The Feasibility Memo That Recommends Not Building It

A feasibility memo that says no, derived from a worked case: four questions, each answered from the record, and the conditions that would turn it into a yes.

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. The case ends with a one-page summary offering three options. The biggest of them, Option C, gets one line of verdict: it "should not be attempted before B has produced a number". The case never writes out the memo behind that line. This piece does, from the record alone, and says where the record runs out.

A feasibility memo is a short document that answers one question, whether this should be built now as proposed, and it is only worth reading if "no" is an answer it is allowed to reach and if it states what would turn that no into a yes.

In short

• The option under test is Option C from the case summary: a full self-service rebuild with automated refunds up to a value threshold. The memo below recommends not building it now.
• Every answer comes from something already published in the case: an open question, a scope table, a list of what the proposed change makes worse. Nothing new is measured or assumed.
• The core of a "no" memo is its reversal conditions: the specific, observable facts that would change the answer, each with who could supply them.
• A no that recommends a smaller step producing the missing fact is a decision. A no with nothing behind it reads as reluctance, and gets overruled.

Why write a memo whose answer is no?

Because the alternative is that the no never gets written, and the proposal survives by default.

Public guidance says so outright. The GOV.UK Service Manual page on the discovery phase puts it plainly: "It's not a failure to stop at the end of the discovery phase if your research shows that's the best thing to do." It also names the condition I lean on below: "If there's a hard constraint which means you are not able to improve on the solution that's currently available, it might be worth stopping at the end of discovery."

HM Treasury's Green Book, the appraisal guidance for UK central government, builds the same habit into option appraisal. It requires the "business as usual" option to be carried into the shortlist every time: "The BAU provides a benchmark against which proposals can be compared. The BAU should be taken through to shortlist appraisal regardless of whether or not it meets the objectives and critical success factors." A memo that compares a big build only with a smaller build has skipped the comparison that most often wins.

Neither document is about retail refunds, and the transfer is mine. What both give you is permission, in writing, to put "no" on the table next to the build.

What exactly is being tested?

The one-page summary in the case lists three options against the same problem: a damaged-delivery case that lacks sufficient evidence at the first call has no named owner and no due date. This is the row under test, verbatim.

Option C is not a bad idea. Its effect column says "Largest", and nothing in the case argues otherwise. The memo does not argue that C is wrong. It argues that C cannot be justified yet, and names what would justify it.

Four questions, answered only from the record

The four questions are my choice for this memo, not a standard, and a longer list would not change the verdict here. The column that matters is the last one.

Read down the answer column and the memo has written itself. One unknown and three noes, and every no has a named way out.

A fifth question, whether the build is affordable, has no row, because the case cannot answer it: "I cannot price internal effort from outside the company, and an invented figure would be the first thing to break in the room." A memo that fakes that row to look complete is worse than one that leaves it out and says why.

The memo, filled in

Written for this piece, not part of the case as reviewed. The two peers who reviewed the case never saw this page, and every sentence in it points back to a section of the published case.

To: Head of Customer Care From: Sebastian Koczyk, business analyst Date: 26 September 2026 Subject: Option C, the full self-service rebuild. Recommendation: do not build it now

Recommendation. Do not fund Option C in this release. Fund Option B, which costs one process change, one form and one queue configuration, and which produces the number this memo lacks.

Why not now. Option C is justified only if cases stall often enough to need the largest fix. How often they stall cannot be answered from outside the company (OQ-06). Building the largest option first would spend a delivery slot on a problem whose size nobody has measured.

What the record says against it today.

1. The condition risk moves after payout in Option B already. Option C runs that same risk at volume, before REQ-017 has been tried once.
2. Automated refunds apply rules that have no named owner (OQ-03), and add a threshold of their own.
3. Photo retention has not been checked by Legal (OQ-01), and that question is due before build starts.

What would change this answer. Any one of these reopens it; all three together make C the recommendation.

1. Two months of stall counts and durations under Option B showing the parked branch is frequent and long enough that B's escalation cannot absorb it. Owner of the number: whoever owns the queue export.
2. A named owner for BR-04 and BR-11, recorded against OQ-03.
3. OQ-01 closed by Legal.

What "not now" costs. Nothing breaks visibly. Cases that B does not fix keep leaving through the customer's persistence, and the store front stays without the self-service rebuild for at least one more release.

What I need from you. Agree that Option C is parked with the three conditions above, rather than rejected, so it returns on evidence and not on the calendar.

Why that memo holds together

• The recommendation is in the subject line and again in the first line. A reader who sees only the preview pane knows the answer.
• "Two months" is my number, not the case's. I chose it and it is visible so it can be argued with, which is what [MINE] means on this site. Nothing about the case says two months is enough.
• The conditions are observable. "When the time is right" is not a condition. "OQ-01 closed by Legal" is.
• Option C is parked, not killed. A parked option carries its conditions with it, so whoever raises it again knows what has to be true first.
• The cost of no is stated in the same honest terms the case uses for doing nothing. A no memo that claims no downside is as suspicious as a yes memo that claims no risk.

The blank memo

Copy the blocks in this order. Delete the instructions in square brackets as you fill each one.

To: [The one person who can fund or stop it.] From: [Your name and your part in the work.] Date: [The day you send it.] Subject: [The option by name, and the recommendation in words. "Do not build it now" fits in a preview pane.]

Recommendation. [One sentence of no, and one sentence naming the smaller step you recommend instead. If there is no smaller step, say what you recommend doing with the slot.]

Why not now. [The single condition the proposal depends on, and why it is not met today. One condition. If you have five, pick the one the others follow from.]

What the record says against it today. [Two to four items. Each points at something that already exists: an open question id, a scope line, a risk you wrote down before this memo. Nothing new is argued here.]

What would change this answer. [Observable facts only, each with who could supply it. State whether any one reopens the question or all are needed.]

What "not now" costs. [Honest. What stays broken, and for whom. If the honest answer is "very little", write that.]

What I need from you. [Parked or rejected, and by when. Parked keeps the conditions attached.]

What separates a no from a weak yes?

"Feasible, subject to further analysis" is a no that did not want to be written down: the project continues, and nothing says who does the analysis or what result would stop it.

Three things make a no stand up in the room.

It includes the option of doing less. The Green Book keeps business as usual on every shortlist as the benchmark. In the case, the comparison that sinks Option C is with Option B, and B stands up because the summary also wrote out Option A, publishing the five working day line and changing nothing else, and said plainly that its effect is "none that anyone will notice".

It turns the missing fact into somebody's job. OQ-06 in the case already carries a fallback: "Instrument first, decide second. No further design work on this branch until there is a number." The memo borrows that rule and applies it to the biggest option on the page. The no is only defensible because the smaller step produces the evidence.

It says what it cannot see. Every row in the four-question table points at a published line in the case, and the one question it cannot answer, the price, is left out with the reason given. A reader can check each claim in minutes. That is worth more than a longer memo nobody can check.

Where this memo is weak

It was not reviewed. The case went through a self-review and two external sessions before its summary was final; this memo went through none of that, and I would expect a second reader to push hardest on condition 1, because two months is a number I picked.

It is built on one option in one case. The four questions travel; the answers do not.

And it inherits the case's limit. Everything here was assembled from public sources, so the memo can say what the record lacks but cannot supply any of it. With access, the first job would be the same one the case names: pull the stall count.

Where to start

Find the largest option on your current shortlist and write its reversal conditions before anyone asks for them. If you cannot name one observable fact that would change your recommendation, the recommendation is a preference.

The memo follows the shape of the one-page summary; the blank is on the decision summary template, and the full case the answers point at is on the case. A business case a CFO can argue with takes the yes side of the same problem apart, one number at a time. The success measure that makes a business case falsifiable asks what result would mean a change did not work, which is the question condition 1 above depends on. How long will the analysis take? measures the same case to the decision rather than the document.

Checks 22 to 25 on the scorecard were written for the one-page summary, and this memo fails two of them as it stands. It gives no date for the decision, because nothing in the case sets one for Option C, and it does not lay the options side by side, because the summary already does. Fix both before you send one of your own. The scorecard is free, yes or no, no email address asked for.

All field notes · The portfolio guide · More in The Document That Gets Read