone can refer to2WhoA technicianThe person the requirement serves3Whatcan mark a job completeThe action, not the screen4Condition you can checkonly after adding at least onephoto and a short noteSomething you can see or count5PriorityMust haveMust, Should, Could or Won't thistime6Acceptance checkGiven a job with no photo, whenthe technician taps Complete,then the job stays openWritten before building startsFossilite | Example from the illustrative job app in this guide.ONE TESTABLE REQUIREMENTAnatomy of a good requirementR4 A technician can mark a job completeonly after adding at least one photo anda short note.MUST1IDR4A number everyone can refer to2WhoA technicianThe person the requirement serves3Whatcan mark a job completeThe action, not the screen4Condition you can checkonly after adding at least one photo and ashort noteSomething you can see or count5PriorityMust haveMust, Should, Could or Won't this time6Acceptance checkGiven a job with no photo, when the techniciantaps Complete, then the job stays openWritten before building startsFossilite | Example from the illustrative job app in this guide.
Fig. 2: The parts of one testable requirement, using a line from the example later in this guide.

The fastest way to improve a draft is to rewrite its vaguest lines. Table 4 shows five common examples rewritten so they can be tested.

VagueTestable
The app should be fast.The job list loads within 3 seconds on a phone with a 4G connection.
Users can manage jobs.An office coordinator can create, edit and cancel a job. A technician can update only the jobs assigned to them.
The system must be secure.Technicians see only their own jobs. Customer phone numbers are visible only to office staff.
Send notifications if possible.When a job is assigned, the technician gets a phone notification within 1 minute.
Easy reporting, etc.The coordinator can download a weekly list of completed jobs per technician as a spreadsheet file.

Table 4: Five requirements before and after. The numbers are illustrative.

Write the acceptance check at the same time

An acceptance check is the test you will use to decide whether a requirement is done. A common format is Given, When, Then, which the Cucumber project (opens in a new tab) uses in its Gherkin language for examples that double as tests. Given sets the starting situation, When is the action, and Then is the result you should be able to see.

R4 acceptance check.

Given a job with no photo attached, when the technician taps Complete, then the app asks for a photo and the job stays open.

You do not need special software to use this format. Written in plain sentences, it gives you and the vendor the same definition of done.

How To Prioritize Requirements

Sort every requirement into Must have, Should have, Could have and Won't have this time. This method, called MoSCoW, is part of the DSDM framework published by the Agile Business Consortium. A requirement is a Must have only if the release would be pointless, illegal or unsafe without it. If any workaround exists, it belongs lower.

The Agile Business Consortium (opens in a new tab) suggests a simple test: ask what happens if the requirement is not met. If going live would be pointless, it is a Must have. If people could manage with a manual workaround, even a painful one, it is a Should have or a Could have.

It also recommends that Must haves take no more than about 60% of the effort, with around 20% kept for Could haves. That is guidance, not a rule, and it is measured in work, not in the number of lines. The point is to leave room: if something takes longer than expected, the Could haves are dropped first so the Must haves can still arrive on time.

SETTING PRIORITIESMoSCoW, sized by effortMust haveUp to about 60%Should haveThe restCould haveAbout 20%Won't havethis timeTHIS RELEASEMust haveNo point going livewithout it. If aworkaround exists, it isnot a Must.Should haveImportant, but the releasestill works with aworkaround.Could haveWanted, and the firstthing dropped if time runsshort.Won't have this timeOutside this release.Written down so it doesnot creep back in.Source: Agile Business Consortium, DSDM Handbook. Guidance, not a rule.SETTING PRIORITIESMoSCoW, sized by effortMust haveUp to about 60%No point going live without it. If a workaroundexists, it is not a Must.Should haveThe restImportant, but the release still works with aworkaround.Could haveAbout 20%Wanted, and the first thing dropped if time runsshort.Won't have this timeOutside this release. Written down so it does notcreep back in.Source: Agile Business Consortium, DSDM Handbook. Guidance, not arule.
Fig. 3: MoSCoW priorities, sized by effort as the Agile Business Consortium recommends.

The Won't have this time list matters as much as the others. Writing down what a release will not do stops ideas from creeping back in halfway through the build. It is also the simplest way to control cost, as our guide to MVP development cost explains.

Software Requirements Document Template

Copy the template below into your own document and replace everything in square brackets. Keep each requirement to one line with its own number, priority and acceptance check. Delete sections that do not apply, and keep the whole document as short as the project allows, usually one to three pages for a first release.

Software Requirements Document: [Project name]

Version [1.0], [date]. Owner: [name]. Approved by: [name].

1. Purpose and scope

We need [one sentence on what the software does] so that [business result].

We will measure success by [measure and today's number].

This version covers [first release]. It does not cover [list].

2. Users and roles

[Role]: [how many people], [what they need to do]

[Role]: [how many people], [what they need to do]

3. How the work happens today

[The steps in order, including the exceptions that happen most often.]

4. Functional requirements

[R1] [Who] can [do what] [under what condition]. Priority: [Must / Should / Could]

Acceptance check: Given [starting situation], when [action], then [result you can see].

5. Non-functional requirements

Speed: [for example, main screens load within X seconds on a phone]

Availability: [when it must work, and how much downtime is acceptable]

Access and security: [who can see and change what]

Devices: [phones, tablets, browsers]

6. Data, integrations and business rules

Data it holds: [types of information, and any privacy rules]

Connects to: [systems, and who controls access to each]

Rules: [calculations, approvals and limits that must always apply]

7. Won't have this time

[Things deliberately left out of this release]

8. Assumptions and open questions

[What you are assuming, and what still needs an answer, with an owner]

9. Changes

Changes are requested by [name], priced by [vendor] and approved by [name] in writing before work starts. Each approved change raises the version number.

Template: Copy-and-paste software requirements document.

A Software Requirements Example: A Maintenance Company's Job App (Illustrative)

This software requirements example is hypothetical. It shows how a 30-person property maintenance company might describe a job-tracking app for its field technicians, replacing paper job sheets and a busy WhatsApp group. It is not a real Fossilite project, and every number in it is an assumption chosen to show the format.

Today, the office takes calls, writes jobs on paper sheets and sends them to technicians over WhatsApp. Photos of finished work get lost in the chat, customers call to ask whether a job is done, and nobody can say quickly which jobs are overdue. The owner's goal for the first release is simple: every job tracked from booking to completion, with proof of the finished work.

IDRequirementPriorityAcceptance check
R1An office coordinator can create a job with the customer's name, address, problem description and due date.MustGiven a new call, when the coordinator saves a job with all four fields, then it appears on the open jobs list.
R2The coordinator can assign a job to a technician, who gets a phone notification within 1 minute.MustGiven an unassigned job, when it is assigned, then the technician's phone shows it within 1 minute.
R3A technician can see today's jobs in time order, and tapping an address opens the phone's map app.MustGiven three jobs today, when the technician opens the app, then they appear earliest first.
R4A technician can mark a job complete only after adding at least one photo and a short note.MustGiven a job with no photo, when the technician taps Complete, then the app asks for a photo and the job stays open.
R5With no signal, the app saves the technician's updates and sends them when the connection returns.MustGiven a phone with no signal, when a job is completed, then the office sees it within 2 minutes of the signal returning.
R6The customer gets an email with the photo when their job is marked complete.ShouldGiven a completed job, when it is saved, then the customer's email arrives with the photo attached.
R7The coordinator can see every job that is more than one day overdue.ShouldGiven a job due yesterday, when the coordinator opens the overdue list, then it is shown.
R8The coordinator can download a weekly list of completed jobs per technician as a spreadsheet file.CouldGiven a finished week, when the report is downloaded, then it opens in the company's spreadsheet software.
R9The app suggests a job category from the problem description. The coordinator must confirm it before saving.CouldGiven 100 past jobs, when the app suggests a category for each, then at least 80 match the coordinator's choice and none is saved without confirmation.

Table 5: An illustrative set of requirements for a technician job app.

The document would also include three non-functional lines: the job list loads within 3 seconds on 4G, technicians see only their own jobs, and the app works on the phones the company already issues. Its Won't have this time list names a customer portal, invoicing (which stays in the current accounting software) and automatic route planning.

R9 shows how to write a requirement for an AI feature. State how often it may be wrong, how you will measure that, and where a person checks the output before it counts. If those three things are missing, the requirement cannot be tested.

Because technicians work from their phones, mobile experience matters for this project. If you want to see how a software team describes this kind of work, SensViz lists its mobile app development (opens in a new tab) services on its site.

Common Mistakes In Software Requirements Documents

Four mistakes cause most of the trouble in software requirements documents: describing screens instead of the work, marking everything as a Must have, skipping non-functional requirements and never updating the document. Each one shows up later as a disagreement, a change request or a missed deadline, and each is easy to avoid while drafting.

  • Describing screens instead of the work. A drawing of a screen tells the developer what you imagine, not what you need. Describe the task and let the designer propose the screen.

  • Everything is a Must have. If every line is essential, nothing can be dropped when time runs short, so the deadline moves instead.

  • No non-functional requirements. Speed, access and availability are easy to leave out and expensive to add after launch.

  • Writing it once and filing it. A requirements document that nobody updates stops describing the software within weeks.

How To Sign Off And Handle Changes

Sign off the software requirements document before development starts, give it a version number, and agree in writing how changes will be handled: who can request one, how it is priced and who approves it. Requirements will change. The goal is not to freeze them but to make each change visible and agreed.

Catching problems early is cheaper. A NASA study (opens in a new tab) found that a requirements error costing 1 unit to fix during the requirements stage cost 3 to 8 units to fix during design, 21 to 78 during integration and testing, and from 29 to more than 1,500 once the system was in use. The study covered large hardware and software systems such as spacecraft and military aircraft, not business apps, so treat the size of the numbers with care. The direction holds for almost any project.

A simple change process has four steps:

  1. Sign off version 1.0 before development starts.

  2. Give every change request a number, a one-line reason and the requirement it affects.

  3. Ask the vendor for its effect on price and dates before any work begins.

  4. Approve or decline it in writing, and raise the document's version number.

Your contract should refer to the signed version. Our guide to the software development agreement explains how scope and acceptance are written into it. How changes are paid for depends on the type of contract, and our comparison of time and materials vs fixed price explains the difference.

How Fossilite Approaches Requirements

Fossilite starts with how the work happens today, not with a feature list. We look at what existing tools can already do, write requirements around the real workflow and its exceptions, and recommend custom software or AI only where it is justified. Every requirement we agree should be one you can check.

Sometimes that means the document gets shorter, because a tool you already pay for covers half of the list. That is a good outcome. The requirements that remain are the ones worth building.

Frequently Asked Questions

These are the questions business owners ask most often about software requirements documents. Each answer stands on its own.

What is a software requirement specification?

A software requirement specification, or SRS, is a document that describes what a piece of software must do and the standards it must meet, such as speed, security and who can use it. It is the same thing as a software requirements document. Developers use it to estimate, build and test, and buyers use it to check the result.

What is the difference between functional and non-functional requirements?

Functional requirements describe what the software does, such as sending a confirmation email when a customer books. Non-functional requirements describe how well it does it, such as how fast pages load, how much downtime is acceptable and who can see which data. A complete document needs both, written so each one can be checked.

How long should a software requirements document be?

As long as the project needs and no longer. A first release or small internal tool often fits in one to three pages. A system used by several departments, or one that holds regulated data or has a fixed price, usually needs more. Precision matters more than length: two vendors should be able to price each requirement the same way.

Who should write a software requirements document?

The business owner or manager who understands the work should write the business side: the problem, users, rules and priorities. A developer or business analyst can sharpen the wording and add technical detail. Whoever drafts it, the buyer should read and sign off every line, because the document defines what is being paid for.

Is a software requirements document the same as an RFP?

No. An RFP, or request for proposal, asks vendors how they would solve your problem and what they would charge, and explains how you will choose between them. A software requirements document describes in detail what the software must do. Attaching a short requirements document to your RFP helps every vendor price the same work.

Can requirements change after development starts?

Yes, and they usually do once people see working software. What matters is changing them on purpose. Give each change a number, ask the vendor for its effect on price and dates, approve it in writing and update the document's version. Changes that nobody records are where budgets and deadlines quietly slip.

What To Do Next

Once your requirements are written, the next step is asking vendors to quote against them. Our software RFP template guide gives you a copy-and-paste request to send alongside your software requirements document.