Ending Inventory Analysis

  • Posted: 27 July 2012 06:17 PM

    I have been looking at Ending Inventory for a bit, and believe I have come up with a way to use it to calculate next quarter sales.

    First of all, in every quarter Apple stated how much Ending Inventory they held, they described it as 4 - 6 weeks sales.  Wrong.

    In 3 of the 7 quarters where management provided Ending Inventory numbers, that inventory failed to provide 4 weeks sales.  In another quarter Ending Inventory exceeded 6 weeks.

    The actual range is 2.8 weeks to 6.4 weeks.  4- 6 weeks may be what Apple is forecasting, but their forecast accuracy is only 42%.  However, the average weeks supply for the 7 quarter study period is 4.6 weeks.

    There are other inferences that can be made from this study, including pre-launch and launch quarter volumes.  But there are only one of each, making the value of such inference limited.

    Nevertheless, I applied the average weeks supply to FQ3’s Ending Inventory and get 24,050,000 iPhones sold during FQ4.  At first blush this would suggest that an iPhone5 launch won’t occur until FQ1/2013.  It also indicates a daily run rate of 203,000.  Max run rate has been 414,000 (FQ2/2012), so a run rate of only 203K is less than half of max capacity.

    Volume of 24,050,000 iPhones this quarter accounts for the dramatic drop in gross margin guidance.  iPhone is the tail that wags the dog.

    This has happened before.  FQ4/2011 daily run rate was 185K, followed by FQ1/2012 daily run rate of 380K.  For FQ4/2011 Apple reported (October) disappointing earnings numbers, and followed that with a historical blow out (January).

    From the study I have come to feel that Apple has begun the new iPhone ramp, but because they have not taken possession of that new model inventory, they do not include it in their Ending Inventory.

    Looking forward to the new iPhone launch, it is important to note that first weekend sales of the iPhone 4S was 4MM units, a record at the time.  I expect launch weekend for the new iPhone will yield 5.5MM units (at minimum).  That means to me that iPhone 3S production has already been ceased, iPhone 4 production greatly reduced as well as iPhone 4S.  Of the 24MM that Ending Inventory infers, I think 4MM of those will be the new iPhone.  I define the launch weekend as September 28, 29 & 30.  However, September 30 is the first day of FQ1/2012, and as such that day’s sales won’t be included in FQ4/2012.

    All this tells me that industry handset seasonality has firmly set in.  That makes the December quarter the #1 sales quarter for the iPhone, with volume trailing off from there.  This is exactly opposite of what we have experienced since the introduction of the iPhone 5 years ago (linear growth).

    OK so what does this mean for FQ1/2013?  Applying the max daily run rate generates 37.7MM units (91 days vs 98 days).  That doesn’t account for an expanded user base, or switcher demand.  Ten percent YoY growth in the daily run rate, without the increase in Ending Inventory Apple reported for FQ1/2012, yields iPhone unit sales of ~44.4MM units during FQ1/2013.

    I hope this helps plan investment strategies for the next several months.

    I will be posting Mac and iPad Ending Inventory analysis in a couple weeks.  Wednesday I leave for two weeks in Kansas, moving my daughter into her new home (University of Kansas).

    [ Edited: 27 July 2012 07:12 PM by Gregg Thurman ]

    Signature

    You can’t do more, make more, be more, than the next guy, if you think like the next guy. Think different.

         
  • Avatar

    Posted: 27 July 2012 07:10 PM #1

    Makes sense.

         
  • Avatar

    Posted: 27 July 2012 09:16 PM #2

    Looks great indeed.  Sounds like its time for a spreadsheet.

    On a related note, I have a spreadsheet which tries to predict quarterly sequential increases in sales with some smoothing features.  It appears I should have left out the smoothing.

         
  • Posted: 27 July 2012 10:20 PM #3

    Tetrachloride - 28 July 2012 12:16 AM

    Looks great indeed.  Sounds like its time for a spreadsheet.

    On a related note, I have a spreadsheet which tries to predict quarterly sequential increases in sales with some smoothing features.  It appears I should have left out the smoothing.

    I use averages. Believe me they are terrific at “smoothing”, and that’s their greatest weakness.

    Signature

    You can’t do more, make more, be more, than the next guy, if you think like the next guy. Think different.

         
  • Avatar

    Posted: 27 July 2012 11:29 PM #4

    Gregg Thurman

    You certainly do your homework and your sharing is greatly appreciated.  I like that you based numbers on run rate as Apple will most certainly sell everything they can produce.

    Signature

    Waiting to be included in one of Apple’s target markets, but I still own an iPod, iPhone and iMac and APPL stock.

         
  • Avatar

    Posted: 28 July 2012 01:08 AM #5

    Gregg Thurman - 27 July 2012 09:17 PM

    I have been looking at Ending Inventory for a bit, and believe I have come up with a way to use it to calculate next quarter sales.

    First of all, in every quarter Apple stated how much Ending Inventory they held, they described it as 4 - 6 weeks sales.  Wrong.

    In 3 of the 7 quarters where management provided Ending Inventory numbers, that inventory failed to provide 4 weeks sales.  In another quarter Ending Inventory exceeded 6 weeks.

    The actual range is 2.8 weeks to 6.4 weeks.  4- 6 weeks may be what Apple is forecasting, but their forecast accuracy is only 42%.  However, the average weeks supply for the 7 quarter study period is 4.6 weeks.

    Of the 24MM that Ending Inventory infers, I think 4MM of those will be the new iPhone.

    The 4-6 week ineventory they refer to is channel inventory and this was defined during this conference call as…” Keep in mind that the channel inventory is only for the indirect channel. It’s not used to support any direct sales and direct sales are Apple retail sales, Apple online sales and Apple education sales…I know that there are some companies out there that refer to their channel inventory in terms of their gross sales but we don’t do that because we feel very strongly that it only supports the channel flows, so that’s how we calculate it.”. 

    So does this change your thinking on the predictive nature of Apple ending inventory?  It does mine.

    Gregg said…“Of the 24MM that Ending Inventory infers, I think 4MM of those will be the new iPhone.”. Huh?  How are these dots connected?

         
  • Posted: 28 July 2012 01:44 AM #6

    madmaxroi - 28 July 2012 04:08 AM
    Gregg Thurman - 27 July 2012 09:17 PM

    I have been looking at Ending Inventory for a bit, and believe I have come up with a way to use it to calculate next quarter sales.

    First of all, in every quarter Apple stated how much Ending Inventory they held, they described it as 4 - 6 weeks sales.  Wrong.

    In 3 of the 7 quarters where management provided Ending Inventory numbers, that inventory failed to provide 4 weeks sales.  In another quarter Ending Inventory exceeded 6 weeks.

    The actual range is 2.8 weeks to 6.4 weeks.  4- 6 weeks may be what Apple is forecasting, but their forecast accuracy is only 42%.  However, the average weeks supply for the 7 quarter study period is 4.6 weeks.

    Of the 24MM that Ending Inventory infers, I think 4MM of those will be the new iPhone.

    The 4-6 week ineventory they refer to is channel inventory and this was defined during this conference call as…” Keep in mind that the channel inventory is only for the indirect channel. It’s not used to support any direct sales and direct sales are Apple retail sales, Apple online sales and Apple education sales…I know that there are some companies out there that refer to their channel inventory in terms of their gross sales but we don’t do that because we feel very strongly that it only supports the channel flows, so that’s how we calculate it.”. 

    So does this change your thinking on the predictive nature of Apple ending inventory?  It does mine.

    Gregg said…“Of the 24MM that Ending Inventory infers, I think 4MM of those will be the new iPhone.”. Huh?  How are these dots connected?

    Baysian statiscal analysis, wherein you can make a fairly accurate estimation with incomplete, or seemingly unrelated data.

    If iPhone sales total 24 MM units this quarter, and the new iPhone is launched on Sepetember 28, then 4MM of the 24MM will be the new iPhone.  I arrived at 4MM by looking back to the iPhone 4S launch in the first weekend (3 days) Apple sold 4MM units i believe launch weekend will exceed that amount by at least 1MM units, with Sunday sales being allocated to FQ1 I failed to include that in my original post. Thank you for your question.

    [ Edited: 28 July 2012 01:54 AM by Gregg Thurman ]

    Signature

    You can’t do more, make more, be more, than the next guy, if you think like the next guy. Think different.

         
  • Avatar

    Posted: 28 July 2012 03:12 AM #7

    Gregg,

    You didn’t address max’s point about this being only indirect channel inventory. iPhones sold at the Apple Stores and apple.com are not counted. Also, we don’t know if the inventory is looking forward at a projected average per week for the whole quarter or just at the first 4-6 weeks of the quarter.

    I suspect it’s the latter. In other words, Apple expects the 8.5 million iPhones in the channel at the end of the quarter to have been sold 4 - 6 weeks into the quarter, on top of any sold at Apple stores and apple.com. This would imply nothing about the rate of sales in later weeks and is not a forecast of the rate of sales for the entire quarter.

    Since Apple doesn’t report how many iPhones are sold directly to end customers vs. through the indirect channel, it makes it hard to use ending inventory to project iPhone sales.

    Signature

    The only way to be truly satisfied is to do what you believe is great work. And the only way to do great work is to love what you do. — Steve Jobs

         
  • Posted: 29 July 2012 03:52 PM #8

    Apple II+ - 28 July 2012 06:12 AM

    Gregg,

    You didn’t address max’s point about this being only indirect channel inventory. iPhones sold at the Apple Stores and apple.com are not counted. Also, we don’t know if the inventory is looking forward at a projected average per week for the whole quarter or just at the first 4-6 weeks of the quarter.

    I suspect it’s the latter. In other words, Apple expects the 8.5 million iPhones in the channel at the end of the quarter to have been sold 4 - 6 weeks into the quarter, on top of any sold at Apple stores and apple.com. This would imply nothing about the rate of sales in later weeks and is not a forecast of the rate of sales for the entire quarter.

    Since Apple doesn’t report how many iPhones are sold directly to end customers vs. through the indirect channel, it makes it hard to use ending inventory to project iPhone sales.

    I thought I had when I referenced Bayesian statistical analysis.

    During the ‘60s the CIA wanted to know how many Cuban troops were in Angola. They gave the task to mathematician Bayes who counted the number of baseball diamonds in Angola, and compared that number to those on military bases in Cuba. Knowing the number of players per team, soldiers required to support a team, number of teams to support a league, etc, Bayes deduced the number of Cuban soldiers in Angola.

    The important aspect of Bayesian statistical analysis is that data need not be complete to make an inference of probability.  It also makes provision for data updates.

    By comparing channel ending inventory supply (incomplete data) you can infer a future probability from past experience.

    For my FQ4 iPhone estimate I compared past ending inventory supply to subsequent quarter results, and applied the average of that ratio to June quarter ending inventory.

    For December quarter results (not knowing September quarter ending inventory) I used daily run rate from the March quarter. I used this quarter vs December quarter run rate because I believe Apple will launch the new iPhone in the latter half of September to provide more ramp up time for December quarter sales. As noted previously, December quarter sales are the handset industry high point of the year.

    Signature

    You can’t do more, make more, be more, than the next guy, if you think like the next guy. Think different.

         
  • Avatar

    Posted: 29 July 2012 05:14 PM #9

    Gregg Thurman - 29 July 2012 06:52 PM
    Apple II+ - 28 July 2012 06:12 AM

    Gregg,

    You didn’t address max’s point about this being only indirect channel inventory. iPhones sold at the Apple Stores and apple.com are not counted. Also, we don’t know if the inventory is looking forward at a projected average per week for the whole quarter or just at the first 4-6 weeks of the quarter.

    I suspect it’s the latter. In other words, Apple expects the 8.5 million iPhones in the channel at the end of the quarter to have been sold 4 - 6 weeks into the quarter, on top of any sold at Apple stores and apple.com. This would imply nothing about the rate of sales in later weeks and is not a forecast of the rate of sales for the entire quarter.

    Since Apple doesn’t report how many iPhones are sold directly to end customers vs. through the indirect channel, it makes it hard to use ending inventory to project iPhone sales.

    I thought I had when I referenced Bayesian statistical analysis.

    During the ‘60s the CIA wanted to know how many Cuban troops were in Angola. They gave the task to mathematician Bayes who counted the number of baseball diamonds in Angola, and compared that number to those on military bases in Cuba. Knowing the number of players per team, soldiers required to support a team, number of teams to support a league, etc, Bayes deduced the number of Cuban soldiers in Angola.

    The important aspect of Bayesian statistical analysis is that data need not be complete to make an inference of probability.  It also makes provision for data updates.

    By comparing channel ending inventory supply (incomplete data) you can infer a future probability from past experience.

    For my FQ4 iPhone estimate I compared past ending inventory supply to subsequent quarter results, and applied the average of that ratio to June quarter ending inventory.

    For December quarter results (not knowing September quarter ending inventory) I used daily run rate from the March quarter. I used this quarter vs December quarter run rate because I believe Apple will launch the new iPhone in the latter half of September to provide more ramp up time for December quarter sales. As noted previously, December quarter sales are the handset industry high point of the year.

    In Bayesian statistics you need two things to infer the (posterior) probability. The prior (probability) and the likelihood function. I assume that the inventory is your prior and the quaterly results your likelihood function?

    Signature

    “Anyone who has never made a mistake has never tried anything new.” Albert Einstein

         
  • Posted: 30 July 2012 03:20 AM #10

    Big Al - 29 July 2012 08:14 PM
    Gregg Thurman - 29 July 2012 06:52 PM
    Apple II+ - 28 July 2012 06:12 AM

    Gregg,

    You didn’t address max’s point about this being only indirect channel inventory. iPhones sold at the Apple Stores and apple.com are not counted. Also, we don’t know if the inventory is looking forward at a projected average per week for the whole quarter or just at the first 4-6 weeks of the quarter.

    I suspect it’s the latter. In other words, Apple expects the 8.5 million iPhones in the channel at the end of the quarter to have been sold 4 - 6 weeks into the quarter, on top of any sold at Apple stores and apple.com. This would imply nothing about the rate of sales in later weeks and is not a forecast of the rate of sales for the entire quarter.

    Since Apple doesn’t report how many iPhones are sold directly to end customers vs. through the indirect channel, it makes it hard to use ending inventory to project iPhone sales.

    I thought I had when I referenced Bayesian statistical analysis.

    During the ‘60s the CIA wanted to know how many Cuban troops were in Angola. They gave the task to mathematician Bayes who counted the number of baseball diamonds in Angola, and compared that number to those on military bases in Cuba. Knowing the number of players per team, soldiers required to support a team, number of teams to support a league, etc, Bayes deduced the number of Cuban soldiers in Angola.

    The important aspect of Bayesian statistical analysis is that data need not be complete to make an inference of probability.  It also makes provision for data updates.

    By comparing channel ending inventory supply (incomplete data) you can infer a future probability from past experience.

    For my FQ4 iPhone estimate I compared past ending inventory supply to subsequent quarter results, and applied the average of that ratio to June quarter ending inventory.

    For December quarter results (not knowing September quarter ending inventory) I used daily run rate from the March quarter. I used this quarter vs December quarter run rate because I believe Apple will launch the new iPhone in the latter half of September to provide more ramp up time for December quarter sales. As noted previously, December quarter sales are the handset industry high point of the year.

    In Bayesian statistics you need two things to infer the (posterior) probability. The prior (probability) and the likelihood function. I assume that the inventory is your prior and the quaterly results your likelihood function?

    Email me at .(JavaScript must be enabled to view this email address)

    Signature

    You can’t do more, make more, be more, than the next guy, if you think like the next guy. Think different.

         
  • Avatar

    Posted: 30 July 2012 04:31 AM #11

    Gregg,

    Thanks for clarifying. Your reply to max seemed like an answer only to his second question beginning “how” rather than his “yes or no” question.

    Also, with respect, and although a lot of what you posted is correct, I think you invited these questions, because your rhetoric described Apple’s 4-6 weeks number as “wrong”, “inaccurate”, and a forecast of the entire quarter sales even though it applies only to the indirect channel.

    Gregg Thurman - 29 July 2012 06:52 PM

    For my FQ4 iPhone estimate I compared past ending inventory supply to subsequent quarter results, and applied the average of that ratio to June quarter ending inventory.

    I got that from your original post, and I think it’s one take worth considering, but I’m trying to gauge how useful it is. While an average ratio of 4.6 weeks between actual total sales and previous quarter ending indirect channel inventory may get you 24 million based on the 8.5 million in the channel, if the actual ratio has varied between 2.8 to 6.4 weeks, why should we think that the 4.6 average gives anything resembling an accurate estimate in any particular quarter? What is the likelihood that 24 million is close to right this time?

    Also, not clear on where the 203k daily run rate comes from.

    Signature

    The only way to be truly satisfied is to do what you believe is great work. And the only way to do great work is to love what you do. — Steve Jobs

         
  • Posted: 30 July 2012 11:39 AM #12

    Big Al - 29 July 2012 08:14 PM
    Gregg Thurman - 29 July 2012 06:52 PM

    I thought I had when I referenced Bayesian statistical analysis.

    During the ‘60s the CIA wanted to know how many Cuban troops were in Angola. They gave the task to mathematician Bayes who counted the number of baseball diamonds in Angola, and compared that number to those on military bases in Cuba. Knowing the number of players per team, soldiers required to support a team, number of teams to support a league, etc, Bayes deduced the number of Cuban soldiers in Angola.

    The important aspect of Bayesian statistical analysis is that data need not be complete to make an inference of probability.  It also makes provision for data updates.

    By comparing channel ending inventory supply (incomplete data) you can infer a future probability from past experience.

    For my FQ4 iPhone estimate I compared past ending inventory supply to subsequent quarter results, and applied the average of that ratio to June quarter ending inventory.

    For December quarter results (not knowing September quarter ending inventory) I used daily run rate from the March quarter. I used this quarter vs December quarter run rate because I believe Apple will launch the new iPhone in the latter half of September to provide more ramp up time for December quarter sales. As noted previously, December quarter sales are the handset industry high point of the year.

    In Bayesian statistics you need two things to infer the (posterior) probability. The prior (probability) and the likelihood function. I assume that the inventory is your prior and the quaterly results your likelihood function?

    In a word, yes.

    Signature

    You can’t do more, make more, be more, than the next guy, if you think like the next guy. Think different.

         
  • Posted: 30 July 2012 11:47 AM #13

    Apple II+ - 30 July 2012 07:31 AM

    Gregg,

    Thanks for clarifying. Your reply to max seemed like an answer only to his second question beginning “how” rather than his “yes or no” question.

    Also, with respect, and although a lot of what you posted is correct, I think you invited these questions, because your rhetoric described Apple’s 4-6 weeks number as “wrong”, “inaccurate”, and a forecast of the entire quarter sales even though it applies only to the indirect channel.

    Gregg Thurman - 29 July 2012 06:52 PM

    For my FQ4 iPhone estimate I compared past ending inventory supply to subsequent quarter results, and applied the average of that ratio to June quarter ending inventory.

    I got that from your original post, and I think it’s one take worth considering, but I’m trying to gauge how useful it is. While an average ratio of 4.6 weeks between actual total sales and previous quarter ending indirect channel inventory may get you 24 million based on the 8.5 million in the channel, if the actual ratio has varied between 2.8 to 6.4 weeks, why should we think that the 4.6 average gives anything resembling an accurate estimate in any particular quarter? What is the likelihood that 24 million is close to right this time?

    Also, not clear on where the 203k daily run rate comes from.

    Questions are always good,  don’t mind them at all.

    (Ending inventory minus beginning inventory plus unit sales) divided by total days in quarter

    Signature

    You can’t do more, make more, be more, than the next guy, if you think like the next guy. Think different.

         
  • Avatar

    Posted: 30 July 2012 03:04 PM #14

    Gregg Thurman - 30 July 2012 02:47 PM
    Apple II+ - 30 July 2012 07:31 AM

    Also, not clear on where the 203k daily run rate comes from.

    Questions are always good,  don’t mind them at all.

    (Ending inventory minus beginning inventory plus unit sales) divided by total days in quarter

    8300k - 8600k + 26028k = 25728k

    25728k / 91 days = 283k

    203k a typo then?

    Signature

    The only way to be truly satisfied is to do what you believe is great work. And the only way to do great work is to love what you do. — Steve Jobs

         
  • Posted: 31 July 2012 02:25 AM #15

    Apple II+ - 30 July 2012 06:04 PM
    Gregg Thurman - 30 July 2012 02:47 PM
    Apple II+ - 30 July 2012 07:31 AM

    Also, not clear on where the 203k daily run rate comes from.

    Questions are always good,  don’t mind them at all.

    (Ending inventory minus beginning inventory plus unit sales) divided by total days in quarter

    8300k - 8600k + 26028k = 25728k

    25728k / 91 days = 283k

    203k a typo then?

    Not sure where you got your inventory numbers.  I show FQ3 Beginning 8,550,000 Ending 5,550,000

    For FQ4 I don’t have an ending inventory number, and assuming a September 28 iPhone launch, I took my units sold estimate (24,050,000) and deducted Beginning inventory (5,550,000).  I divided that result by 91 to get the 203K run rate.  If Apple were to complete FQ4 with 8,000,000 iPhones (by way of example) the “indicated” run rate would increase to 291K.

    Signature

    You can’t do more, make more, be more, than the next guy, if you think like the next guy. Think different.