Turns out that FactSet is missing filings (mostly interims) in USA as well. There are about 50 companies in the USA missing some filings and about 100 in Europe. It's troubling because some major companies are missing filings like CRM:USA JNPR:USA, ANET:USA, FBIN:USA
You can see it for yourself by pulling up the snapshot page for the tickers above. You will see a small number of bars in like below
The reason the Snapshot looks this way is that it's a point-in-time view, and when we notice a hole in the filings we stop adding processing statements. If you change the As-Of date to two years ago you will see that problem is gone for CRM:USA since the missing statement is in the future.
We're started feeding FactSet these issues and pressing them to make the necessary fixes so it stops happening. We are giving them issues a bunch at a time so we don't overwhelm them and keep them focused. We made this data QA part of our process: periodically we run a screen to find problems, examine them, and report them to FactSet.
We are thinking of revisiting our logic of doing hard stops if there are holes (which makes it obvious something is amiss). Perhaps we keep the hole and continue loading the statements. This will not fix TTM factors that are NA due to the holes, but it will allow other factors to compute, like ones that use annuals.
Let us know your feelings. Thanks
PS. This is how CRM looks on FactSet. Not very flattering
Frustrating! I'm still happy with the overall data quality, glad to see you are pushing FactSet. You'd think they'd be very happy for free help with their data!
Having a QA process is a prudent choice. It's crucial for P123 to critically analyze the data from its providers to avoid being seen merely as a superficial layer over these data sources. By scrutinizing the data, P123 can truly add value for its users.
Moreover, have you considered the periodic creation of complete snapshot backups? These backups would be reserved exclusively for quality assurance and not for user access. Such a strategy could be particularly beneficial for diagnosing and understanding issues that become irreproducible once the data providers address and rectify the underlying errors—like these 'data holes'—without any notification.
I found that one of polish stocks I hold SNK:POL (small/mid cap) have missing financials for 4Q/2023. This stock published 4Q/2023 report on 22/03/2024.
How I can reliably backtest (European) strategy in this environment ? In backtest I use available data 'as of date', but do not know if there was a substantial lag between the time the report was published to public and availability in Factset.
We are currently only monitoring gaps. This is different: FactSet is taking a long time to process the latest filing. This is not uncommon unfortunately, and other providers do it too (but they will expedite specific requests from their clients).
You can see which are late with the rules below. About 65 companies are late with MktCap>250
Now it is even worse...All the data is lost.
21 days earlier, SNK:POL had data for 3Q/2023 and previous periods. After new 4Q/2023 data arrived, all the previous data is lost.
I ran the screen to find issues, checked that it is in fact a data issue on Factset's workstation, and reported the companies below to them today. There are at least 25 more that are big enough that I will report in the next few days. I think it's best to report a handful at a time. I'm also asking them why they don't run periodic queries to catch these issues.
rules:
Country("POL") and MedianDailyTot(10) > 10000 and mktcap > 10 and price > 1 // in PLN
InterimMonths(0)=3 and FHist("salesq",13*9)!=NA and (LoopSum("Sales(CTR,QTR)=NA",8)>0 OR SalesPTM=NA)
unbelievable 
