How do you make the case for headcount or a platform investment, and what do you do about a stakeholder who keeps going around you?

Ask for the outcome, not for the people

"I need three engineers" invites a negotiation about the number, which you will lose, and which reveals nothing about why. "Our onboarding funnel is losing customers at verification and we cannot fix it while the same team owns the payments migration; here is what fixing it is worth and here are three ways to resource it" invites a decision about a business outcome, and lets the person deciding choose the shape.

Present options rather than a single request. A version that solves the problem fully, a version that solves the worst half, and the do-nothing option with its consequences named. Three options change the question from whether to fund you into which of your framings to accept, and they demonstrate you have thought about the constraint the decider is under.

The components that make an ask credible are the same regardless of size.

Component	What it must contain	Common weakness
The outcome	A business result, in the sponsor's language and metrics	Framed as engineering activity
Cost of the status quo	Quantified from assumptions you show, and attributable	Asserted as "significant tech debt"
The ask	The smallest credible version, with alternatives	Padded, which invites a haircut
What you stop	The work being displaced, named explicitly	Silence, which reads as capacity you were hiding
Evidence of working	A leading indicator visible within a quarter	A benefit that only appears in year two
Kill criterion	What would make you hand the money back	Absent, which makes the case unfalsifiable

The last two rows are what separate a senior case from a request. Volunteering a kill criterion — the measure that would tell you this was the wrong bet, and a date to check it — buys more credibility than any amount of confidence, because it signals you are optimising for the company's capital rather than for your headcount.

Pricing the status quo

Do the arithmetic in the open, from assumptions the audience can challenge, and never invent a figure. If four teams each lose roughly half a day a week to a flaky deployment path, say that, state the assumption plainly, multiply it out, and let people argue with the half day rather than with your conclusion. A number derived visibly from a stated premise survives scrutiny; a confident number of unclear origin dies to the first sceptical question and takes your credibility with it.

Where the cost is a risk rather than a running loss, price it as exposure and likelihood instead of pretending it is a certainty. "A restore has never been tested; if it fails we are looking at a multi-day outage of the ledger" is a legitimate and forceful case that does not require a fabricated probability.

Platform investment has an attribution problem

Platform work is harder to fund because the benefit lands on other teams' metrics. Three moves help. Express the benefit as a tax being removed, measured across the teams paying it, and get one of those teams' leads to say so in their own words — a peer's testimony is worth more than your estimate. Attach the investment to something already funded and already at risk, so it becomes an enabler of a committed outcome rather than a competing priority. And commit to an adoption metric rather than a delivery metric, because a platform nobody uses is worse than no platform, and volunteering that measure shows you know it.

Deliver it in thin slices with visible results in the first quarter, for the same reason any long programme needs early value: the sponsor who approved it may not be there in a year, and the successor will fund what they can see.

Finally, never take the case to the forum cold. Walk the two or three people who can kill it through the argument individually beforehand, ask what they would object to, and incorporate it. A decision meeting should ratify a case that has already been agreed in private, and a manager who treats the forum as where the persuasion happens will keep losing to people who do this.

When someone routes around you

Start with the diagnosis, because the cause determines the response and the instinct to defend process is almost always wrong. The usual causes are unflattering: your intake is slower than their deadline, your last answer was a no with no alternative attached, they had a direct relationship with an engineer before you arrived, or they have learnt that the side channel works. Only occasionally is it a genuine attempt to undermine you.

Have the conversation about their problem rather than about the protocol, because a conversation about protocol makes you the obstacle in the story they tell afterwards.

"You went straight to Anya for the reporting change, and I want to make sure I understand why rather than making it about process. My guess is that coming to me felt like it would take two weeks to get an answer. If that's it, that's on me and I'd like to fix it — I can give you a yes or no inside two days from now on. What I can't do is have work land on Anya outside the plan, because she's on the migration and you'd be getting a quiet delay somewhere else rather than a fast answer here."

Then earn the front door. Publish how requests are made and what response time they get, answer inside it, and make sure a no always comes with a reason, a date when it could be reconsidered, or a smaller alternative. Most bypassing stops when the official route is genuinely faster than the unofficial one.

Handle your engineer's side privately and without blame in the first instance. They usually said yes because saying yes is pleasant and because nobody told them what to do instead. Give them a script — "happy to help, let me get it into the plan with my manager so it does not fall over" — and be clear that the expectation is transparency, not refusal. If it continues after that, it becomes a straightforward expectations conversation.

Escalate only when you have made the front door work and the bypassing continues, and escalate the effect rather than the etiquette. "Three unplanned requests this month moved the migration date by two weeks" is a business fact your director can act on. "They keep going around me" is a complaint about status, and it makes you look small in exactly the room where you need to look reliable.

Fund an outcome and price the status quo honestly, then treat every bypass as a measurement of your own intake before you treat it as a challenge to your authority.