These are the same business metric, one is revenue the other is count of orders. Why the 1st section has enough data and the other section has not enough data? how this works? Is it a bug?

Each metric has their own min sample sizes

We checked and the two have the same min sample sizes. We didn't edited this value recently. Still not sure why they don't align on the experiment results. Why we should make it lower the 'count'? They are aligned at 150 metrics. We see this is as a Growthbook bug.

@helpful-application-7107 any thoughts?

I think the confusion is coming that the "revenue" metrics is wrongly based on amount $ value. In statistics globally speaking would make sense to have the same sample size, and that should be a numeric value like "count of items" and not "amount of revenue". As some products are having $2 cost, and some could have $499. Wiring these two into a sample size doesn't make sense.

Yes, calling the field

Sample Size

is a misnomer.

Also on the top two charts you need to add the unit eg $ or count.

We should change it to be Minimum Metric Value or the like.

Do you have something planned for a revenue metric that is based on "sample size" param? I am after this param that you don't have.

What do you mean by this: > Also on the top two charts you need to add the unit eg $ or count. I'm not following the chart you mention.

here you need to display 150$ the $ is missing.

Oh, yes. That's a good suggestion, thank you.

but I am against favoring a $ amount metric

but I am against favoring a $ amount metric

What do you mean by this?

can you hop on a call, it would make sense to explain easier.

Sure, is a slack huddle ok?

k

To summarize for posterity... There are three potential "minimums": 1. Minimum sample size (we don't have this): based on the N users in your experiment 2. Minimum total metric value (what we currently call

Minimum sample size

): based on the total of the numerator in your experiment 3. Minimum total metric rows/counts (e.g. for a revenue metric, then this would essentially be the N of orders, not just the sum of the order values). We should immediately rename 2, and add 1 in the short term. Adding 3 requires changing fundamental models on our side and is a bit more complex and also niche. As for Marton's other queries: 1. The

No data

state occurs when there is a 0 in the numerator of your baseline metric. In this case, it is impossible to show Chance To Win and other metrics, but we can do a better job explaining what is going on. 2. We should also show

$

for the renamed minimum sample size on the metric overview page.