A website project that is late, over budget and still full of rough edges is rarely a mystery. When teams look back at one, the symptoms usually trace to a single cause: nobody decided which of fast, cheap and good would be allowed to bend. This post treats that as a diagnosis. It lists the symptoms, works back to the cause, and ends with a remedy that costs one meeting and a paragraph of text.
The symptom cluster
Start with what the team actually reports. Developers say the scope keeps growing. The owner says the budget was clear from the beginning. The designer says feedback arrives too late to act on. Meanwhile the launch date has moved twice, and a list of small defects is being carried forward under the label “post-launch polish”.
Taken one at a time, each complaint has a local explanation. Taken together they have the signature of an unchosen trade-off. The schedule was promised, the cost was capped, and the quality bar was described as professional, which is a promise of all three. Something has to give, and because no one picked the casualty in advance, it gets picked by accident, usually in the final two weeks.
Gathering the evidence
Before blaming people, check a few records. They are cheap to collect and they tell you which corner has been quietly sacrificed.
- Schedule drift: compare the first milestone dates with the current ones. A steady slide points to speed being treated as negotiable.
- Change requests: count the additions accepted without a matching change in date or budget. Each one is a hidden withdrawal from the cost or time account.
- Defect backlog: a growing list of known issues that are deferred rather than fixed is the clearest sign that quality is the corner actually being spent.
- Test coverage: check whether the final round of checks shrank as the deadline approached.
If three of the four show movement, the diagnosis is firm. The project is running on an unspoken trade, and the unspoken part is the problem.
Differential diagnosis: what else could it be?
Good practice means ruling out alternatives. Poor estimation can produce late delivery without any trade-off involved. A weak brief can cause rework that looks like scope growth. A skills gap can inflate defects. These are real, and a project can suffer from several at once. The distinguishing test is simple: if you could ask everyone to name the corner they would protect and they would give three different answers, then the triangle is part of the cause, whatever else is also wrong.
In that situation, fixing estimates or hiring more people will help less than expected, because the new capacity will be spent on the same unresolved argument.
A short case shows how the cluster plays out. A small company asked for a refreshed site in six weeks, with a fixed fee, and a polished look across every page. Four weeks in, the agency was trimming testing to hold the date, the owner was adding pages to the list, and the designer was reworking layouts twice. Nobody had broken a promise on purpose; each had simply been handed three promises that could not all be kept. Once the owner chose to protect the date and the fee, the pages beyond the core set moved to a second phase, and the remaining work settled within days.
The treatment: choose the corner and write it down
The remedy has three parts, and none of them requires new software.
- Rank the three. The owner states which pair matters most for this launch. If the date is tied to a campaign, fast is protected. If the brand is on the line, good is protected.
- Name the relaxed corner. Put it in writing, for instance: “Quality of secondary pages may be reduced to hit the date.” Specifics matter, because “some polish may slip” invites argument later.
- Attach a trigger. Say what happens when a new request arrives: either the date moves, the budget grows, or something else leaves the list.
This turns scope changes from silent erosion into visible decisions. It does not slow the work. It reallocates the arguing to the start, when it is cheaper.
Notice also that the diagnosis is non-judgemental. It does not require anyone to have made a mistake. The people involved were doing reasonable things inside an incomplete agreement, and the fix is to complete the agreement.
Applying it to common web decisions
Concrete cases make the principle easier to see. A team that wants a site live within a month on a small budget can protect fast and cheap by shipping fewer pages. A guide such as this walk-through on which pages a business site needs at minimum helps decide where that line falls. The pages that are left out are the quality or breadth the team has chosen to give up for now.
A team that wants depth and a quiet launch can protect good and cheap, and let the schedule stretch. That choice benefits from preparation before any code is written, along the lines of a pre-development planning routine for a business website. Preparation is the cheapest way to make slow also produce good.
A team that wants fast and good has to accept a higher bill, because experienced people and tested components cost more. Page-level decisions follow the same logic. A strong service page that turns visitors into enquiries takes real writing and design time, and rushing it with a cheap template changes the quality corner whether or not anyone says so.
Where the reference reading helps
If you want a plain-language companion for the operations side of running a site once it is live, a blog about cloud operations that publishes short readable articles describes itself as covering that area. It is useful background for non-specialists who sit in launch meetings and need the vocabulary, though it is not a substitute for the decisions above.
Prognosis once the choice is on paper
Teams that adopt a written trade-off typically notice three things within a few weeks. Meetings get shorter, because the default answer to a request is already known. Arguments move from blame to options, since everyone can see the agreed corner. And launches stop feeling like ambushes, because the thing that was given up was named long before it mattered.
It would be dishonest to promise that the project then runs without trouble. Estimates will still be wrong sometimes and requirements will still change. The difference is that the team has a shared frame for dealing with them, and a shared frame is worth more than a perfect plan.
A one-page template
A single sheet is enough, and five minutes covers the review.
- Protected corners (two) and the reason for each.
- Relaxed corner (one) and exactly what that means in the work.
- The trigger rule for new requests.
- The person who can change the choice, and how they will announce it.
Keep it where everyone can see it. Revisit it at each milestone, since a decision that suited the design phase may not suit final testing. Revising the choice openly is healthy; drifting away from it silently is the thing to prevent.
Summary of the diagnosis
A late, expensive and untidy web project is most often an untreated disagreement about priorities. The evidence is in the schedule history, the change log and the defect list. The treatment is to rank the three goals, name what gives, and give every new request a defined consequence. Teams who do this early deliver the same work with fewer surprises, and teams who do it late usually do it under pressure and at a higher cost.