The Conflict You Avoid Becomes the Escalation You Can't Control

Almost every escalation that lands on a director's desk was, five or six weeks earlier, a ten minute conversation somebody decided not to have. Usually that somebody was the technical leader closest to it. Usually they had a reason that sounded a lot like maturity at the time.

The comforting story about not raising something is that you're giving it room to settle. Sometimes that's true. Mostly what you're giving it is room to organise.

Here's what happens in the weeks you spend not having the conversation. Both people keep working, so both keep accumulating evidence for the version of events they already hold. They each tell a colleague, then another, and every retelling sands off the ambiguity until what's left is clean, coherent and unfair to somebody. Positions that were tentative in week one are public by week three, and a public position is enormously expensive to abandon. Meanwhile the underlying thing, which was probably a scheduling clash or an unowned interface, hasn't moved a millimetre.

Then the cost lands somewhere you didn't plan for. Someone's manager hears the story before you say anything. Someone's supplier mentions it on a call you weren't on. A senior person forms a first impression of a problem you understand better than anyone, and they form it from a source that isn't you.

That's the actual price of avoidance. Not the delay. The loss of framing.

A problem belongs to you for exactly as long as you control when and how it gets raised. The second it reaches a senior person through somebody else's account, you've moved from deciding to responding, and responding is a far weaker place to stand.

A pension payments platform integration at a third-party pensions administrator, two suppliers involved, go-live set for October. A team lead and one supplier's delivery manager disagreed about who owned the masked test data environments. It wasn't a complicated question. It was just awkward, the supplier relationship formally belonged to procurement rather than to the lead, and the whole thing felt political. So the lead let it sit. Five weeks, three polite emails, no resolution.

In week six the supplier's account director wrote to the administrator's CIO about "blockers on the client side". That was the CIO's first exposure to the issue, and in that version, the administrator was the problem.

What followed was three meetings, a joint remediation plan, a request for a weekly status paper, and about a fortnight of the lead's life spent correcting the record rather than fixing anything. When somebody finally sat down with both sides and asked what the actual disagreement was, it took an hour and one shared document. The supplier had assumed test data was in scope for the administrator because of a line in a statement of work written nine months earlier. Nobody had read it in the same way twice.

One hour of work. Five weeks of avoidance and a fortnight of damage control to buy it.

The reason good engineers avoid escalation is that they've seen it done badly, and done badly it looks like telling on someone. Fair enough. That version deserves its reputation.

Done properly it's a different act entirely. You put the decision in front of the person who can make it, with the work already done: the disagreement in one paragraph, what each side wants and why it's reasonable, what you recommend, what you need decided, and by when. Add the consequence of no decision, in plain terms, with a date attached.

Notice what that does. You're not handing anyone a problem. You're handing them a choice you've already narrowed, on a clock you set, framed by the person who understands it best. That's the opposite of losing control.

Two failure modes sit either side of this, and both are common enough to name.

The first is absorbing it. You decide that raising it would look weak, so you carry the disagreement personally for two months, mediating the same conversation over and over, protecting everyone from a decision that was never yours to make. You'll be praised for it, right up until the delivery date moves.

The second is filing it. You send the email, you note it in the risk log, you tell yourself it's someone else's now, and you stop pushing. That isn't escalation, it's paperwork with an alibi attached. Escalating something doesn't transfer your accountability for the outcome. It transfers one decision, and only that decision.

Escalate the decision, not the person. Nobody needs your assessment of a colleague's character in writing, and you don't need it in your history either.

Adopt a threshold and use it this week: if a disagreement has survived two direct attempts and seven days, it isn't going to resolve itself, and waiting is now a decision you're making.

Then take the one you're currently sitting on, the one you thought of in the first paragraph, and write four lines today. The decision that needs making. Each position in a sentence. Your recommendation. The date you need an answer by.

Send it to whoever can actually decide, and copy the person you disagree with. That copy is the part that feels uncomfortable, and it's the part that keeps the whole thing honest.