The Simulator Told Me What No Real Call Would

I scored 19 out of 100 on a sales call.

Not a real one. A simulated buyer — a synthetic engineering leader with a verification problem, built from a client's actual market and wired to push back the way that kind of person pushes back. Four minutes and twenty-one seconds. Nineteen out of a hundred, with a written breakdown of why.

I have been selling for years. I have never once been given a number.

What the number was actually measuring

The feedback was not about tone, which is where I would have looked first. Tone came out fine. I was respectful, I did not oversell, I made no claim I could not support, and I did not talk over the incumbent vendor he already had. On every axis I would have graded myself on, I passed.

The nineteen came from sequence.

I pitched before I understood anything. He named a problem and I answered it with our capability instead of asking what the problem costs him. He declined, and I responded by restating the value rather than acknowledging the no. And at the end, with nothing agreed, I offered to send material and left the next step as a vague gesture at a colleague possibly getting in touch.

Three errors, one family: I was moving toward my outcome instead of toward his situation. The restraint I was proud of was real, and it was decoration on top of a conversation that had the wrong shape.

Why a simulation was the only thing that could tell me this

A real prospect who experiences that call does not give me the diagnosis. He gives me "send me something" and then silence, and the silence is unattributable. It could be budget. It could be timing. It could be that I pitched ninety seconds too early. I will never know, so I will carry on doing the thing that caused it, and I will interpret the occasional yes as evidence that the approach works.

Outcomes are terrible feedback for a process with a long lag and a low hit rate. A deal that closes six weeks later cannot tell me which of forty moves mattered. A score attached to a four-minute transcript can.

The second run bears this out. Same synthetic buyer, same market, two days later: 53. Thirty-four points, and I can name exactly what bought them — I asked about his friction before I said what we do, and I stopped arguing with the refusal. Still not good. But I now have a gradient to climb, and a gradient is the thing selling has never given me.

The specific habit, written down so I stop relitigating it

The order, which I apparently do not have as a reflex:

1. Ask what the friction actually is, in his words, before naming anything we sell. 2. Quantify it — what it costs in time, rework, or delay. A problem without a number attached is not yet a problem he will spend money on. 3. Widen before narrowing. Who else in the account has this. One team's irritation is not a purchase. 4. Leave with a named person and a date. Not "I'll send something over." Not "maybe a colleague will reach out." A specific next step with a specific human on a specific day.

Point four is the one I keep failing on, and it is the cheapest to fix. Everything before it is judgment under pressure. The last one is a sentence I can simply decide to say.

Where this goes wrong

The simulator optimises for the scorecard, not for the buyer. Whoever built the rubric encoded a view of good selling, and I can learn to satisfy the rubric while getting worse at reading an actual room. Scores rising is evidence about scores. The only real test is whether a live conversation goes further than it used to, and I do not have that evidence yet — two synthetic calls and a 34-point delta is not a result, it is a hypothesis about myself.

And the fictional call generates fictional artefacts. The simulated buyer produced a simulated follow-up commitment, which has every surface property of a real task. I caught one of those about to migrate into a live outreach list. A practice call that leaks into the pipeline is worse than no practice call, because now the pipeline contains a person who does not exist.

The honest counter-argument: nineteen out of a hundred on a four-minute call with a machine may just mean I perform badly when I know it does not count. Low stakes make me careless in a way the real thing does not. I think that is partly true and mostly a comfortable story, and the way to find out is to score a real recorded call against the same rubric rather than to argue about it here.

The thing I actually want to keep: I have been practising sales for years without ever practising it. Doing something repeatedly in front of consequential, unexplained outcomes is not practice. Practice needs a rep, a score, and another rep — and I had to build a fake buyer to get one.