Comparing two very different software quotes

A founder holding a $40,000 quote and a $180,000 quote for what both vendors describe as the same product has no way to tell whether one is a bargain, a misunderstanding, or an opening position to be renegotiated later. The gap is almost never margin. The US Government Accountability Office, whose cost guidance exists because governments keep buying software that costs more than they were told, names the mechanism plainly: "a contractor proposal is a document that represents the contractor's best estimate of cost, and proposals tend to be influenced by the amount the customer can spend" (US Government Accountability Office, Cost Estimating and Assessment Guide, GAO-20-195G, March 2020). Both of your quotes are partly a measurement of the work and partly a guess about you. Six things separate them, and every one of the six is askable before you sign anything.

One: they are not quoting the same thing, and this is most of the gap. Two vendors reading one brief produce two different products, because a brief describes an outcome and a quote describes a build. One read the line about user accounts as email and password; the other read it as single sign-on, role-based permissions and an audit trail, because they have been through an enterprise security review and you have not yet mentioned that your first customer is a hospital. Neither is dishonest. Until both are pricing the same list, the two numbers are not comparable at all, and normalising the scope is the buyer's job because only the buyer holds both documents. The fast version: take the more detailed proposal, reduce it to one line per deliverable, and send that identical list back to both firms.

Two: the seniority mix, which is the largest lever nobody names. Software is people multiplied by months, and people are not interchangeable at any price. Inside a single occupation, in a single country, the spread is wide enough to account for a doubled quote on its own. The US Bureau of Labor Statistics reports that in May 2025 the lowest-paid 10 percent of software developers earned less than $82,460 and the highest-paid 10 percent more than $214,670 — that is the software-developers row of the Occupational Outlook Handbook, not the combined occupation that includes quality assurance analysts (US Bureau of Labor Statistics, Software Developers, Quality Assurance Analysts, and Testers, Occupational Outlook Handbook, May 2025 wage data). The same source shows the industry moving the median as well: $164,550 in software publishers against $132,050 in computer systems design and related services. So the question to put to both vendors is not about price at all: how many people, at what level, for how many months, and are the people who scoped this the people who will build it. What an MVP build actually costs turns that answer into a labour floor you derive yourself from public wage data rather than from anybody's assertion.

Three: the exclusions, which is where a cheap quote keeps its discipline. GAO's definition of a comprehensive estimate is the right thing to hold both proposals against. Comprehensive estimates "completely define the program and reflect the current schedule and technical baseline," are "structured with sufficient detail to ensure that cost elements are neither omitted nor double-counted," and where judgement was required, "assumptions and exclusions on which the estimate is based are reasonable, clearly identified, explained, and documented." Read that last clause as a checklist for your own inbox. The items that routinely sit outside a low quote are not exotic: environments and hosting, migrating the data you already have, third-party service fees that scale with usage, design beyond the screens listed, accessibility, an automated test suite, documentation and the handover itself, and the first year of keeping the thing alive. None of them disappear because a proposal omitted them. Two quotes a factor of four apart can both be honest if one is pricing a build and the other is pricing a build plus everything around it — and the only way to see that is to demand the exclusion list in writing from both.

Four: whether it is a number or a range with a method behind it. This is the most diagnostic question on the list and almost nobody asks it. GAO is explicit that a single figure is supposed to sit inside a distribution: "High-quality cost estimates usually fall within a range of possible costs, the point estimate being between the best and worst case extremes." It is equally explicit about what the range is for — "having a range of costs around a point estimate is more useful to decision-makers because it conveys the level of confidence in achieving the most likely cost and also informs them on cost, schedule, and technical risks" — and about the alternative: "an estimate without risk and uncertainty analysis is unrealistic because it does not assess the variability in the cost estimate from such effects as schedules slipping, missions changing, and proposed solutions not meeting users' needs." You will not get a Monte Carlo simulation from a twelve-person studio and you should not ask for one. Ask the three questions underneath it instead: what is the cheapest this finishes for, what is the most expensive, and which single assumption would have to break for the second number to happen.

Five: who carries the risk of being wrong. Every estimate is wrong; the commercial question is whose problem that is. A fixed price transfers the risk to the vendor, who prices the transfer — which is one entirely legitimate reason a number is higher. Time and materials leaves the risk with you behind a lower headline. Fixed scope with a defined end is the third shape, and it is the one that makes the trade visible instead of hiding it. GAO's fourth characteristic of a reliable estimate covers this directly: credible estimates "discuss and document any limitations of the analysis, including uncertainty or bias surrounding source data and assumptions," vary the major assumptions "to determine how sensitive it is to changes," and "include a risk and uncertainty analysis that determines the level of confidence associated with the estimate." A proposal offering a confident total with no stated assumptions has not transferred the risk to anybody. It has postponed the conversation.

Six: what a change costs after signature, because that is the price you will actually pay. Two quotes can converge or diverge entirely once changes are priced, and the mechanism is usually one paragraph nobody reads. Ask for it explicitly: who may request a change, who prices it, on what basis, how long the repricing takes, and what happens to the delivery date. Then ask for a worked example from a finished project — the change, the number, and what came out of the current scope to pay for it. A vendor who has never done this will describe an intention. A vendor who has will describe a specific argument they once had with a customer. What separates the top software development companies treats that behaviour as a signal you can observe in a sales call; here it is a line item you can price.

There is a seventh reading, and it points at your side of the table rather than the vendor's. GAO warns its own estimators that bias "can originate from different sources, such as over-optimism, group think, dominating personalities, inexperience, or pressure from management," and defines confidence bias as the case where "the estimator is overly optimistic about the success of the program." A founder comparing quotes is exposed to exactly that, in a predictable direction: the low number is also the number that makes the plan work, so it receives a benefit of the doubt it has not earned. Notice when you have started arguing for a quote rather than examining it. That is the moment the comparison stopped.

The case where the cheap quote is a bid to be renegotiated later is real, and GAO describes the mechanism in the setting where it has been studied hardest: "During source selection in a competitive environment, for instance, lower proposed costs may increase the chances of receiving a contract award. In this situation, cost estimators should analyze the cost data for realism." Cost realism analysis is the professional name for the thing a buyer does by instinct and then talks themselves out of. Your version of it is arithmetic. Take the headline number, divide by the months, divide by the people named in the proposal, and ask whether the implied monthly cost per person is a figure any firm could employ those people for. If it is not, the proposal is describing a different team than the one in the meeting, a different labour market, or a different scope than the one you think you are buying — and the third is the one that arrives later as a change order.

What none of this tells you is which firm is better, and that is worth saying plainly. Price is evidence about scope read, seniority and appetite for risk. It is not evidence about judgement, and judgement is the thing a first-time buyer is actually short of. Two proposals that survive all six tests and still differ by a factor of two are usually describing two genuinely different products, which makes the decision a product decision wearing a procurement costume: do you need the version with the audit trail in release one or not. A 150-branch student information system — per-branch revenue and royalty rules, payroll, accounts, an integrated LMS and role-based portals for four different kinds of user — is a different animal from one workflow for one user type, and a quote that prices them alike is not a quote. How to scope an MVP so it ships is the decision rule for settling that before the quotes rather than after them.

And sometimes the answer is neither, yet. If the two numbers are four times apart after you have normalised the scope, the most likely explanation is that the scope is not finished, and no amount of comparison repairs an unfinished scope. The cheap correction is a small, paid, bounded piece of scoping work with a written specification at the end of it — something both vendors can quote against — priced so that being wrong about it is cheap. That is the whole idea behind fixed-scope sprints: a named capability with a defined end rather than an open discovery meter, which is also the thing you should refuse to sign in either proposal.

To be explicit, because an article about quotes invites the inference: none of the figures above are ours. The $40,000 and $180,000 describe the shape of a founder's inbox, not a price list, and our pricing page publishes fixed deliverables and no rate card on purpose — one rate across every project shape means the simple work is overpriced or the hard work is underpriced. Every number in this article comes from the Government Accountability Office or the Bureau of Labor Statistics, and you can open both sources yourself. The criteria a top saas development company should be held to are set out on the page this article supports, including the one we publicly fail, and they apply to whoever sent you the expensive quote exactly as they apply to us.