Cwa-documentation: In-App statistics update too late

[whiskey]

First the basics:

Avoid duplicates

• [x] Bug is not mentioned in the FAQ
• [x] Bug is not already reported in another issue

Technical details

• Device name: any
• iOS Version: any
• App Version: any since the statistics

Describe the bug

The in app statistics are great but updated too slowly. How can it be that DLF announces the latest numbers at 6 in the morning and the app still shows at 08:00 the stats from the previous day? Please update earlier, you have the 'direct line' to the RKI.

Steps to reproduce the issue

1. Get up early in the morning (yes, this still happens, despite home office)
2. Listen to the radio at 06:00
3. Compare announced numbers to the in-app statistics

Expected behaviour

Numbers in the app are updated to the latest statistics.

Possible Fix

Only Server sided. No need to adjust the app code.

Additional context

n/a

---

Related: https://github.com/corona-warn-app/cwa-documentation/issues/532
Internal Tracking ID: EXPOSUREAPP-5448

[dsarkar]

Dear @whiskey,

Thank you for your contribution. As far as we know, yesterday February 26, there was actually a delay in showing the updated numbers. Please, stand by, we will come back to you. Best wishes, DS

---

Corona-Warn-App Open Source Team

[Ein-Tim]

First of all, I noticed that too.

@dsarkar & @whiskey
Second, may I suggest to transfer this to either the documentation repo (since it affects iOS & Android) or directly to the server repo?

[whiskey]

[…] directly to the server repo?

Would have made more sense… I'm too biased which repositories I use :)

[dsarkar]

@whiskey @Ein-Tim Moved to documentation repo. The server repo is not related to RKI statistical data.

[dsarkar]

@whiskey @Ein-Tim Thanks. Internal Tracking ID: EXPOSUREAPP-5445

[MikeMcC399]

The server repo is not related to RKI statistical data.

See https://github.com/corona-warn-app/cwa-server/issues/1223 where it was confirmed that the cwa-server just has a reformatting and pass-through function as far as RKI infection statistics is concerned.

This repository (https://github.com/corona-warn-app/cwa-documentation) is the best one for functional problems affecting both the iOS and the Android app.

[dsarkar]

New Internal Tracking ID: EXPOSUREAPP-5448

[GisoSchroederSAP]

All,
regarding the "too late" question:
The consolidation cycle of the current day's in-app statistics on server side starts at 0:10 a.m. Multiple (at least 4) data sources are examined and reviewed.
Not before 05:30 a.m. daily, the "ESRI dashboard" will be updated. Once this task is performed, the data analysts review and validate the different data streams, all data are consolidated, processed, and condensed to the statistics later displayed in the CWA - data quality and consistency is crucial before releasing the data records.

Unfortunately, there is no real "latest time to expect statistics" - at least not of my knowledge.
However, when the internal alerting for any delay or service outage on the server side gets triggered, the data analysts and operational experts will jump in to analyze the root cause and fix it accordingly - if necessary, even with manual intervention. Once finished, the new statistical records will be made available to the app.

So, yes, as multiple data sources are combined and those are not always available/up-to-date at the same time with the same level of data quality, the complexity for the CWA-feeding background process is much higher than scanning one single data pot. Subsequently, the required effort for presentation with respect to time and data quality is higher.

As those in-app statistics shall support users and help them adjust their own behavior accordingly, we consider data accuracy and consistency are key. Still, we are working on continuous improvements and acceleration of the process.

[MikeMcC399]

@GisoSchroederSAP
Thank you for the insight into the process!

Today, Wednesday, March 3, 2021 at 08:00 all RKI statistics values shown by the (Android) app are consistent and up-to-date with other RKI information sources available online (Dashboard and Fallzahlen). 👍

[MikeMcC399]

Today, Thursday, March 18, 2021 16:20 all values in the app have been updated except the 7-Day Incidence, which is still showing the value 86.2: the same as it showed yesterday. The new value of 90 was published by RKI to Fallzahlen.html this morning at 09:40 and the value of 90.4 is also shown in RKI's Fallzahlen_Kum_Tab.xlsx Excel file (in cell BL_7-Tage-Inzidenz'!LG20).

So I'm wondering what happened to the data feed for the 7-Day Incidence to the CWA app which I would expect to show 90.4 right now, instead of 86.2.

[dsarkar]

@MikeMcC399 Indeed, the values were updated late today. Will try to get info about the 90.4 / 86.2 discrepancy.

[GisoSchroederSAP]

I can confirm: 8:35 p.m. The value of the 7-day Incidence is still not yet updated.
So, I'm going to force the data analysts one more time...

[GisoSchroederSAP]

Update: I'm going to ask tomorrow morning, if there is a way to indicate early, when there is an issue that prevents an in-time update of the statistics. Someone must be able to have transparency on the process.

[MikeMcC399]

@GisoSchroederSAP
It's strange that only one value is missing. Thank you for following up!

Even the official PDF reported which is published every evening has the value 90.
https://www.rki.de/DE/Content/InfAZ/N/Neuartiges_Coronavirus/Situationsberichte/Maerz_2021/2021-03-18-de.pdf

[GisoSchroederSAP]

Let me share one more fact:
I already mentioned, those statistics will be consolidated separately based on various different sources, right?
Having this in mind, our implementation works as follows:

• For all tiles, we start in the early morning to collect data from different sources.
• If those data are available, they will be consolidated.
• If all data for _one_ statistics tile are available, the data will be processed and the validity review happens.
• If the latest review is successful, the result(s) will be pushed to the server, where the app can them pick up and update the the value(s) on the tile.

There a lot of "if" conditions, aren't there?
So, if one "if" fails, this respective tile will not updated.
On the other hand, the process may still succeed for other tiles and the user gets at least, what we have so far.
The alternative would be: Wait until all calculations are finished successfully. But this would mean, either all or none of the tiles get updated. We decided, not to go this path,, though.

Does this help?

[Ein-Tim]

As of now, all statistics are up to date.

[MikeMcC399]

@GisoSchroederSAP
Thank you for explaining about the granularity of the publishing process.

Did you find out why the value 7-day Incidence 90.4 was not published during the working day yesterday, March 18, 2021?

@Ein-Tim

As of now, all statistics are up to date.

Agreed. Today, Friday, March 19, 2021, went much more smoothly.

Today's 7-day Incidence is 95.6 and this is showing in the app. The site COVID-19: Fallzahlen in Deutschland und weltweit was updated today at 07:50 and Robert Koch-Institut: COVID-19 Dashboard showed today's data even earlier (it is currently showing "Letzte Aktualisierung: 19.03.2021, 09:06 Uhr, although that was not the first update of today).

[MikeMcC399]

@GisoSchroederSAP

Did you learn any more about the reasons for past problems?

The data in the app seems to be getting updated reliably the last days. Are you monitoring this automatically?

[GisoSchroederSAP]

I'm validating the content/values from time to time, still confident about correctness of the calculation.
Update: Meanwhile, there was a new technical interface established that allow faster and more often updates on the analytics engine. This would somehow confirm your comment about reliability - and it seems to have a impact to the timing (early update each day) - keep the fingers crossed.
Last update: I documented all statistic tiles and translated them (DE/EN) - these texts are now under review. Once released, we are going to make them available on the FAQ page.