How Long Will the Analysis Take? An Estimate You Can Defend

Three clocks on one published case: thirteen hours of work, seven to nine hours a plan could name in advance, and twenty-eight days of calendar.

Numbers in this note carry source tags ([POLICY], [THREADS n=10], [MINE]). What the tags mean.

This piece stands on the returns case published on this site: damaged-delivery refunds at a mid-size online retailer, built from a published returns policy [POLICY], help centre articles [HELP], ten public complaint threads I selected myself [THREADS n=10], consumer law [LAW] and a walk through the published customer journey [WALKTHROUGH]. No client, no engagement, no interview, and nobody inside that company was asked anything at any point. What the case does have is a published timeline: every version dated, both review sessions dated, the hours written down beside them. The estimate below is derived backwards from that record, not from a feeling about how long this sort of thing takes.

An estimate for analysis work answers three separate questions with three separate numbers: how many hours the work costs, how many of those hours a plan can name before the work starts, and how many days pass before anyone can decide. Quoting one number for all three is how the date gets missed.

In short

• The case cost about thirteen hours. A step plan for the same case names seven to nine of them plus an evening. The gap is the work that happens after the document already looks finished.
• The same case took twenty-eight calendar days. Six of them sit between a finished draft and the end of the second review session, on dates two other people set. Effort does not shorten that stretch.
• Three of the five requirements in that register did not exist in v1.0. Cutting a review pass to hit a date does not cut polish, it cuts more than half the register.
• The document was reviewed on 9 April and the decision it asks for was due on 30 April. Quoting the date the analysis ends answers a question nobody asked.

What did the work actually cost?

About thirteen hours, spread over four evenings and one Saturday. The largest single block was the process model, and most of that block was one full redraw.

That figure is on the case sheet, so it can be checked against what it produced: a brief and six artifacts, namely a stakeholder map, an as-is and to-be model, a requirements register holding five requirements, a rules register, a review log and a one-page summary. It was noted rather than timed, and the word in the case is "about".

It holds no time for arranging interviews, chasing a data export or rescheduling a workshop, because there was no access to arrange. The thirteen hours are mine; the two reviewers read the document as a favour.

Which hours could a plan have named in advance?

The same case is published with a five-step method, and each step carries a time. Reading those numbers as a forecast, then holding them against what the work cost, is the useful part.

Add the hours actually stated and you get seven to nine, plus an evening not expressed in hours at all. The recorded cost was about thirteen.

The difference is not a rounding error and not sloppiness in either number. It is the work the step list does not name: the self-review against the scorecard, the redraft after each review session, and writing up the review log. All three happen after the document already looks finished, which is exactly why a plan skips them.

So an honest planning number has two lines: the hours you can name, and what the passes after the first draft cost. Here the second line is about four hours on top of nine.

The calendar clock: twenty-eight days

Hours answer what it costs. Days answer when it lands.

Twelve March to 9 April is twenty-eight days. Thirteen hours of work, spread across four weeks.

Look at rows four and five. Six of the twenty-eight days sit between a draft being finished and the second session ending, on dates set by when other people could read. That stretch does not shorten if you work harder. It shortens if you ask earlier or ask more people, and both are decisions made at the start rather than rescues at the end.

This is the number to quote when somebody asks when it will be ready, and the reason to quote it with the waiting on its own line. An estimate that hides the waiting inside the work looks smaller and is wrong in the one direction that gets noticed.

What did the extra passes actually produce?

This is where cutting a review round stops looking like a saving. Of the five requirements in that register, three did not exist in the first draft. REQ-018, the guard rail stopping an automated pre-check from refusing anybody, came out of the self-review against the scorecard, and so did the fallback on REQ-016. REQ-015, escalation ownership, and REQ-017, the condition dispute that moves after the money, both came out of the external sessions, along with a rewritten start event on REQ-014, a rule BR-12 nobody had written down, a page of rationale for BR-11, and one scope exclusion nobody had noticed.

Five objections across two sessions: four accepted, one held with the reason written down beside it.

So the passes a plan does not name, and a deadline squeezes first, produced more than half the register. The first draft was the input to the part that found the missing pieces, not a nearly-finished document waiting for corrections.

The estimate card, filled in from that case

Copy the columns, not the content. Anything on this card that I chose rather than measured says so.

The blank card

One card per piece of work. If a row has no number, write what would have to happen for it to get one.

Three rules for using it. Give the waiting its own line, because it moves the date and a single number hides it. Never quote the hours and the calendar as if one implied the other: thirteen hours landed on day twenty-eight here and could as easily have landed on day nine. And when asked to cut the estimate, cut scope or access assumptions out loud rather than quietly cutting a pass, because a pass is where three of those five requirements came from.

Why is "analysis finished" the wrong date to quote?

Because the work exists to produce a decision, and the decision has its own calendar.

The reviewed version of that case is dated 9 April 2026. The one-page summary it ends with asks for a decision by 30 April 2026 and names the three things it needs answered: whether five working days is a commitment or a target, whether partial refunds ship in release one, and who owns the weekend rota. Twenty-one days sit between the document being finished and the question being settled, and none of them belong to the analyst.

The open questions register makes the same point twice over. Six questions, each with a due date and a fallback, and one of them still carrying an open owner slot. OQ-06 asks how many cases actually stall today, and the case says plainly that if parked cases turn out to be rare and short, this work is not worth a release slot. That question is dated to the first week of access, a week that had not started when the document was finished.

An estimate measured to the document answers a question nobody in the room asked. Measure it to the decision, and say which part of that span is yours.

What do the standards put a number on?

Less than people expect, and never on the thinking.

The Scrum Guide is the document most often reached for when somebody wants a norm here. It puts a length on the Sprint and time boxes on the events, and it gives no figure for preparing a backlog item or for analysis of any kind. Sizing there sits with the people who will do the work, not with a standard. Teams quoting a number for refinement got it from their own history, which is the right place to get it and not a place anybody else can borrow from.

HM Treasury's Orange Book, the free risk management guidance for UK government organisations, asks for something a schedule can use: "clear processes for bringing significant issues to its attention more rapidly when required, with agreed triggers for doing so". That is written about risk, not estimates, and the transfer is mine. It is still the most useful sentence I know here, because the weakness in most estimates is not the number, it is that nothing is agreed in advance about what makes it wrong. Hence the trigger row above.

What this estimate does not cover

One case, one person, public sources, no client waiting. Treat the figures as mine and the shape as transferable: hours, the subset a plan can name, calendar days, waiting shown separately.

A real engagement adds a class of waiting absent here entirely. Interviews get scheduled and moved, access gets requested and granted late, a workshop needs a room full of diaries rather than two. Each of those lands in the calendar column and none of them in the hours column, which is why the two numbers have to travel separately.

And the thirteen hours were noted, not timed. If you want a figure you can defend to somebody paying for it, time your next two jobs. Two measured ones beat any number of remembered ones.

Where to start

Take the last piece of analysis you finished and write down two dates: when you started reading, and when the decision it supported was actually made. The span between them is the estimate your stakeholders experienced, whatever number you gave them at the start.

Three pieces on this site pick up where this one stops. Definition of ready for a requirement is the gate deciding whether a pass is finished, run over the same register. A risk register with owners, dates and a trigger derives its rows from artifacts that already existed, the same move this piece makes with a timeline. The questions a reviewer asks about your work sample is what those two sessions looked like from the other side of the table.

The twenty-five checks behind the artifacts named here are on the scorecard: yes or no, free, no email address asked for. The worked case this timeline comes from is at the case, and the six blank templates are on templates.

Nobody can tell you how long business analysis takes, and a page answering that with one number would be worth less than this one. What can be published is a real timeline with the waiting left in, next to the artifacts it produced, so you can hold your own against it.

All field notes · The portfolio guide · More in The Case File