Best way to write a test with unknown thresholds

I have a script that has 314 GET requests being called and with just 1 virtual user the results are as follows.

I’m trying to create a test based on those numbers so that when I ramp up the virtual users it can stay close to those numbers. Which metric would be best to create a threshold and ensure the requests can handle a certain amount of virtual users until a certain point. I currently am not sure how many users it can handle and ultimately trying to figure that out but no certain which metrics to pay attention to.

Take a look here, to understand what the different metrics mean: https://k6.io/docs/using-k6/metrics#http-specific-built-in-metrics

To start with, I’d suggest focusing on some http_req_duration percentile, unless you have something else specific you care about. For example, something like that:

export let options = {
    thresholds: {
        // We want the 99th percentile of all HTTP request durations to be less than 300ms
        "http_req_duration": ["p(99)<300"],
    },
};