The Finish Line That Wasn’t

The first thing Henrik Schneider ever said to me - the very first thing, before hello, before introductions - was: “We need to move fast on this.”

I wrote it down. I underline things that worry me, and I underlined that twice.

Henrik was Head of Platform Engineering at Delos Interactive, which made video games or interactive experiences or something in between - the kind of company where everyone’s job title sounds made up but the revenue is very real. He’d come to us through our AE, Brooke Mitchell, who’d done the discovery call and passed it along with what I can only describe as aggressive optimism. Her notes said: “Strong fit. Multiple stakeholders engaged. Ready for technical validation.” Which, translated from sales into English, meant: “They took a second meeting and someone asked about APIs.”

I don’t blame Brooke. She was doing her job. But I want to talk about the handoff, because that’s where the crack opened - the one everything else fell through.

Brooke’s discovery notes were two paragraphs. One about Delos’s tech stack. The other a bullet list of “key pain points” that read like a horoscope: vague enough to apply to anyone, specific enough to feel personal. No org chart. No success criteria. No timeline beyond “Q3.”

I asked Brooke if Henrik had mentioned specific numbers during discovery. Latency thresholds, throughput targets. She was sure he had.

“Did you write them down?”

She checked. She had not.

I should have stopped here. Should have gone back to Henrik and said: before we build anything, tell me what good looks like. But Brooke had already sent the intro email. Henrik had already replied. The momentum was moving and I told myself I’d figure it out as we went, which is a thing I tell myself when I am about to not figure it out as I go.

The first call with Henrik went well enough. They were evaluating three vendors. Their current system was “showing its age.” They needed to move fast.

I asked what success would look like at the end of the proof of value.

He said: “We just need to feel confident it can handle our workload.”

I wrote this down. I did not underline it. I should have underlined it, boxed it, pinned it to my wall - because feel confident was the sound of a man who knew what he was running from and had not yet turned around to see where he was running to. But it sounded reasonable in the moment. Most dangerous things do.

I put together a scoping document. A page and a half. What we’d build, what they’d test, how we’d know if it worked. I included a section called “Success Criteria” with three blank rows and a note: To be completed with Delos team before build begins.

Henrik said it looked “really solid.” The three rows stayed blank. I never heard about the document again.

What I heard instead was scope creep dressed as curiosity. Week two: Kafka ingestion. Week two, day four: Charlotte Beaumont, a principal engineer who appeared on the third call like a side quest I hadn’t signed up for, wanted custom serialisation testing. Week three: Henrik asked me to spin up a second environment for Lorenzo Rossi in Berlin, who dialled in at hours suggesting either dedication or insomnia and asked questions that had been answered two calls ago.

Neither Charlotte nor Lorenzo had been in discovery. Neither had read the scoping doc. I found out about them not from Henrik, not from Brooke, but from Slack notifications at odd hours. Lorenzo had left a comment on my dashboard at 11pm: “Where is the Kafka connector? Henrik said there would be a Kafka connector.”

I messaged Henrik. He said: “Oh yeah, I told Lorenzo you were working on that. Hope that’s okay.”

It was not okay. But I said it was okay, because Delos was my biggest open deal, and there is a version of me that is very bad at saying things are not okay when they are, in fact, not okay.

I was building sandcastles in a rising tide. Every time I finished one thing, a wave came in that washed the edges off the last. And the worst part was that everyone kept saying we were “almost there.”

Almost where?

I started asking in every call. “Henrik, what does ‘there’ look like for you?” And each time he answered a different question than the one I’d asked. “Charlotte thinks the throughput looks promising.” “I think we’re in good shape.” Statements that felt like answers but weren’t. I’d nod. Nodding is what calls are for.

Week four. I’d built three demo environments, written custom connectors for two data sources outside the original scope, and stayed up until midnight hand-tuning a dashboard because Charlotte had mentioned - offhand, at 5:47 pm - that their CTO liked visualisations.

The CTO. Not mentioned once in discovery. Apparently had veto power over the whole thing.

I went back to Brooke. “Did you know about the CTO?”

“Henrik told me he had authority to make this decision.”

“Did he say that, or did you ask him if he was the decision-maker and he said yes?”

Pause. “What’s the difference?”

The difference is everything. People will agree they’re the decision-maker because it feels true, because saying “actually my CTO has veto power” is an admission of organisational complexity nobody wants to make on a first call. Brooke had asked a leading question and got the answer it was leading towards, and now I was four weeks deep building for someone who might not be able to say yes.

That night I sat at my kitchen table with instant coffee - the bad kind, the kind that tastes like someone described coffee to a machine - and stared at my notebook. Arrows going nowhere. A doodle of what was meant to be a maze but looked more like a nervous breakdown in biro. Every page the same: tasks completed, questions unanswered, a growing list of things I’d built next to a conspicuously empty space where the reason should have been.

I’d been so busy proving value that I’d never made anyone say what value meant. And the thing that sat in my stomach like cold coffee was this: I wasn’t even sure Henrik was avoiding the question. I think he genuinely didn’t know. He had a problem and he thought evaluating vendors meant watching them build things. Nobody had asked him to turn the feeling into numbers. Not Brooke. Not me. We’d all just kept moving because movement feels like progress, even when you’re walking in circles.

The next morning I wrote an email to Henrik. Not long. Not aggressive. Just clear.

Henrik - I want to make sure this PoV gets you to a decision. Can we get 30 minutes with you, Charlotte, and whoever else needs to weigh in, and nail down three things: what we’re testing, what good looks like, and when we’ll know?

He replied in eleven minutes. “That’s a great idea. Should have done this earlier.”

I read that line four times. Should have done this earlier. I wanted to reach through the screen - not to strangle him, just to gently take him by the shoulders and say: Henrik, I sent you a document with three blank rows four weeks ago. The earlier was right there.

But what I actually did was send a calendar invite. Because that’s what professionals do. They swallow the scream and send a calendar invite.

The call happened two days later. Charlotte was there. Lorenzo dialled in looking mildly confused, which I think was just his face. And a new person appeared - Astrid Lindqvist, VP of Engineering, actual budget holder, the CTO’s direct report.

She’d never heard of the PoV.

Let me say that again. Astrid Lindqvist, the person whose approval was required for any vendor contract over €50,000, had never heard of the proof of value I’d spent four weeks building. She thought we were “doing some testing.” She did not know there was a deal attached.

She asked, with the calm authority of someone who signs cheques, what problem we were solving.

I let Henrik answer. He did well - better than I expected. Having to say it out loud, to someone with authority and no prior context, forced a clarity that weeks of Slack threads never had. He talked about latency during peak events. The serialisation bottleneck. The operational cost of their current system. And for the first time, I heard him use specific numbers. Numbers that had been living in his head the entire time, waiting for someone important enough to draw them out.

By the end of the call we had three success criteria. Written down. Agreed. Latency under 200 milliseconds at peak. Throughput of 50,000 events per second sustained. Compatibility with their custom serialisation without a forklift rewrite.

Forty minutes to get what I’d spent four weeks trying to reverse-engineer from vibes.

Astrid said, before she dropped off: “I wish someone had looped me in earlier.” She said it without accusation, and it landed on the call like a stone in a pond. Henrik said nothing. I said nothing. Lorenzo looked confused, but that was just his face.

We hit all three benchmarks inside a week. Not because the product had changed. Because I finally knew what the finish line was, which meant I could stop running in circles and start running in a straight line.

Brooke closed the deal six weeks later. She sent me a message: “Delos signed! Thanks for the heavy lifting.”

Heavy lifting. She wasn’t wrong. But most of the weight had been unnecessary. The weight of building without a blueprint. The weight of performing competence for an audience that hadn’t decided what it was watching for.

That evening I found my notebook. The page from week four. The maze doodle. The underlined words that didn’t connect to anything. I wrote underneath, in small letters, the thing I’d been circling for weeks without saying:

The finish line was never missing. We just never stopped long enough to ask where it was.

I underlined it once. Then I closed the notebook, because some lessons don’t need underlining twice. They just need you to finally hold still.