Why two honest quotes can differ threefold
You send the same two-page brief to three agencies. The quotes come back at £12,000, £24,000 and £40,000. The instinct is that someone is overcharging. Usually nobody is.
A brief describes what you want. A quote prices what will be built. Between those two sits a long list of decisions the brief did not mention, and each agency has silently made its own:
- Is content migration included, or do you paste 60 pages in yourself?
- Who writes the copy?
- Which browsers and devices are tested, and how many?
- Is accessibility being met to a standard, or just "we'll keep it sensible"?
- Is there automated test coverage, or manual clicking before launch?
- Who handles the 301 redirect map for existing URLs?
- Is training included? Documentation?
- What happens in month two when something breaks?
The cheap quote is rarely more efficient. It has usually assumed less. That is not automatically wrong — if you genuinely have someone in-house to write copy and migrate content, paying an agency to do it is waste. It becomes a problem when nobody realises the assumption was made until the invoice for "additional works" arrives.
The single most useful question you can ask a quote is what it excludes. A good agency will have that list ready.
Fixed price versus time and materials
Both models are legitimate. Each fails in a specific way when applied to the wrong kind of work.
Fixed price
You agree a scope and a number. The agency carries the risk of it taking longer than expected.
Right for: marketing sites, defined integrations, POS rollouts, migrations — anything where the end state can be described before starting.
The failure mode: the agency prices in a risk buffer you cannot see, and every conversation becomes a negotiation about whether something was in scope. On a genuinely uncertain project, fixed price makes the agency defensive precisely when you need it flexible.
Time and materials
You pay for time spent, with an estimate and regular reporting.
Right for: exploratory product work, R&D, projects where requirements will legitimately change as you learn.
The failure mode: no ceiling and no urgency. This needs a capped budget per phase and genuinely weekly visibility, or it drifts.
The model to refuse
A fixed price on a brief nobody has interrogated. It looks like certainty and is not. It resolves one of two ways: change requests that push the real cost above what an honest estimate would have been, or corners cut in the places you cannot see — testing, accessibility, security.
Why discovery is charged for
This is the part clients push back on most, so here is the reasoning plainly.
To price a custom system accurately you have to understand the process it replaces. That means sitting with the people doing the work and documenting what actually happens — including the undocumented spreadsheet that everything depends on, and the step everyone forgot to mention because it is "obvious".
That takes one to three weeks. An agency that does it for free has to recover the cost somewhere: either in the price of the work, or by not doing it and guessing.
Discovery with us costs £1,500 to £4,000 and produces a written specification, an architecture outline and a fixed build quote. It is yours to keep. If you take it to another agency and get the build done cheaper, that is a legitimate outcome — and it happens.
It also regularly concludes that you need less software than you thought. That result is worth the fee on its own.
What actually drives the number
For most projects, cost is dominated by five factors — and none of them is "how many pages".
- Integration count. Every external system is unpredictable. A poorly documented legacy API can consume more time than the rest of the project.
- Data migration quality. Migrating clean data is cheap. Migrating a decade of inconsistent records with duplicate customers and three date formats is not.
- User roles and permissions. One user type is simple. Five roles with overlapping permissions and an audit trail is a genuine subsystem.
- Content volume and who produces it. Sixty pages of copy someone has to write is weeks of work, whoever does it.
- Non-functional requirements. Accessibility standards, performance budgets, uptime guarantees, data residency. Each is real work and each is usually absent from the brief.
Notice that "how many screens" is not on that list. Screens are the cheap part.
Reading a quote properly
A quote you can rely on will tell you:
- What is in scope, itemised
- What is out of scope, itemised
- What it assumes about your input — content, decisions, access, availability
- The payment schedule and what triggers each stage
- What happens to the price if scope changes, and how that is agreed
- Who owns the code and what you receive at the end
If a quote is a single number and a paragraph, you are not being given enough information to compare it against anything.
Holding contingency properly
Hold fifteen to twenty per cent of the build cost yourself, outside the agreed price.
This is not an allowance for the agency being wrong. It is for the things you will want once the work is real in front of you — and you will want things, because seeing something working generates better ideas than reading a specification did. A project with no contingency turns every good idea into a difficult conversation.
The honest summary
Software estimation is not precise, and any agency claiming otherwise is either padding heavily or about to disappoint you. What a good estimate offers is not certainty but transparency: here is what we are assuming, here is what we are excluding, here is how we will handle it when reality differs.
If you are trying to budget a project and want a straight answer on the range, we will give you one on a call before any paperwork.
Frequently asked
Because a brief is rarely specific enough to price precisely, so each agency fills the gaps with its own assumptions — about testing, browser support, accessibility, content migration, and who writes the copy. A cheap quote usually assumes less, not that the work is more efficient. Ask each quote what it excludes and the gap normally explains itself.
Fixed price suits work that can be specified up front — a marketing site, a defined integration, a POS rollout. Time and materials suits genuinely exploratory work where the requirements will change as you learn. The dishonest option is a fixed price on an unspecified brief, which always resolves through change requests.
Because a number produced without understanding your process is a guess, and someone pays for a bad guess — usually the client, through change requests. Discovery costs £1,500 to £4,000, produces a written specification and a fixed build quote, and is yours to keep even if you take it elsewhere.
Fifteen to twenty per cent of the build cost, held by you rather than padded into the quote. It is not for the agency getting it wrong — it is for the things you will want once you see the work, which is a normal and healthy part of any project.
- pricing
- budgeting
- procurement
- project management