USCIS processing - The Ugly, The Bad and The Good

[Zipline 23]

I decided to take my own statistical jab at the USCIS data, as others have. I think it was the worth the effort and I found some interesting things to highlight.

Since the K-1 process is already depressing enough, I will start with the most negative findings and work my way up to the positives (The Ugly, The Bad and The Good) and try to end on a positive note.

I decided to stick w/ raw numbers and not graphs. The one thing I find very bizarre is the volatility of the forms processed per month at both the CSC and VSC and I wanted to put some figures together to show just how bizarre it truly is.

Here are the 3 charts (I-129F CSC, I-129F VSC, and I-8219 DACA CSC), and below are my explanation of the the stats and my findings ...

csc.tiff

vsc.tiff

daca.tiff

Explanation of derived stats:

==========================

Pro-rated March 2013 for DACA ... Since only the first 14 days of March 2013 were available for DACA, I decided to pro-rate the data for the month of March rather than throw away the data. Reallize though that this March 2013 DACA

data is an approximation, not real numbers.

Column G - Backlog (number of forms). This one is simple. It is the previous month's Pending work carried forward to the first of the month. This is presented for the following month so we can show how much backlog (total work) there was to be worked on at the start of the month.

Column H - Current Month, Projected Completion (in months). This is the most telling figure, in my opinion. This column essentially says ... "Hypothetically, if no new forms at all were received from this month onward and the service center only worked what it had in it's backlog at the rate it worked this month how long would it take to finish the work?"

Column I - Average Rate, Projected Completion (in months). Similar to Column H, this column asks ... "Hypothetically, if no new forms at all were received from this month onward and the service center only worked what it had in it's backlog at the monthly rate it worked on average, how long would it take to finish the work?" For I-129F, the average is over the last two years Feb 2011-Jan 2013. For DACA it's Sept 2012-Mar 2013 (pro-rated).

Column J - Maximum Rate, Projected Completion (in months). Similar to Column H, this column asks ... "Hypothetically, if no new forms at all were received from this month onward and the service center only worked what it had in it's backlog at the highest monthly speed demonstrated, how long would it take to finish the work?" Again, for I-129F, the average is over the last two years Feb 2011-Jan 2013. For DACA it's Sept 2012-Mar 2013 (March prorated).

Column K - Versus Goal (in months). Shows how far ahead/behind the service center is of reaching it's National Service Goal, working at the speed it's working this month. Positive numbers indicate they will likely complete work ahead of time by X months on average, while negative numbers show they will likely complete work behind schedule by X months on average.

Column L - Ahead/Behind (number of forms). This stat is helpful because it shows whether the processing center gained (positive) or lost (negative) ground when looking at the forms finished during the month as compared to the forms received that month.

================

The Ugly:

================

DACA - This is wonderful news for DACA petitioners and very disconcerting news to us at CSC or VSC. According to the USCIS website, I-821D DACA has a 6 month National Service Goal. Yet, CSC seems to process its DACA forms at a 3 month rate (H25-H27 DACA, green) and a 2-month rate for March prorated (H28 DACA, blue)! Considering the enormous amount of DACA work, why are they finishing 2-3 times ahead of schedule?! What are they trying to prove by working so fast, when no one ever has waited anywhere close to 6 months for a DACA? This is a huge question to ask.

VSC - You will notice the backlog numbers are generally higher at the VSC and reached a maximum of 18,898 (G22 VSC, dark red). For folks who have sufferred at the VSC, we can see that, for whatever reason, historically VSC has carried a larger backlog than the CSC and that the VSC folks have paid the price with longer waits, up until Jan 2013 where they have probably turned a corner based partially on these Jan 2013 figures and what we've seen on VJ.

CSC - These numbers are really troubling. In this data (H25 and H26 CSC, red) we can see that the CSC was shockingly slow working K-1's in Dec 2012 and Jan 2013. If the CSC worked at its December 2012 pace indefinitely to finish it's K-1 backlog, it would take 20.11 months to finish the work. If the CSC worked at its January 2013 pace indefinitely to finish it's K-1 backlog, it would take 19.04 months to finish the work. Notice these figures are worse than any anemic single-month pace at the VSC (VSC's worst was 15.5 in July 2012). This is a worst case scenario, but it's conceivable many people would wait 15-20 months now from NOA1 to NOA2, four times the service goal, if CSC puts forth its worst effort indefinitely.

================

The Bad:

================

VSC - The VSC got way behind in May 2012-July 2012, falling more than 900 forms behind each month (L18-L20 VSC, light red)! It is really good thing the VSC turned things around in the last 4-6 months or they could have reached a 20k form backlog - easy.

CSC - Notice that in May 2011, CSC was working on a pace to complete it's backlog in 2.78 months (H6 CSC, salmon). And about the time that VSC was very bogged down, CSC was working on a completion pace of 3.34 and 2.99 in July 2012 and Aug 2012, respectively (H20-H21 CSC, purple). So why is this bad news? Because it warps expectations. We have probably all heard from friends "oh yeah, so-and-so went to through the CSC and it only took 3 months!" prior to filing a K-1 at CSC. This shows that, intentional or not, the USCIS are the ones manufacturing the drama with such wild variations in completion figures per month. The numbers show that if USCIS could just stabilitize and produce at a steady pace, they could easily do the work in 5 months instead of subjecting people to a 3 to 11 month crapshoot. Wouldn't you rather just wait the 5 months rather than risk waiting 9-11 months?!

The Good:

================

DACA - The USCIS did indeed get bombed with DACA forms from Sept 2012 - Nov 2012 (almost 100k per month), but receipts now seem to have stabilized at about 30k per month. The DACA work has lessened and should in theory be quite manageable now by the USCIS.

VSC - The good is certainly the 4,439 completion number from Jan 2013 (C26 VSC, purple). In that month they really worked ahead of the National Goal for the first time and cut out a huge chunk of backlog. We'll see what the Feb 2013 and March 2013 numbers say, but perhaps the VSC backlog is a thing of past and it's finally down to a 5-month waiting time, if not a little sooner.

CSC - The good news is that, as of Jan 2013, our backlog is really not that bad - 10,929 (G26 CSC, grey). If they turned around and started processing at an average rate they could be caught up in a little over 6 months (I26 CSC, blue). If they really attacked the problem at their previous monthly maximum, they could make sure every K-1 waiting any amount of time from today backward is served within 3.1 months (J26 CSC, brown)! This proves they don't have to move Heaven and Earth to serve us and solve the problem at CSC, they just simply need to increase output to what they've demonstrated before.

I have to be a little fair to the USCIS though and realize there are many kinds of USCIS forms out there and resources are constrained and there's bureacracy, inefficiency, etc. So a modest goal, would be to have CSC and VSC would be kicking out their K-1 Visas at a 5 month rate, following the National Service Goal and column K would be as close to zero as possible). After all, most of us probably saw the National Service Goal of 5 months and the other figures close to it when we filed at CSC. Longer than I would prefer to wait? Certainly, but it's what I got when I signed up.

Oscillating between 3,525 per month and 484 per month at CSC is just madness. If they want to keep people from going ape about the CSC for a few months and then things being super hunky dory in a few months, all they need to do is work at their maximum pace for a single month to weed out those waiting too long and then drop back to their average pace to serve at approximately 4-5 months from here on out.

Edited April 5, 2013 by Zipline

[I & B 273]

Great work. I've been working on something similar with the projections but I have largely been stumped on how to move forward. The problem is that, if you really get into the dashboard data, it doesn't add up.

Specifically, you would think that the number of apps pending in one month would be a simple function of how many apps were pending in the previous month plus the previous month's receipts and minus its completions. But that's not the case at all. There is a huge, inexplicable divergence in what that number of pending would be and the number actually reported. Over the two years of dashboard data I have, i calculate there is an average discrepancy of 1700 pending apps per month. I cannot explain it, and, since I cannot explain it, I cannot figure out the actual relationship between pending and completions from month to month and project a consistent estimate based on it.

I'd love to hear your thoughts on how you dealt with this. It drove me crazy all last weekend and I'm gearing up to have another go at the problem this weekend.

[Zipline 23]

Great work. I've been working on something similar with the projections but I have largely been stumped on how to move forward. The problem is that, if you really get into the dashboard data, it doesn't add up.

Specifically, you would think that the number of apps pending in one month would be a simple function of how many apps were pending in the previous month plus the previous month's receipts and minus its completions. But that's not the case at all. There is a huge, inexplicable divergence in what that number of pending would be and the number actually reported. Over the two years of dashboard data I have, i calculate there is an average discrepancy of 1700 pending apps per month. I cannot explain it, and, since I cannot explain it, I cannot figure out the actual relationship between pending and completions from month to month and project a consistent estimate based on it.

I'd love to hear your thoughts on how you dealt with this. It drove me crazy all last weekend and I'm gearing up to have another go at the problem this weekend.

I noticed the disparity too. I'm not sure where it comes from. I think they probably have a goofy timing on when an app goes from Received to Pending. And maybe it also has something to do with the Awaiting Customer Action figure. You've also got me wondering if the proper Backlog number is really Last Month's Pending + Last Month's Receipts - Last Month's Completions. I may re-run the numbers with this and see what the effect is.

[I & B 273]

I noticed the disparity too. I'm not sure where it comes from. I think they probably have a goofy timing on when an app goes from Received to Pending. And maybe it also has something to do with the Awaiting Customer Action figure. You've also got me wondering if the proper Backlog number is really Last Month's Pending + Last Month's Receipts - Last Month's Completions. I may re-run the numbers with this and see what the effect is.

Yah, the disparity is mind boggling. And, as far as I can tell, it cannot be related to the awaiting figure from month to month, either, as the fluctuations in that number seems to have no meaningful correlation with the direction or degree of the discrepancy.

So as to not send you on a wild goose chase, let me be even more specific about why I found that the discrepancy was posing a very serious problem to projecting delays. In order to project a completion date, I had one basic assumption to go on--that the applications would, on average, be processed serially from the date they were submitted. This means that a given completion date would, on average, be the day upon which all applications that were already pending at the time of an app's submission were completed. So the key variables for determining the completion date of a given app should be the following:

Total Pending At the Time Of Submission (= P + R - C)

1. P: The number of apps pending at the end of the month prior to the submission month.

2. R: The number of apps received in the submission month up until the date the app was submitted (monthly receipts prorated by the day of the month of the submission).

3. C: The number of apps completed in the submission month up until the date the app was submitted (monthly receipts prorated by the day of the month of the submission).

Completions Per Day (This is a variable iterative function over time that I cannot properly notate here, so I'll just describe it's constituent parts as best I can)

1. Completions in the Submission Month Per Day: The sum of number of apps completed per day after the app was filed in the submission month (monthly completions prorated by the number of days left in the month after the submission date).

2. Completions in the Following Months for Which There Is Dashboard Data per Day: The sum of the number of apps completed per day for each of the months that follow the submission month, based on the published completion figures (basically just the monthly completions data divided by the number of days in the month).

3. Hypothetical Completions in the Following Months for Which There is No Dashboard Data: The sum of the number of apps completed per day for each of the months that follow the months for which there is USCIS data, based on any hypothetical completion rate you want (hypothetical monthly completions data divided by the number of days in the month).

The function would basically count down, with on iteration equaling one day, from the Total Pending at the Time of Submission through each of the above Completions Per Day variables (taking into account the order of the months and their respective completion rates) until the number (+1 to account for the submitted app) was equal to zero. When the number reached zero, the number of iterations is equal to the number of days until completion.

If the USCIS's monthly apps pending figures were consistent with the receipt and completion numbers, this formula would work without a hitch. But, since they are not consistent, the pro-rating of receipts and completions per month produces significantly divergent outcomes for apps filed in between two months. This leads to nonsensical results, such as an app filed on 9/30 being completed 1.5 months before an app filed on 10/1 because the apps pending number in October is 2000 apps higher than completions and receipts would allow.

Ultimately, I think the only way to fix this is to simply smooth the pending data in a way that reconciles it with the completions and receipts data, perhaps by assuming every month there really is just some other process that functions like receipts and adds significantly to the over all workload and thus adding that number (which would be the difference to the receipts number and the actual increase in the number of pending) to the receipts figure for the purposes of prorating. Still gotta think about that a little more deeply, however, to make sure I'm not just making a formula that can't accurately predict anything, though...

Edited April 5, 2013 by I & B

[TexasMojo66 3]

You guys are really working overtime on the statistics of this whole process. It is very much appreciated as it gives a lot of people backup documentation when contacting the 'powers that be'. Thank you very much.

[rjm_cmyk 75]

Another thanks for all the hard work. With all the wailing and gnashing of teeth (some of them mine) it's good to see some numbers.

[mamichou 8]

Thank you for ur hard work

[Azsara 110]

Wow impressive!

[Andreea&Kevin 38]

CSC - The good news is that, as of Jan 2013, our backlog is really not that bad - 10,929 (G26 CSC, grey). If they turned around and started processing at an average rate they could be caught up in a little over 6 months (I26 CSC, blue).

Nice work, but I do see an issue with this part. The 10,929 is from Dec. 31, 2012. Really should be looking at the 11,898 from the end of Jan. For that matter you could conservatively estimate the backlog of Feb as 13000 and March as 14000. I think that is conservative because CSC hasn't improve the last 2 months and March has been the highest receipts the last 2 years. I realize you don't have a completion % for Feb & March, but using your average completions number, it would at least give a more up to date estimate for CSC to "clear" the backlog.

[Zipline 23]

Nice work, but I do see an issue with this part. The 10,929 is from Dec. 31, 2012. Really should be looking at the 11,898 from the end of Jan. For that matter you could conservatively estimate the backlog of Feb as 13000 and March as 14000. I think that is conservative because CSC hasn't improve the last 2 months and March has been the highest receipts the last 2 years. I realize you don't have a completion % for Feb & March, but using your average completions number, it would at least give a more up to date estimate for CSC to "clear" the backlog.

Good points. I wish the USCIS would publish their Feb and March data. I wrote the Ombudsman specifically yesterday asking when this data is going to be presented.

[glenmar 34]

Nice work! Thanks for sharing. I agree with Andreeea&Kevin, the current backlog (as of March) should be around 14,000. I would also say by the number of vj approvals this week, the backlog has continued to grow so far in April.

[Zipline 23]

Bump. Using this info for my Youtube project. Hoping the USCIS publishes February and March 2013 figures for I-129F soon