Conflicting Requirements Are an Unmade Decision

Conflicting requirements are an unmade business decision. The one-page card I write instead of mediating, filled in on a 40 EUR refund threshold.

A finance lead who will not move off the 40 EUR threshold. A head of customer care who wants partial refunds available on any order, whatever it is worth. Both right by the numbers each of them answers for, both outranking the analyst, and both waiting for that analyst to write down how it will work. A requirements conflict is a business decision that nobody with the authority to make it has been asked to make.

One thing straight before the method. That standoff is the shape of the conflict, not a meeting I am reporting, and it is certainly not a scene from the returns case published on this site. That case has no client, no meeting and no interviews, and nobody inside the retailer was asked anything, so nobody argued in front of me there. What it does have is the thing underneath a conflict like this one: a published 40 EUR threshold whose owner I could not find anywhere public, carried in the register as an open question rather than filled in with a plausible department.

In short

• Conflicting requirements are a decision, so the question worth asking is who gets to make it, not who is right.
• Most of them dissolve on one question: what happens if you do not get this?
• Three signals say the argument is standing in for a missing owner: nobody can name who set the rule, the argument restarts instead of resuming, and both sides hand the choice to you.
• The card that moves it off your desk: what is disputed, the options, the cost of each, your recommendation marked as yours, a named decider, and a date with a consequence attached.

For years I read a standoff like that as a communication problem. Get them in a room, find the common ground, keep the temperature down, leave with something everyone can live with. It sometimes worked, and I could never explain why.

It worked on the days the disagreement was not real. When it is real, the room has nothing to decide with, and neither do I. I have no authority to resolve it, and neither does anyone else who is likely to be sitting there. What I owe instead is a name for the thing being decided, an honest price on each option, and a form the decision can be taken in.

Is this conflict real?

Most of what arrives as a conflict is two people defending solutions whose underlying needs do not actually clash. Nobody has said the needs out loud, so the solutions do the arguing. One question separates the two: what happens if you do not get this?

Finance is not defending 40 EUR. Ask what breaks without it and the answer is that below some order value the manual check costs more than the refund it is checking. The number is a proxy for that sentence. Customer care is not defending "no threshold" either. Ask the same question and you get a customer with a damaged 30 EUR item, told the policy has nothing for them, calling three times.

Written that way, the needs stop excluding each other. If the check below the threshold stops being manual, the cost the threshold protects against goes with it, and so does most of the argument.

There is a step before that, and in the returns case published here it was the whole job: nobody could produce a reason for the number. The threshold appears in the published policy and in two help centre articles, always worded as a hard rule, and nowhere does anything say why it is 40 rather than 25. You cannot price options against a rule whose purpose nobody can state.

Only when the needs still exclude each other do you have a real conflict. Then it is a decision, not a discussion.

Three signals the conflict is fake

Fake does not mean insincere. It means the argument is standing in for something else, and that something is usually a process with no owner.

1. Nobody can tell you who set the rule, or when. The rule is published, the owner is not. Two departments are arguing over a threshold neither of them chose, defending a decision made by someone who has left, or never made deliberately at all. In my worked case it sits in the register as an open question with my own name against it: until it is answered, neither position can be taken seriously.

2. The argument restarts instead of resuming. A conflict with an owner resumes: we decided this in March, the reason was this, what changed? A conflict without one begins at zero every time, often with different people in the chairs. No record of the last round means there was no decision in the last round either.

3. Both sides ask you to decide, and then accept whatever you say. This one arrives disguised as trust. Two directors who genuinely disagree do not both hand the decision to the analyst. What they are handing over is not the choice, it is the consequence. Take it and you own a rule you cannot defend in six months, and the complaint lands on your desk rather than theirs.

When those show up, do not facilitate harder. Put the vacancy into the card as the first thing that has to be decided: who owns this rule.

The card I send when requirements conflict

One page, written before the meeting and sent ahead, so the room argues about the options instead of about what somebody said last time.

Filled in, on the threshold:

That card comes from the returns case published here, which is built from public sources. Both positions are reconstructed from a policy and two help centre articles rather than quoted from interviews, and the costs are inferences marked as inferences. On a live project you would have the quotes and the volumes. The card looks the same.

Four rules for filling it in.

Do not average. "Let us say 25 EUR, it is roughly in the middle" produces a number nobody chose, nobody can explain and nobody defends six months later. Splitting the difference usually fails both of the needs it was assembled from.

Do not leave a cost cell empty. The empty cell is where the decision goes wrong. Fill it, or write the open question with a name and a date against it.

Do not let the date be decorative. Write what happens when nobody answers, and make it something the owner would rather avoid.

Keep the card after the decision, with the answer written into it. Six months later somebody will ask why partial refunds stop where they stop. "Card of 30 April, decided by the owner of the published promise, reasoning in the register" ends that conversation. It does not answer who set the threshold originally, which no public source says and my register still leaves open. Silence starts a new one, on worse terms, with people who were not in the room the first time.

The row that fails most often is "who decides", and it fails earlier than this card does. If your map of the project holds job titles instead of decision rights, there is nothing to put in that cell, which is the argument in the note on the stakeholder map that survives contact with the project.

What would a reviewer say about this card?

"Your recommendation is you deciding, with extra steps"

Partly fair. The recommendation is not neutral, and hiding that would be worse, so it is labelled as mine and sits under the options rather than above them. The test: write the option you did not recommend well enough that its owner would sign your description of it. If Option A reads as a strawman, you wrote a decision and dressed it as a card. The second test is what happens afterwards. Record the decisions that went against your recommendation, in the same file, in the same tone, with no commentary. A register in which the analyst always turned out to be right is a register nobody believes.

"This is bureaucracy for a five minute conversation"

Usually right. Most disagreements are cheap, reversible and settled by two people at a desk, and putting a card in front of those is how an analyst becomes the person meetings get scheduled around. My rule: would reversing this after the build cost more than a sprint, or has it already come back twice? Then it gets a card. Otherwise one line in the log, which is enough to stop the third rerun.

"You do not have those cost numbers"

Often true, and truest early, when the numbers sit inside a system I cannot reach. The answer is not an invented figure and it is not a blank. It is "unknown, two days of queue data would answer it, owner me, by 12 March", written into the cost cell. That does two things a guess cannot: it puts a price on staying undecided, and it gives whoever owns the data a reason to release it. More than once the measurement closed the argument before the meeting happened, because one option turned out to cost almost nothing.

The part that stays

The threshold argument was never about 40 EUR. It is about which department absorbs the cost of a case nobody wants, and that is a question about the business, not about the requirements document.

An analyst who settles that quietly, by picking whichever option is simpler to build, has taken on a responsibility nobody gave them. Writing it down, pricing it and handing it to the person whose job it is looks like less work and is the harder thing.

Take the disagreement your project has been carrying for a month and put it on one page tonight: what is disputed, the options, what each one costs, your recommendation, the named person who decides, the date. If you cannot fill the "who decides" cell, you have not found a conflict. You have found a vacancy.

All field notes · The portfolio guide · More in Requirements That Survive Review