Your first run: taking a baseline

What a run is

A run puts each of your prompts to the engine, captures the full response, and then parses it: was the brand mentioned, which sources were cited, which competitors appeared, how was the brand characterised. The value is not the score at the top — it is the corpus of actual answers underneath it.

On the free tier a run covers ChatGPT. That is one engine of several, and the results will not generalise perfectly to Perplexity or Google's AI surfaces, which favour different sources. Know that limitation rather than forgetting it: your conclusions are about ChatGPT until you measure elsewhere.

Write your prediction down first

Before you trigger the run, write down what you expect. Which prompts do you think you will appear in? Which competitor do you expect to be named most? What do you think the engine will get wrong about you?

This takes two minutes and it is the difference between learning something and confirming what you already believed. Without a written prediction, whatever you see will feel roughly like what you expected, because that is how memory works. With one, you get a genuine surprise, and the surprises are where the work is.

Reading the first result honestly

The first number will probably be lower than you hoped. That is normal and is not yet a finding — it is a starting coordinate. A baseline is not a grade. It is the thing you will subtract from later.

Resist two temptations. The first is to immediately re-run because the result looked wrong; that spends a second unit of a three-unit budget on a sample you have no reason to distrust. The second is to start fixing things immediately. You have not diagnosed anything yet — Module 4 is where the answers get read properly, and acting before diagnosis is how teams ship changes that address a problem they do not have.

Where this lives in SolCrys

• Trigger a run from the workspace. On free, this is a manual check rather than a scheduled daily one.
• The runs surface shows the run and its status; results land on the workspace dashboard when it completes.
• Read individual responses, not just the aggregate. The per-prompt view is where the real information is.

Lab: Run 1 of 3 — the baseline

Needs a free SolCrys account (no credit card). Spends run 1 of 3.

1. Write your prediction down somewhere outside the tool — three lines is enough.
2. Trigger the run.
3. When it completes, read every response in full before looking at any aggregate number.
4. Note the three things that most surprised you against your prediction.
5. Do not change anything yet.

If the run surfaces an obviously broken prompt, resist rewriting it. Note it, finish the experiment, fix it in the next cycle.

Start Free — free plan, no credit card.

Check yourself

• Did you write a prediction before looking? What did you get wrong?
• Which engine does this baseline actually describe, and what does that mean for how far you can generalise?
• Why is re-running immediately a bad use of the budget?

Go deeper

This lesson is the spine. These guides are the depth.

Continue

• Previous: Choosing your ten prompts
• Next: Reading presence without over-reading it