Sink your Teeth into the Data - Time from Order to Delivery

[Mdg56]

I've been a (mostly) silent observer of this forum for a while, but couldn't help but dig into the database built and maintained by turbo2ltr (great job!). I looked at the elapsed time between order dates and actual delivery dates for those who've received their cars to date, and plotted them in a couple of ways on control charts. I'll explain some of what these charts mean for those not versed in statistical process control -- you can gain a lot of insight with these tools.

First, I plotted the elapsed time from order to delivery (let's call it "span" for short) on an Individuals-Moving Range chart, in order of actual delivery date. The top chart here is simply a time series plot of the span sorted by actual delivery date. Each point is a delivery. The points on the bottom chart (the "MR" chart) are simply the difference between the span of one delivery date minus the span of the previous date. This is useful in estimating the expected variation from date to date: this variation is represented by the two red lines on the top chart -- the Upper and Lower Control Limits (UCL and LCL). You'd be able to tell that a process was stable and consistent if all of the black points fell inside the Control limits on the top chart. Points outside the control limits on this chart are flagged as red with the footnote of "1" -- this is the result of a statistical test showing that these points are unexpected, or "out of control".

It's obvious at first glance that the Nissan Delivery process is neither consistent nor stable. Focusing on the top chart, you can see that the first two points were the mad rush to deliver the first Leafs in December 2010 (hooray for Gudy!). Then starting in January, the initial batch of orders were received with reasonable consistency; however, you can see the span creep up slowly and really ramp up to approaching 200 days before the beginning of March. Then, in mid-April, you see the process go haywire. Some people start receiving their cars in 100 days or less, while others are hanging out for two-hundred days or more (Note: I'm only plotting delivered vehicles, so I'm not showing orders that haven't arrived).

One can hypothesize a number of "special causes" associated with this unpredictable behavior. The Tsunami, of course, and the software update...but I'd challenge someone in the supply chain logistics group at Nissan to explain this unpredictable behavior.

I've got another way of looking at this data that I'll present in a second post. In the mean time, I'm interested in what other people make of this. What other theories do you have based on what you see?

Cheers,

 

[Mdg56]

Ok, Here's another way to look at the same data. We're still looking at span, but now sorted by order date rather than by actual delivery date. So this shows how the span (again, elapsed time from order to delivery) varies as people who have received cars have placed orders.

Again, you can see that the folks that ordered in August 2010 by and large received their cars (in general) in 150 days or less; however, it seems that there's a tail prior to the orders on the first of the month where the span increases a lot. People who ordered between September and November experienced elapsed times up around and over 200 days.

But then here's the crazy thing. If you ordered your car in December or later, you were more likely to get your car in around 100 days. This indicates in part that Nissan did ramp production, but something else is going on -- it appears they started to straighten some things out around their delivery process (or maybe their order process).

This doesn't tell the whole story either, though. There's a lot of data that's not here, represented by people who have ordered but not yet received their cars (including me). See my next post for that one.

 

[Mdg56]

Ok, last post. This chart is called an "Xbar-R" chart. It's similar to the charts I've shown before, but the data's organized a little differently. Each point in the top chart represents the average of the elapsed time from order to delivery (Span) for all orders placed on a given date. Each point in the bottom chart represents the range of span for those orders placed. A high number in the range chart indicates a large amount of variation in span for people that ordered on a given day.

In this data set, I've included orders that have not yet been delivered -- those are the gaps in the data (see if you can find yours). I've also done some fancy footwork to simplify how the control limits are calculated -- these control limits now show the expected variation in span based on the average range of span across all the order dates.

This plot shows things a little differently. If you ordered between August and October, you could expect to have received your car in 130-220 days (there are some noteably painful exceptions). Then, from November and December, there are a lot of gaps. Some people received their cars in the neighborhood of 150 days, but many of us have yet to receive our vehicles.

And then, starting in January 2011, the elapsed time from order to delivery for some dropped precipitously. I'd be very interested in what factors drove these people to get their cars so much sooner. I've seen speculation in other threads that cars may be batched by boat (likely), or by color, or by region...I haven't taken the time to mine the spreadsheet further as yet.

What are your theories? How would you look at the data?

A final thought: this chart elicited the following quote from my wife: "If we had waited a few months more to order, we'd probably have our car by now." :lol: :lol: :lol: :lol:

 

[walterbays]

Very interesting charts. Thanks for posting!

I wonder if you might do some of the same analysis looking at reservation date rather than order date. Reservation date is truly an independent variable, in that it was selected by the customer. Order date is mostly another dependent variable, like delivery date, in that it was determined by Nissan, not by the customer. A large fraction of the initial orders, stretching over months, came from reservations all made within a few days in April. True wait time should be measured from reservation, not from order.

 

[Glenn]

A final thought: this chart elicited the following quote from my wife: "If we had waited a few months more to order, we'd probably have our car by now." :lol: :lol: :lol: :lol:

Very interesting analysis, M!

And, with a January 20th Order Date, if we had waited just one more day to order, we almost assuredly would have our car by now!! Ah well...

 

[planet4ever]

I've mentioned my theory a couple of times in other threads. It's only a theory; no one has so much as whispered to me that anyone inside Nissan has admitted to it.

The theory is that someone in Nissan NA, or some contractor they hired, set up a database to track orders, but very foolishly made the order date a character string rather than a proper date field. Most Americans use an extremely illogical date format -- month/day/year. [Why on earth should the least significant part of the date go between the two more significant parts?] But that's the format I suggest this clueless person used. Everything started along fine for orders in 09/xx/2010 through 12/xx/2010. The earliest orders sorted lowest, and were the first ones they worked on filling. Then came the 01/xx/2011 orders. These suddenly popped to the top of the list, as if they were the earliest ones, crowding out the fall orders that had not yet been filled. Of course the cars couldn't be built and shipped instantly, but starting in late March and early April, everything went to hell.

Obviously it doesn't really matter whether the format was 01/xx/2011 or 01/xx/11, the result would have been the same. Curiously it would still have been the same if dates were entered as 1/xx/2011 or even 1/x/2011 because in the ASCII collating order "/" comes before "0". So 1/29/2011 would still come before 10/1/2010.

I can't imagine this insanity (if it truly existed) could have continued for very many weeks before someone inside Nissan realized they had a serious problem. But by then the damage was done and, given the obviously poor quality of Nissan's customer interactions (the My Account website, the ill-prepared customer service people, the crazily illogical date estimates), I doubt if they had the skills and resources to undo it.

Ray

 

[DaveEV]

I posted elsewhere what Brendan Jones @ Nissan told me about their order/production sytem.

Essentially - there are 2 queues - pre-production order queue where orders wait to be scheduled and can be re-ordered and a scheduled-production queue where orders can not be re-ordered or assigned.

What happened is that after the quake scheduled-production runs were canceled due to parts shortages and had to be put back into the pre-production queue. But by then the production queue already started filling up with some Jan/Feb orders.

So if you were unlucky and your car's production had to be re-scheduled - that would essentially mean a 1-month delay - and some lucky Jan/Feb orders that got into the production schedule.

If you were really un-lucky - your car might get to the front and be rescheduled again - leading to 2-month delay. Of course - if this happened I would expect to see a couple bands in the data - I think I see that in the "I-MR Chart of Days from order to delivery"...

The timing of when the out-of-order deliveries started showing up fits - about a 6 weeks after the quake which would have been when the first post-quake built cars should have started reaching dealers...

 

[BlueSL]

As fond as I have become of Brendan Jones for finding me a car and personally calling me, I do not believe this explanation for one minute.

By March 24 (just one day after the Oppama factory re-started production if memory serves), some on this board were becoming apoplectic over the obvious mis-assignment of VINs to January and February orders. (Indeed, March 24 is the day on which Nissan corporate restored Ttweed's delivery status with a new April delivery projection). There was a spread of 30 such lines of the spreadsheet at that time, ranging from lines 359 to 389. See below from the May Deliveries! Or ... May(be) thread:

I called my PD today and requested more info on what "Pending" means right now, with last 2 ships of Leafs in Long Beach. He called his Nissan contact and got back to me soon after, with this info:

Nissan is now well aware of discontent from Leaf orders in the middle of the queue that were passed over for later orders. They are going to try to be extra careful to be fair to the original order when they process cars off these two ships.
They could not confirm whether or not I would get a car from one of these last ships, but it did not look likely.

The inability to fix the problem by simply making a lot of Leafs fast was affected by the earthquake, no doubt. But, the screw up in order assignment had nothing to do with the earthquake. It pre-dated that event. Indeed, about three weeks after the quoted post, orasadil and BrianSanDiego and others began getting their pre-earthquake cars -- roughly 90 days after they placed their orders.

 

[rnkepler]

I thought it was a combination of the two. Dree's explanation for the mess up post earthquake and something along the line of what planet4ever is theorizing for the September/November mess. Wasn't someone (maybe Tronz?) told by Nissan that a vendor had released a batch of orders out of sequence that caused the Jan/Feb orders to get ahead of Sept/Oct?

 

[occ]

Cool! Thanks Mdg56 for the analysis.

We all know that Nissan screwed up the order queue. I don't know if they have admitted publicly, but definitely admitted through our persistent MNL members. The pre-production/re-schedule explanation is lame, IMHO that tries to deflect a major screw up.

I think planet4ever has a good theory, software get that kind of mistake all the time, esp. when it is being applied without testing -- I think Nissan rushed the whole process and screwed it up. Well...good that they rushed it for us consumers, bad that they screwed up.

As for the apparent saw-tooth patterns within the small windows of data, that's very "batch" like process. Other outlier points may be due to good "squeakers" among MNL members that got some attentions and cars earlier. Obviously, Nissan did nothing to fix the snafu as a whole, but did somethings for the squeakers.

 

[BlueSL]

I know of one "squeaker" situation where a spreadsheet participant was completely hosed by an incompetent dealer and Nissan made it right and advanced his order priority by several months to correct for the dealer's mistake. There had to be others.

Planet4ever's theory made a ton of sense to me until I remembered that not all January orders got the special treatment. IIRC, no order before 1/6 was advanced to the front of the line. I have a late March spreadsheet I saved and will double-check this.

It is almost as if a human being was doing cut and pastes from the Nissan North America Leaf software reservation system into the Oppama factory system and simply cut and pasted the wrong block of orders -- failing to go back to early September where he/she had left off.

Drees, what do you think about my timing critique of the "explanation?" I do not think the "two sequence" concept holds up.

 

[onlyjaymoo]

This is great data -- I'd also add that I noticed by looking at the Spreadsheet that there is a noticeable gap in the VINs between 2600 -> 3250. I wonder what happened to these cars -- the rest of the Spreadsheet seems to have a "normal" distribution of gaps between VINs...

I wonder what Nissan is going to do "in the future" about this whole process, their vendors, etc... but I think it would take some smart analysis like this and maybe hiring some smart folks to redo it If you're reading this, Nissan, I'd be happy to come help you guys out!

 

[OilFreedom]

Great analysis, Mdg56! Thanks for sharing.

 

[turbo2ltr]

Two similar graphs are on the second worksheet in the spreadsheet.

Drees is now maintaining the spreadsheet, so thank him!

 

[leafkabob]

I'd also add that I noticed by looking at the Spreadsheet that there is a noticeable gap in the VINs between 2600 -> 3250. I wonder what happened to these cars

Dealer demos probably.

 

[LEAFfan]

As fond as I have become of Brendan Jones for finding me a car and personally calling me, I do not believe this explanation for one minute.
By March 24 (just one day after the Oppama factory re-started production if memory serves), some on this board were becoming apoplectic over the obvious mis-assignment of VINs to January and February orders. (Indeed, March 24 is the day on which Nissan corporate restored Ttweed's delivery status with a new April delivery projection). There was a spread of 30 such lines of the spreadsheet at that time, ranging from lines 359 to 389. See below from the May Deliveries! Or ... May(be) thread:
The inability to fix the problem by simply making a lot of Leafs fast was affected by the earthquake, no doubt. But, the screw up in order assignment had nothing to do with the earthquake. It pre-dated that event. Indeed, about three weeks after the quoted post, orasadil and BrianSanDiego and others began getting their pre-earthquake cars -- roughly 90 days after they placed their orders.

+1! Exactly! He also told me that there were 60-80 early orders that never were forwarded to production so that comp programmer could have been the reason.

 

[nissanvirgin]

Planet4ever's theory made a ton of sense to me until I remembered that not all January orders got the special treatment. IIRC, no order before 1/6 was advanced to the front of the line. I have a late March spreadsheet I saved and will double-check this.

Planet4ever's theory as currently formulated is insufficient although elegant. Indeed our order was 1/12 and we are at the very end. Glenn says his was 1/20 and is at the end. Apparently 1/21 was the cutoff.

If you look at the iconic representation of the data at http://www.richard.kelly.ws/wheresmyleaf you will see 1/21 was the day a new batch of orders were allowed. Apparently everybody before 1/21 was in an earlier batch (and I could have ordered in September if I had noticed I didn't get my RAQ email).

Thus I think a better theory would be something along the lines that Nissan added a second sheet to their list starting on 1/21, kind of like adding a second worksheet to a spreadsheet file. Maybe the first had filled up some aspect of their storage system or their customer service system, or maybe Nissan switched contractors and needed to start with a fresh sheet after 1/21. Then the two sheets were supposed to be processed sequentially but some signal was issued that falsely indicated the first sheet was finished.

Another possibility that I have considered is that there were some VERY VERY important people in the post-1/21 order batch who had friends in high places in Nissan and Nissan thought they would let these people cut in line, hoping the victims wouldn't notice.

I would love to know the truth. Incompetence or deviousness? Regardless bad customer service, made worse by their refusal to offer a plausible explanation.

 

[BrendanDolan]

I would love to know the truth. Incompetence or deviousness? Regardless bad customer service, made worse by their refusal to offer a plausible explanation.

They aren't being devious. It happened to many people in my order queue, and I can say with 100% conviction that those people didn't have any back channels to Nissan to get things moved ahead. I'm calling database error personally, but that's just one man speculating.

 

[Reddy]

Well I ordered Jan 14th and my delivery estimate is still Month of August. Nissan wouldn't allow me to order earlier because I didn't jump on the reservation system on the first day. Whatever happened to the database, it happened after that date. I guess I missed being a "lucky ducky" by just a few days. Hmmm, does that make me an "unlucky ducky"?

Reddy

 

[buddf]

Wouldn't it be funny if this all stemmed from one person? One person with big pull wanted their car sooner - they said sure, we can move around one vehicle. Then the system freaked out, started assigning cars from orders at that point on. When it was finally noticed it was too late?! That one person would not even know what kind of mayhem they and nissan inadvertently caused...