Delivery
Software estimates: why a single number is a bid, not an estimate
Early on, an honest software estimate is a wide range with its assumptions written down. Here is where the range hides and what to ask a vendor.
Computing America. Applications, integrations, data analysis and AI/ML engineering.
Send the scope
In short
- An honest early estimate has three parts: a range, the assumptions behind each end of it, and each risk priced on its own.
- Before requirements are understood, an estimate can be off by a factor of four in either direction. Discovery narrows it by learning about the problem, not by estimating harder.
- A feature list leaves out the edges where budgets often go: failure paths, permissions, data migration, integration faults and reporting. None of them appears in a demo.
- A fixed price quoted before the risk is known has to cover that risk somewhere.
- Ask a firm to price one feature twice, once as a demo and once with every failure path named. The gap between the two numbers is the cost of the edges a demo leaves out.
A single number is a bid because the honest estimate is wide. Boehm's cone of uncertainty, as Construx publishes it (opens in a new tab), puts an estimate made at the initial concept, in the best case, at anywhere from a quarter to four times the final cost. Once requirements are complete, the range is two-thirds to one and a half times. Firms quote one number anyway. Both sides treat it as a commitment, and the rest of the project goes to defending or renegotiating it.
The dishonesty isn't in being wrong, since forecasts are wrong. It's in presenting a forecast with a precision it doesn't have.
The range narrows by learning, not by estimating harder¶
Discovery is how it narrows. In one to four weeks, for a fixed fee, we observe the process, model the domain, map the integrations and run a technical spike against the riskiest assumption. The first release then takes four to six weeks.
GAO's cost-estimating guide (opens in a new tab) treats sensitivity and risk analysis as a step in producing a credible estimate. A buyer can ask a software vendor for the same discipline.
multiple of the final cost
- Estimate at the initial concept0.25-4x
- Estimate once requirements are complete0.67-1.5x
A single quote at the initial concept hides a best-case range of 0.25x to 4x the final cost; once requirements are complete the range is 0.67x to 1.5x.
SourceBoehm's cone of uncertainty as published by Construx, best-case bounds.
At the end of discovery you should hold:
- A range, with a plain statement of what would push the number to either end of it.
- The assumptions: which systems have usable APIs, which data is as clean as claimed, who is available for decisions and how fast.
- The risks, each priced in a sentence. If the ERP integration needs a vendor engagement, say, that's four to six weeks and a cost.
Much of the range hides in the edges¶
A feature list describes the states you want the system in. Much of the engineering is in the states you don't, and those don't get written down because they're what you're trying to avoid. Five categories recur, and you can find each in advance by asking.
| Edge | What gets missed | Ask |
|---|---|---|
| Failure paths | "Submit the order" is one line. The work is the submission that times out after the order exists, the duplicate from a double click, the partial save, and the retry that must not charge twice. | What should happen if this half-finishes? It's a business decision, and it's cheaper before the code exists. |
| Permissions | The costly question is who can see this about whom. A regional manager covering a colleague sees two regions for three weeks. A contractor sees one client's records but not the client list. A former employee's name still appears on their work. | Which of these cases does the quote model? Get it wrong and you get an exposure or a spreadsheet workaround. |
| Data migration | Duplicate customers with slightly different names, a status field repurposed years ago, dates typed as text. A strict new model must accept what the old one accepted loosely. | How clean is the data we're moving? Treat it as a discovery input. Finding out late reprices the project. |
| Integration under fault | Quotes cover the happy path the documentation describes. They skip the ERP's maintenance window, work created during the outage, and who finds out when the two systems disagree. | What does the system do if the integration is down for four hours? "It won't be" means the outage was left out. |
| Reporting | Arrives as a late line and behaves like a second system. The shape that's fast to write to is rarely fast to query. | A few well-chosen queries, or a separate read model? Decide while the budget is being set. |
Fixed price isn't safety¶
Fixed price feels like risk transfer, but it reprices the risk. A vendor quoting before the risk is known has to cover it somewhere. Both sides then have a reason to argue about scope, and every change request becomes a negotiation.
We quote a fixed price for the build only after discovery, when the uncertainty is small enough to carry without padding it beyond recognition. Before discovery it would mean gambling or overcharging. We estimate that build slice by slice, so each boundary between slices is a place to stop.
Fix the budget and date, and let the scope vary¶
Ship the most valuable slice first, and make each slice independently useful, so what the last one taught you informs the next. A fixed price for the whole scope often doesn't give you the option to stop when you have enough.
Sources
- 1.Cost Estimating and Assessment Guide: Best Practices for Developing and Managing Program Costs (GAO-20-195G) (opens in a new tab), U.S. Government Accountability Office,
Next step
Send us the thing you need estimated.
Describe the scope in whatever shape it is in today, a spreadsheet or an email. We will come back with a range, the assumptions holding each end of it, the edges sitting behind the feature rows, and what would narrow it.
- Phone
- (214) 723-2510
- Reply
- A person replies, not a sequence: within one business day, from someone who would be on the engagement.
Packages
Start with a package
Four packages, each with its price on the page. Pick one and buy it online.
$199/month
Billed monthly
$182.42/month
Billed yearly
$499/month
Billed monthly
$457.42/month
Billed yearly
- Scammers stopped from sending email as you
- Your website and systems scanned every week
- Your email settings checked every week
- Everything in the 1 to 10 plan
- Covers up to 50 people
$199/month
after a one-time setup fee
- A site planned, designed and built for you
- Hosting and updates every month
$499/month
Billed monthly
$457.42/month
Billed yearly
$999/month
Billed monthly
$915.75/month
Billed yearly
- Your Google profile, fully set up
- A Google post every week
- 2 social posts a week
- Everything in Get found
- 4 social posts a week
- Your top 10 Google searches tracked
from$500/month
after a one-time setup fee
- Your data gathered from the tools you use
- A first analysis that answers your questions
- New findings every month
Related reading
Engineering3 min readLegacy software modernization: why rewrites get canceled, and a safer bet
A big-bang rewrite is a bad bet, not a bad pitch. Replacing the system one slice at a time changes the bet and gives the budget review something it can approve.
Strategy3 min readBuild vs buy software: sort each process step before you write requirements
Sort every process step as commodity, differentiating or contextual before anyone writes a requirements document.
Delivery4 min readStatement of work template for software projects: what a SOW must include
A checklist to run before you sign, ending in a test you can give to someone who has never heard of the project.
Delivery3 min readSoftware requirements: the one page you need before a vendor call
Write five answers on one page and send them to the firms you're considering. The long specification can wait for their questions.
