When looking at gender content gap on the public folder that was created on July 29th, I don't see drop as you did.

Running this snippet based on Miriam's notebook

import pandas as pd
gendata=pd.read_csv("https://analytics.wikimedia.org/published/datasets/knowledge_gaps/content_gaps/knowledge_gap_index_metrics_csv/content_gap=gender/part-00000-4001ddac-9eed-47bb-b0ae-251784f815ca.c000.csv")
wikis = [ w for w in list(gendata["wiki_db"].unique()) if w.endswith("wiki")]
gendata=gendata[gendata.wiki_db.isin(wikis)]
gendata['gender3category']=[i if i in ['male','female'] else 'non-binary' for i in gendata['category']]
gen3snapshot=gendata.groupby(['gender3category'])['article_created_value'].sum().reset_index()
gen3snapshot

gender3category article_created_value
0 female 2386030
1 male 9898221
2 non-binary 13474

This is consistent with April data you pasted. Were you using the same file as in the snippet above?

I tried a few things -

1. created a new notebook today for calculating the values using the gender content gap csv. and I'm now getting values similar to yours, though slightly different than yours.

gender3category article_created_value
0 female 2381837
1 male 9873151
2 non-binary 13463

2. I re-ran the previous notebook where I got erroneous numbers and I'm still getting the same wrong numbers.

3. The last thing I tried was using gendata=pd.read_csv("https://analytics.wikimedia.org/published/datasets/knowledge_gaps/content_gaps/knowledge_gap_index_metrics_csv/content_gap=gender/part-00000-4001ddac-9eed-47bb-b0ae-251784f815ca.c000.csv") instead of downloading the csv and uploading it to jupyter nb gendata=pd.read_csv(indir+'gender-june.csv') and this also yielded the correct numbers!

In conclusion, I think it is best to use the csv link in the read_csv command to avoid errors that could possibly be introduced due to manual handling of the csv file. One other major take-away is to find a way to QA the numbers we get (other than just comparing and seeing if it is in line with the previous months).

Lastly, I'm still curious to know why our numbers are different even though we are both using the same csv file.. any idea?