mPulse RUM Data Loaded through September 2023, with INP + BFCache
By Nic Jansma on
Table of Contents
Latest mPulse RUM Data Loaded
We've completed an export of mPulse daily RUM data into the BigQuery
akamai-mpulse-rumarchive project up through 2023-09-30.
This brings the RUM Archive's mPulse dataset to a running total of 394 straight days of data available since 2022-09-01. (There is also first-of-month data going back to 2021-10-01).
Interaction to Next Paint (INP) Data
INPAVG(FLOAT64): Weighted average
INPSUMLN(FLOAT64): Sum of the natural logarithms
INPCOUNT(INTEGER): Number of measurements taken
At this time, INP data will not be as common as other timers such as Page Load Time (
PLT*) or even First Input Delay (
FID*), as capturing the INP measurement requires mPulse customers to be on a recent (1.766+) version of boomerang.
Let's review the current rate of INP data vs. something like FID:
SELECT SUM(BEACONS) AS BEACONS,
SUM(INPCOUNT) AS INP_BEACONS,
true) as INP_MEDIAN,
SUM(FIDCOUNT) AS FID_BEACONS,
true) as FID_MEDIAN
WHERE DATE = "2023-09-30"
So as of 2023-09-30, out of 169.6m beacons, there were 5.8m datapoints for INP, vs. ~22.8m for FID. The INP rate should grow over time, as mPulse customers upgrade their boomerang.js versions, and as more visitors use browsers that support EventTiming/INP.
One caveat for INP data in mPulse: boomerang.js only measures INP up to the point of the first beacon, which is usually the page load event. For Single Page App (SPA) sites, INP data is also split by SPA Soft Navigations. Due to this, mPulse INP measurements may be less than CrUX INP measurements.
Back-Forward Cache (BFCache) Navigation Data
BFCache data is still very small in the mPulse dataset since it requires an update to boomerang.js 1.785+.
As of 2023-09-30, we're seeing about 0.05% of beacons coming from BFCache navigations:
-- Relative popularity of each Beacon Type
SUM(BEACONS) AS BEACONCOUNT,
WHERE DATE = '2023-09-30')
) AS BEACONPCT
WHERE DATE = '2023-09-30'
GROUP BY beaconType
ORDER BY BEACONCOUNT DESC
This should not be interpreted that only 0.05% of navigations across the web are BFCache, as the incoming BFCache data is limited to only mPulse customers using the latest cutting-edge version of boomerang.js. I'd expect the BFCache rate to rise as adoption of the new version of boomerang.js increases.
For sites that have upgraded to the latest version of boomerang.js, and that don't have any blockers for BFCache, they are seeing a much higher rate of BFCache vs.
page views. However, the mPulse dataset has data from the Top 100 sites mixed together, so we can't separate the data to only review sites that are on the latest boomerang.js.