Raw page data doesn't match HTML results

In performing a few manual tests recently, I’ve noticed that the values in the Raw Page Data CSV do not necessarily match up with the HTML output from WPT. Particularly looking at Speed Index (the last value in the CSV).

First View: 6039 vs 6107
Repeat View: 4104 vs 4157

Also, for Doc Bytes In…
First View: 2,644KB vs 2707218 Bytes (2,707KB)
Repeat View: 701KB vs 71722 Bytes (717KB)

Is this an expected behavior?

Secondly - what is User Time? It doesn’t appear to be a metric in the CSV.

Do you have a link to test with these conflicting results?

I feel like any random test you perform off the WPT site reproduces this. Here’s a simple washingtonpost.com test:

Speed Index in the HTML results is 3513 vs 3545 in the CSV.

Doc Bytes In in the HTML is 5,727KB vs 5,864,020 Bytes in the CSV.

Fully Loaded Bytes In in the HTML is 5,916KB - is that meant to match up with “Bytes In” in the CSV (6,057,843 Bytes)?

While these discrepancies are small, Speed Index is going to be an important metric to us and I i’m just curious why there’s a difference in the results.

That is odd, but it appears to be specific to the CSV export. If you get the JSON version of the test results you get the expected Speed Index value - https://www.webpagetest.org/jsonResult.php?test=170218_BX_GV2

I agree, perhaps its an issue with the CSV export.

Is there an official place to report this?

Will otherwise move to the json results.