Fundemental Data II

Report a problem or error
1

Hey -

Raised this last weekend and raising again because I’m starting to think I have to find another source of data to do my weekend work and it’s really frustating having been a customer for 15+ years.

I’m trying to review fundamental data and every stock I look at that had earnings this week going back to Monday is not updated in the system yet.

I’m specifically speaking about high growth stocks like UPST, AFRM, TTD, PLTR, ZNGA etc. These are not insignificant companies; TTD has sales of close to $1B year. I even checked WYNN - which is not something I trade but reported on Tuesday and has a 10B market cap - and this data is also not updated. I believe when I think back into the past it was not uncommon that results from Thursday or Friday don’t make it in by the weekend but I’ve never seen it as bad as I have the last few weeks where I can’t find literally one updated set of fundamental data when earnings have changed on Monday/Tuesday that is updated in the system for reasonably large stocks.

Just giving you a heads up because at least for me in invalidates a lot of the work I do around momentum/earnings that I run on weekends and is extremely frustrating.

2

That does seem odd. TTD & ZNGA for example have data in S&P Compustat. I always felt that FactSet was faster processing new filings. We’ll investigate. Thanks

3

We confirmed that FactSet has not yet processed the press release (there is no sec filing yet). Below see a screenshot of their workstation. I checked TTD & ZNGA and they only have updated the analyst actual. We will report this to them. Thanks for reporting this

ttd.png
ttd.png832×403 98.3 KB

4

Upon further investigation the data does appear in the “As-reported” tab of the workstation

We apparently only get data once it has been cleansed (standardized). We will inquire about getting the data even if it has not been processed yet.

ttd.png
ttd.png838×783 172 KB

5

Thanks. There are a ton of stocks that this is happening with. Thanks again.

6

Thanks. There are a ton of stocks that this is happening with. Thanks again.

7

We asked Factset and unfortunately that’s how it is for now:

Preliminary data reported by companies is only made available to their datafeed clients once it has gone through the standardization process. The delay varies depending on the company.

They do have a planned enhancement to provide ‘as reported’ data in the datafeed , but there is no ETA for now.

If a client submits a ticker for missing data they will likely get to it faster and standardize the data.

As a consolation… while they do provide ‘as reported’ in their workstation it is not used for calculating ratios like PE and Price2Sales.

8

Marco,
Given your comments above is Factset backdating the effectivity of the data once they make it available?

I find that I can run a screen after the Saturday morning data update and get a given result. I can run the same screen Monday morning after the late Sunday/early Monday data update and occasionally see a minor change or two. I can then run that same screen after the Monday, Tuesday, Wednesday evening data updates using the previous Saturday or Sunday effective date and oftentimes get some significant differences in the results.

I would expect to get the same results as the initial Saturday run.

I don’t recall this being an issue with the previous data provider.

9

That suggests that any fundamental data used in simulations is not PIT, as the date it shows up in the simulation is potentially before the date it showed up on FactSet’s system. And the date it shows up in FactSet’s system is the date that matters.

10

sklatte, data flows continually to their systems , it’s up to us to define cutoffs. We do a Monday 3am update that is part of the ‘weekend update’. The reason for the Monday AM update is to have the freshest data for Monday rebalances. The differences you see are due to this 3AM update. There’s no way at the moment to specify an as-of date that exclude the 3AM Monday update.

philjoe, FactSet gives us the effective data of when the standardized data gets in the system, not the date of the press release. You can verify with TTD with the following screener rule: Ticker(“TTD”) and SalesQ. Run it wint the ‘Screen Factors’ report.

Today is shows SalesQ of 301, with the as-of 11/14/2001 you get 280. The company pre-announced on 11/8/21. I believe it got into the system on Monday 11/15 , after our 3AM update. Note that historically we only keep weekly fundamentals so setting the as-of date to Tuesday or Monday resolves to the weekend.

It’s not ideal of course , sucks actually, that FactSet isn’t releasing this “dirty” , unstandardized, as-reported data from the press releases. Some of these press releases are quite complete and accurate ( some are not and barely have any data ). We’re pressuring FactSet to commit to redistributing this data (which they don’t even use in their platform – they just display it) . As I said before right now there’s no ETA from them.

11

Marco,
Maybe I’m not following you. If I run a screen on Saturday 11/13/21 after the morning data update and I run that same screen four days later with an “As Of Date of 11/13/21” shouldn’t I get the same results?

12

For the most part yes but for the new data that FactSet processes between Sat 11AM and Mon 3AM. This data is incorporated as the “weekend update”. Only a small subset of companies are affected.

13

I’m curious: I notice that on the System Status page, the “Estimate revision data” was last updated on 11/06/21, 11 days ago. This is usually updated weekly, right?

14

Yes , estimate revision data is stale. We were told it was resolved a few days ago but it did not on our end. We’ll investigate. Sorry about this.

NOTE: estimate revision data , like Up/Down revision last week, is a weekly datapoint that does not change during the week

15

That’s fine as long as they use the date that its standardized and available in their system instead of the press release date. In fact its good news since there is room for improvement and more alpha.

16

So, a simple question: Is P123’s data PIT now or not? The above would suggest that maybe it is and there would be no better news for me than to find this out.

I appreciate that P123 is working hard on this. Working hard on a difficult task. But I shut down all ports based on any FactSet data not too long ago. P123 has become a method for using InList for me. InList to import data this has been proven out-of-sample. For the simple reason that out-of-sample data is PIT data.

I continue to see differences in my sims and ports out-of-sample. Significant differences in the performance, differences in the tickers selected and differences in the ranks recorded for the tickers between the sim and port in the transactions. Differences in the ranks for transactions on the same day.

I had stopped posting on this because I figured P123 had done all it could. I gave up and went to using InList.

I post now only because it seems possible that the differences in the sims and ports are something trivial like the 3 am update or an error on my part.

Just tell me the data is PIT now with minor differences in sims and ports due to something trivial like the 3 am update if you can, please.

I saved my ports–stopping rebalance–if I made an error that P123 wants to show me. And if I did make an error, my apologies. It is not my purpose to make any point on this forum. I would just prefer to know for sure if I can use P123 for more than importing data using InList.

Put in the simplest terms, my sim continues to perform well. Continues to do well out-of-sample. Overfitting the sim is not a problem here. But the port did not do as well as the sim continues to do, and honestly, I do not think the sim is realistic (image below). But I leave it as a question: Realistic?

If my perception is wrong on this I would be happy to simply go back to using it. Granted, it is a 5-stock model but it is very liquid and very tradable. If it were real, it could be (and should be) a small part of a diversified portfolio. Average of High Low, liquid universe with variable slippage.

And would the 25-stock sim–which also does well–be a good port?

Is there a type of data I can remove to get realistic sims? I am happy to do some cross-validation to get a more robust port, but if it is data problem I will never be 100% sure the port will be as good as the sim no matter how many cross-validations I do. I will be left guessing if the port is good and will bail on any significant drawdown. Even if it is a good port it will not end up working for me because of any doubts I have about the data.

Happy for any information on this. Confirmation, prove me wrong, doesn’t matter as long as I learn the best path forward for this sim/port and ones like it in the future (if I ever run another sim).

Jim

Sim does well out-of-sample..png
Sim does well out-of-sample..png2024×1048 240 KB

17

No, FactSet does not provide point-in-time data. Compustat kept effective dates for all changes that they made. So if a number changed from 135 to 160 on November 3, 2015, we knew that prior to November 3, 2015 it was 135 and after that it was 160. FactSet does not give us effective dates for changes and they do not inform us of those changes. Moreover, they do not give us the dates on which they made their data available to subscribers. So if a company announced on November 3, 2015 but FactSet didn’t process the data until November 12, 2015, we have no way of knowing. We have taken several measures to mitigate the effects of these things, including adding a one-day lag for preliminaries and final statements and keeping track of effective dates as best as we can post-June 2020.

18

Yuval,

Thank you. Thank you very much!!! Very clear and to the point.

I would still actually prefer to be shown that my concerns have no rational basis that I am OCD or just full of it. I am open to the idea that there is little reason for concern. But to just keep my wishes out of it:

So, a fundamental value ratio like EBITDA/EV could have a change around the time of an earning’s announcement. Perhaps a change in EBITDA/EV that is large enough (and enough of a surprise) to affect the price. A sim may have that data early enough to buy a stock and make a profit. But the port could be behind by some number of days and we are not sure how many days. Be behind and not see much of a return from the new information. Making the sim and port different in a way that could matter.

I get that you are doing everything you can do with the data and recently added a lag to the earning estimates data and undoubtably are doing other things we are not aware of.

Also it is also a little hard to be completely sure with what is going on with my sims and ports. And I keep wanting that sim to work as a port. Sorry I keep coming back to this. It basically happens every time I think about putting money into that sim/port.

This helps a lot!!! I remain open to any ideas which basically say: Jim, you are right you do not have a complete understanding of what is going on with your sims and ports. And it isn’t quite like you have been thinking. Very open to I got the above wrong.

Best,

Jim

19

Isn’t this at odds with what Marco said and I previously quoted:

20

We HAVE the effective date of when the data gets in the system (11/15), and that’s what we display. This is true post-June 2020.