The following story is an invented composite, told because it shows a pattern that engineering managers recognise at once. A small dispatch team at a regional distributor needed a tool to replace a shared spreadsheet. They did what most teams do: one message to a freelance developer, “need a simple web app to track deliveries.” What followed is a useful case for anyone who plans to hire outside help for a software job, because the turning point was not a technical decision. It was the moment the team wrote down its requirements.
Week one: a request that looked complete
Priya, the operations lead, sent four lines to three freelancers. Each replied quickly and confidently. One proposed a full dashboard with maps. One suggested a lightweight form plus a database. One recommended an off-the-shelf product and offered to configure it. The proposals differed in scope by a factor of several, and none could be compared with another. The team spent a week in meetings arguing about which developer seemed smartest, when the real problem was that the question had been too soft for any answer to be wrong.
The first lesson is analytical, not emotional. A request that leaves out outcome, deadline, budget and acceptance tests has a very large solution space. Responses will land all over it. The spread of the replies is a measurement of the vagueness of the request.
Week two: the note
Priya’s colleague Arjun, who had once managed a vendor migration, proposed a different step. Before replying to anyone, write one page. They gave themselves an hour. The page had six blocks.
- Outcome: drivers and dispatchers see the same live list of today’s deliveries, and a driver can mark one complete from a phone.
- Deadline: in use by the start of the busy season, about eight weeks away, with a soft launch to two drivers two weeks earlier.
- Budget range: a stated range, covering build, a short training session and one month of fixes.
- Constraints: data stays in their existing database; must work on low-end phones with patchy signal; no new logins for customers; the existing spreadsheet export must keep working during transition.
- Done means: five scripted scenarios pass during a hands-on test with the dispatch team, including a driver marking a job complete with no signal and syncing later.
- Context: what the spreadsheet does today, which columns matter, who owns each, and what went wrong during last year’s peak.
They borrowed structure from a detailed article on how engineers lay out requirements for a web build, trimming it to a single page because their audience was a freelancer, not a committee.
Week three: the second round of replies
Priya sent the page, unchanged, to the same three people and to two new ones. The replies looked different. Two began with confirmation of scope in their own words, which exposed a misunderstanding about offline use that the team corrected on the spot. One asked whether the existing export format was fixed, a good question that prompted a new line in the note. One declined, saying the offline requirement was outside their experience, which saved both sides a month.
Compare that to round one. The spread of scope narrowed, and the questions moved from “what do you actually want” to “have you considered this edge case”. The team had also begun to use the note internally. When a dispatcher asked for a map view, Arjun pointed to the outcome line and asked how the map would help a driver mark a job complete. The feature went to a “later” list instead of expanding the project.
Where the reach came from
Two of the five candidates came from the team’s own contacts. Priya found the other two on ProfessNow, a marketplace where people look up professionals by trade and location and send them a booking request. She pasted the same page into each request, which meant the first message already carried everything needed to respond properly. The directory solved the discovery problem only; the page solved the communication problem.
Month two: delivery and the finish line
On delivery day the team ran the five scripted scenarios. Four passed, one failed: a driver who lost signal mid-update saw a duplicate record after syncing. Because the test was written in advance, nobody had to argue about whether it counted as a defect. It was a failed line in a document both sides had accepted. The developer fixed it within the agreed support window, the scenario passed on the second run, and the team signed off.
The project closed close to the budget range and within a week of the original date. Neither outcome is guaranteed in real life, and the story does not claim a perfect result. What it shows is a change in the type of problem: from disagreements about intentions to checkable items.
What the team measured afterward
In the retrospective, Arjun made a short table in his head rather than on a slide, and it is worth repeating because it is checkable. How many clarifying questions came back in round one compared with round two? How many proposals could be laid side by side without translation? How many requests arrived after kickoff that touched the written outcome? He did not need exact numbers to see the direction, only a rough tally from the email thread. Round two had fewer basic questions and more specific ones, and after kickoff only two requests collided with the outcome line, both of which moved to the later list.
The retrospective produced one more change. The team kept the note as a living template for future vendor work, with the six block headings and a few prompts under each, so the next request would start from structure rather than a blank message.
What the case shows
- Vague input produces a wide spread of answers. The spread is diagnostic. If replies disagree wildly, fix the request, not the reader.
- Acceptance tests are the most valuable lines. They convert opinions into pass or fail checks that can be scheduled.
- A stated constraint filters early. One candidate left in week three, which cost nobody anything.
- The note becomes a steering tool. It gave the team a polite way to defer new ideas.
- Preparation is cheaper than rework. An hour of writing replaced what could have been weeks of correction.
Before your own project
If you are about to brief a developer, consider the questions that serious projects should answer first. A checklist-style piece on what to settle before commissioning a bespoke web application pairs well with this exercise. Write your own six blocks, test them by asking a colleague who knows nothing about the project to explain the goal back to you, and revise until the explanation matches your intent.
Teams moving from paper or spreadsheets, as Priya’s did, will also find the broader perspective in a business-focused account of moving off paper onto digital platforms useful. The brief is small, but it sits at the front of everything that follows.

