In September 2026 we read eight of the guides that rank for this decision. Five published no cost figures at all. Of the three that did, the percentages carried citations and the figures in pounds, dollars and months didn’t: no derivation, no region, no year. One priced the custom option per year and the alternative per month in the same table, and never multiplied the two into the same unit.
So the useful thing left to do is the one nobody has done. Take a single requirement, price it three ways, in one currency, over the same five years, using prices the vendors publish themselves, and show every step of the arithmetic so you can argue with it.
That’s this article. It won’t tell you that custom software is better, because on the numbers below it often is not. It will give you a page you can take to a board meeting.
The One Requirement We Are Going to Price
Everything below prices the same thing: a trade-customer portal for a 140-person UK equipment-hire business. Four hundred trade accounts need to raise quotes, book kit and view invoices without ringing the hire desk. Sixty internal staff across four branches need seats. Five-year horizon, sterling, VAT excluded.
Two numbers in that paragraph do all the work, and most comparisons price only one of them. Sixty is your staff count, which is what a subscription multiplies. Four hundred is your customer count, which is what a white-label platform multiplies and what a subscription usually licenses separately. A build multiplies by neither, which is its whole economic argument and also the reason it can be the wrong answer.
One honest caveat before the figures. The published prices below come from real vendors, but none of those vendors sells an equipment-hire portal. They are anchors for what each pricing model does to a bill over time, not a claim that any named product does this job. Substitute your own shortlist and the shape of the answer won’t change.
What Build, Buy and White-Label Actually Mean Here
Build means commissioning software that belongs to you. Buy means subscribing to a product somebody else runs and configuring it. White-label means licensing somebody else’s product, putting your brand on it, and selling or giving it to your customers as though it were yours. The differences that matter are not features. They are who sets next year’s price, and who can leave.
Two words get used interchangeably and should not be. White-label means the vendor’s product with your branding on the surface, and the vendor still runs it. Private label usually implies a deeper arrangement where the product is made to your specification and sold only under your name. OEM sits closer to embedding somebody else’s engine inside a product that’s otherwise yours. Reselling is none of the above, since the customer knows whose product it is. If a contract calls itself white-label, check which of those four it actually describes, because the exit terms differ sharply.
There is also a fourth route the frameworks usually call partner or extend: buy the boring platform, then build the thin layer that’s actually yours on top of it. It rarely gets its own column because it’s harder to sell as a package, and it’s frequently the right answer. It will show up again in the arithmetic.
Year One: What Each Route Costs to Get Live
On published list prices, year one runs from about £52,800 to about £150,720 for a subscription once you licence both the staff and the trade accounts, against £73,000 to £128,000 for a custom build including the audit week. White-label starts near $12,000 a year and can’t be totalled from a price list at all, for a reason worth understanding.
Start with buy, because third-party vendors rarely publish sterling prices and one does. Salesforce publishes UK list prices for Sales Cloud: a free edition, then Starter at £20, Pro at £80, Core at £156, Advanced at £316 and Max at £440 per user per month, with only the £20 edition billable monthly and every edition above it billed annually. The page notes that it’s for information and subject to change.
Take Core, the middle paid edition, at £156. Sixty staff seats is £9,360 a month, which is £112,320 a year at list. Pro at £80 gives £57,600 a year for the same sixty people. Starter at £20 gives £14,400. That is a £97,920 annual spread across three editions of one product, for an identical headcount.
Then price the part almost every comparison forgets. Four hundred trade accounts are not staff seats, and external users are usually licensed separately. The same Salesforce page publishes partner relationship management from £8 per login per month, which for four hundred accounts is £38,400 a year. Add that to each edition and year one becomes £52,800 on Starter, £96,000 on Pro and £150,720 on Core. The external users cost more than the staff seats on the cheapest edition.
Read those two spreads before anything else. Which edition your requirement lands on is decided by two or three features you’ll discover in month two rather than in the demo, and whether your customers need licences at all is decided by which product you shortlist. Both matter more than build against buy.
Now build. Empyreal Infotech publishes a fixed £8,000 audit week, and most multi-tenant builds we sign land between £65,000 and £120,000 on a ten to sixteen week fixed-price sprint. So year one is roughly £73,000 to £128,000, with the software live at the end of it, the code in your own repository, and no per-user line at all.
White-label looks cheapest and is the only column that can’t be totalled honestly. Vendasta publishes a subscription that bills at zero in any month your qualifying spend on its own products meets a committed minimum: $99 on Starter with no contract, $499 on Professional and $999 on Premium, both on one-year contracts. You pay the fee at signing, third-party purchases don’t count toward the minimum, and if you don’t spend you pay the subscription. White-labelling, meaning your logo in the app and the vendor’s name removed, starts at Professional, so the real entry point is $499 a month committed for a year.
HighLevel publishes three agency tiers, of which it prices two: Unlimited at $297 a month and Agency Pro at $497. Rebilling with your own markup is only available on the $497 plan. A white-label mobile app is $497 a month, and its branded client portal app, a branded mobile app for a sub-account’s community members, is $49 a month per enabled sub-account.
Now try to total it. The agency plan and the white-label app come to $11,928 a year. Every further enabled account adds $588 a year. At one account the bill is about $12,000. At four hundred, the same published prices give roughly $247,000 a year, which would make white-label the most expensive route on this page rather than the cheapest. Whether your four hundred customers map to four hundred billable accounts depends entirely on how that platform defines an account, and that’s the point: the published prices don’t let you work out your own bill.
Three other things spoil the cheap first impression. The fees are in dollars while the build quote is in pounds, and nobody converts them. Neither published price list contains a revenue-share percentage, because these platforms don’t sell one: you buy wholesale and mark up, so your margin is whatever the vendor leaves you next year. And the commitment runs the wrong way, since the white-label tier is the one carrying the twelve-month contract.
Year Five: Where the Three Lines Cross
Over five years at list prices, licensing sixty staff and four hundred trade accounts comes to £264,000 on the cheapest edition, £480,000 on the next and £753,600 on the middle one. A custom build across our whole published band, including the audit week and four years of maintenance, lands between £112,000 and £248,000. On these figures the build is the cheaper column, and the honest thing to say next is that this result flatters the company writing it.
Here is the working, so you can break it. Staff seats: £112,320 a year at £156, £57,600 at £80, £14,400 at £20. Trade accounts: 400 logins at £8 a month is £38,400 a year. Added together that’s £150,720, £96,000 and £52,800 a year, so over five years £753,600, £480,000 and £264,000, before a single renewal uplift.
For the build, run both ends of our published band. At the bottom, £8,000 plus £65,000 is £73,000 in year one. Maintenance is the line nobody agrees on, so treat it as a dial: at fifteen per cent of build cost a year that’s £9,750, and four further years give £112,000. At the top of the band, £8,000 plus £120,000 is £128,000, and four years at twenty-five per cent adds £120,000 for £248,000. A midpoint build of £92,500 sits between, at £156,000 to £193,000. Pick your own percentage and redo it in thirty seconds.
Neither column includes your own people. A subscription needs an administrator, an integration owner and somebody who negotiates the renewal. A build needs somebody who owns the backlog. Put the same internal cost on both sides or leave it off both, but don’t let a proposal charge it only to the build.
The build column is missing two more things, and both are ours to declare. It excludes hosting and third-party services, which a subscription price includes. And it excludes our own optional advisory retainer at £5,000 a month, which our pages describe as six to nine months in one place and six to fourteen in another. Take the longer reading, add £45,000 to the build column and a hosting line on top, and the cheapest build is nearer £160,000 than £112,000. It still wins here. It wins by less.
What Would Flip This Result
The build column wins above, and we sell builds, so here are the four things that would reverse it. Each is checkable in a morning.
- Your customers might not need licences. Plenty of vertical products include a customer portal at no extra charge. If yours does, £192,000 leaves the buy column over five years and the cheapest edition wins outright at £72,000.
- The edition you land on. Between Starter and Core the same sixty staff differ by £97,920 a year. Find out in a pilot which edition your must-haves actually require, not in the sales deck.
- The build overrunning. Every figure above assumes a fixed price that holds. A build that doubles moves straight past the mid editions, which is the risk a fixed-price scope exists to remove.
- The renewal uplift. Every subscription figure assumes today’s price for five years, which no subscription does. That one moves the numbers the other way, so ask for the cap in writing as a percentage before you sign year one.
What Happens to the Price When You Grow
Growth prices the three routes very differently. A subscription bill grows with headcount, so hiring twenty people adds twenty seats at the tier rate. A white-label bill often grows with customer accounts, so winning customers costs you money before it makes you any. A build’s cost of growth is mostly infrastructure, which at this size is usually the smallest line on the page.
Run the hire-business forward. Sixty staff becomes ninety after two good years. On Core that’s another £56,160 a year, every year, for thirty logins. On Pro it’s £28,800. Win fifty more trade customers at £8 a login and that’s another £4,800 a year on top. The build doesn’t notice either number, because nothing in it’s priced per person, though its hosting line does move a little.
The white-label version is the one people model least and regret most. If the platform charges per enabled account and you’re handing the portal to four hundred trade customers, every new customer switches on another charge. That can still be excellent economics if you’re charging those customers for access. It’s poor economics if the portal is a service you give away to keep them loyal, which is exactly what a hire business is doing.
So the growth question isn’t which is cheaper today. It’s which of your numbers is going up fastest, and whether the route you’re choosing multiplies by that number.
The Question This Comparison Never Asks: What Does It Cost to Change Your Mind?
Every route is reversible, at very different prices. Leaving a subscription costs you the export, the data you can’t get out in a usable shape, and the rebuild of every integration. Leaving a white-label platform costs you the product itself, because your customers were using the vendor’s software with your logo on it. Leaving a custom build costs you the least, provided you own the code.
Nobody writes this section because nobody selling you the decision wants you thinking about the exit on the day you sign. It’s the most useful question in the whole comparison, and it takes three sentences to ask: what do I get on the last day, in what format, and how long do you keep it after I stop paying?
Here’s what each route actually hands back.
- Buy: an export and a deadline. Ask what the export contains, whether history and attachments come with the records, whether a rate limit makes a bulk export take weeks, and how long the data is kept after termination.
- White-label: often nothing you can run, because the software was never yours and your customers’ accounts live in the vendor’s tenancy. That’s not a reason to avoid it. It’s a reason to know before you put your brand on it.
- Build: the repository, the running system and the documents that explain it, assuming the contract assigned ownership as the work was done rather than at the end.
For build, the answer is checkable in a sentence. Our own position is that your code lives in your GitHub organisation from the first commit and the intellectual property assigns on commit rather than on final payment, with a handover pack of architecture brief, decision records, runbook, on-call playbook and data processing agreement. The test to apply to any development company, including us, is simple: if you vanished tomorrow, what would we have? A specific answer means a real one.
White-Label’s Quiet Clause: A Processor You Did Not Choose
If you put your brand on somebody else’s platform and your customers’ data goes into it, you’re likely to be a controller and the platform likely to be your processor. The roles turn on who decides the purposes and means of the processing rather than on whose logo is on the screen, and many platforms are controllers in their own right for part of what they do, which can make you joint controllers. Either way the relationship has to be governed by a contract, and any further supplier the platform uses becomes a sub-processor with obligations you never negotiated.
The ICO sets out what such a contract has to contain, in guidance that is currently under review following the Data (Use and Access) Act. Its guidance on controller and processor contracts says the processor should not engage a sub-processor without the controller’s prior specific or general written authorisation, that it must impose the same Article 28(3) obligations on that sub-processor, wording that need not mirror the contract exactly but must offer an equivalent level of protection, and that the processor remains liable to the controller for the sub-processor’s compliance. Where authorisation was general, it should tell the controller about intended changes and give it a chance to object.
In plain terms: your customer complains to you, your regulator asks you, and the software is three companies away from your office. That’s manageable, and thousands of firms manage it. What isn’t manageable is discovering the chain after an incident. Ask for the list of sub-processors before signing, ask how you are notified when it changes, and ask what happens if you object. None of this is legal advice, and a solicitor should read the actual contract, but the questions cost nothing to ask.
The commercial version of the same clause is shorter. The vendor sets the wholesale price, the feature roadmap and the deprecation schedule. If your margin is the gap between their price and yours, your margin is theirs to change.
What an Acquirer’s Diligence Team Asks About Each Route
If you ever raise or sell, a technical diligence team will look at this decision and ask three things: what do you own, what can you not lose, and what happens to the customer relationship if a supplier walks away. None of the eight guides we read mentions diligence at all, which is odd, because it’s the moment the decision is scored by somebody who isn’t selling you anything.
A custom build scores well if, and only if, the ownership is clean. Assignment of intellectual property on commit rather than on final payment, dependencies that are open source or owned by you, and decision records that let a stranger understand the architecture. It scores badly if the code sits in an agency’s account, if a licence somewhere is non-transferable, or if nobody can explain why the system is shaped the way it is.
A subscription is usually neutral, and occasionally a problem: check whether your contract survives a change of control, because some don’t, and an acquirer doesn’t enjoy discovering that the platform running operations can be renegotiated the week the deal closes. A white-label arrangement gets the hardest questions, because the product your customers use isn’t yours, the contract may be terminable, and the revenue attached to it may not transfer.
Source Code Escrow, and When It Earns Its Fee
Escrow is an arrangement in which an independent third party holds a copy of the source code and releases it to you if defined events occur, typically the supplier’s insolvency or a failure to meet support obligations. It costs an annual fee and it’s the standard answer to the fear this whole genre raises and never resolves: what if the company behind my software disappears.
It’s worth the fee when you depend on software you didn’t write and could not rebuild quickly, which usually means a bought platform running something operationally critical. It’s worth much less on a custom build where you already hold the code. Vendors selling one product to many customers often decline a single-beneficiary arrangement, though multi-beneficiary escrow exists for exactly that case, so ask which they offer and what the annual fee is. Expect a figure rather than a range.
Four Questions That Settle It in an Afternoon
You don’t need a scoring matrix. You need four answers, and each one points at a route rather than a preference. Write them down with numbers, and the decision usually makes itself before the shortlist calls are booked.
- Is this thing how you compete, or is it plumbing? If a customer would notice and care that yours works differently, that’s a build. If it’s invoicing, buy it, and spend the money on the part customers notice.
- Which number is growing fastest: seats, customers or transactions? Subscriptions multiply seats, white-label platforms often multiply customer accounts, builds multiply neither. Choose the route that doesn’t multiply your fastest-growing number.
- What tier does your must-have list actually require? Get this from a pilot, not a demo. On published prices the gap between the entry and middle editions was £97,920 a year for sixty staff seats, which is larger than most of the arguments people have about this decision.
- What do you get on the last day? Ask each supplier what arrives when you leave, in what format, and how long they hold your data afterwards. A supplier who answers well has been asked before, which is itself information.
If the answers split, which they often do, that is the partner-and-extend route telling you it exists. Buy the platform for the plumbing, build the thin layer that is genuinely yours, and keep the boundary between the two written down so that either half can be replaced without the other.
If you would like a second opinion on a shortlist you already have, send us the requirement in five lines and we’ll tell you which column we think it belongs in, including when that column isn’t ours.
How Empyreal Infotech Prices a Custom SaaS Build
Our prices are published rather than quoted by mood. The audit week is fixed at £8,000. After that every build is line-itemed into a scope document covering workspace and authentication, billing, core flows, admin tooling, observability and the handover pack, so you can see the cost of each piece and cut any of them. Most multi-tenant builds we sign land between £65,000 and £120,000 on a ten to sixteen week fixed-price sprint.
Ownership is the part worth checking against the column above. Your code lives in your own repository from the first commit, intellectual property assigns on commit rather than on final payment, every dependency is open source or owned by you, and the handover pack includes the architecture brief, the decision records, a runbook, an on-call playbook and a data processing agreement. If we vanished tomorrow you would have a working repository and the documents explaining every meaningful call.
There is an optional advisory retainer at £5,000 a month for a senior voice on your engineering calls, cancellable with thirty days’ notice in any month. Our own pages describe the typical run as six to nine months in one place and six to fourteen in another, so budget for the longer one. The broader SaaS development services page sets out what the build itself includes.
Now the part that costs us work. If your external users turn out not to need licences, the entry edition costs £72,000 over five years against £112,000 for the cheapest build we would quote, and buying wins on price without an argument. That’s the single assumption most likely to decide this, and it is checkable in one call with the vendor. Our published promise is that if we cannot hit your budget we say so in week one and you walk away with the audit brief and no commitment, and the same applies when the arithmetic points at buying. The honest case for a build starts when your customers need licences too, when the edition you actually need is a middle one, or when the thing being built is the reason customers choose you.
FAQ: Build, Buy or White-Label
What is the third option in the build versus buy decision?
There are two. Partner and extend means buying a platform for the commodity work and building only the layer that differentiates you. White-label means licensing a whole product and putting your brand on it. Partner and extend suits companies with one unusual requirement. White-label suits companies selling software to their own customers.
What is the difference between white label and private label software?
White label is the vendor’s product with your branding applied, still built and run by them for many customers. Private label usually means a product made to your specification and sold only under your name. OEM means embedding another supplier’s component inside your own product. Reselling means the customer knows whose software it is.
What is the total cost of ownership for build versus buy?
On the published figures in this article, sixty staff seats over five years cost £72,000 on the entry edition, £288,000 on the second and £561,600 on the middle one, and licensing four hundred external users adds £192,000 to each. A custom build across our published band lands at £112,000 to £248,000 with maintenance. Add your own internal staff cost to both sides, and hosting to the build side.
Is custom software always more expensive than off-the-shelf?
No, and it is not always cheaper either. Custom costs more on day one and stops multiplying after that, while a subscription costs less on day one and multiplies by seats or accounts for as long as you use it. The crossover depends on your seat count, your growth and which pricing tier your requirements need.
How much does it cost to build a SaaS platform in the UK?
Empyreal Infotech publishes £8,000 for a fixed audit week, with most multi-tenant builds landing between £65,000 and £120,000 over ten to sixteen weeks. Prices elsewhere vary widely and are rarely published at all, so ask for a line-itemed scope rather than a single number you can’t interrogate.
What to Take to the Board
One page. On it: the requirement in a sentence, your seat count and your customer count, three five-year totals with the arithmetic visible, the tier assumption behind the subscription figure, the maintenance percentage behind the build figure, and the answer each supplier gave to what arrives on the last day.
That page is unusual not because the sums are hard but because the industry that sells this decision doesn’t produce it. Of the eight guides we read in September 2026, five published no cost figures, and none of the three that did showed how a single one was derived. A board can argue with arithmetic. It can’t argue with an adjective.
If the numbers point at buying, buy, and put the money you saved into the part of the business customers actually choose you for. If they point at building, send us five lines about the requirement and Mohit will reply inside 24 hours on weekdays with a clear yes, a clear no, or the one question that decides it.
Price the same requirement three ways. The column that wins will surprise somebody in the room, and that is the point of doing it.