How a more customer-centric “Newsvendor Model” can increase profits in the face of uncertain demand: an iPhone 12 case study.
As is the case every fall, Apple is gearing up for the release of a new model of iPhones: the iPhone 12. However, behind the scenes, Apple is undertaking the daunting task of accurately matching supply and demand; a task rendered exceptionally difficult given the current Covid-19 crisis. The challenge stems from the fact that Apple must commit to a certain supply before it can observe the (stochastic) demand for the upcoming iPhone 12. Therefore, the company risks producing too many (i.e. supply exceeds demand) or producing too few (i.e. demand exceeds supply). The Newsvendor Model is ideal for tackling this dilemma because it allows Apple to calculate an optimal production quantity that accounts for these risks.
In this article, we will use the iPhone 12 case to demonstrate the importance of collaboration between the Operations and Marketing functions, and specifically, the role that a customer segmentation can play in reducing mismatch costs and maximizing profit. While we will use the upcoming iPhone 12 as the subject of our analysis, the key take-outs are relevant in many industries where firms face long lead times and relatively short selling seasons characterized by stochastic demand, such as fast fashion, perishable goods, consumer electronics; and in industries where there is variability in demand coupled with inflexible supply, such as in the airlines industry, the hotel industry or the car rental industry.
We will start by introducing a classic single-period Newsvendor Model for the upcoming iPhone 12 in the US market, and then look at two possible extensions to increase profits. The first extension is an Operations-driven Reactive Capacity model, most likely used by Apple, that significantly reduces mismatch costs. The second extension, especially relevant when Reactive Capacity is not an option, involves a yield management approach. This approach explores the Marketing / Operations interface and highlights the importance of accounting for key differences in customer reactions when faced with a stockout. Indeed, some customers may be more “forgiving” and opt for another Apple product, while others may switch to a competitor. Coupling the Newsvendor Model with protection levels for specific customer segments results in a more sophisticated and customer-centric approach to operations management, thereby minimizing risk and maximizing profit.
(Disclaimer: we do not have access to internal data. All data used in this analysis is based on publicly available information. Furthermore, there are certain simplifying assumptions that were made. These assumptions are clearly indicated throughout the article.)
The upcoming iPhone 12 and Apple’s timeline of events
Apple announced the new iPhone 12 will be released on October 13th, 2020[efn_note]https://www.tomsguide.com/news/iphone-12-release-date-price-specs[/efn_note] . The company is planning the release of four different models, as shown in Exhibit 1. The models range from the basic iPhone 12 model priced at $649 to the high-end iPhone 12 Pro Max priced at $1,199. Despite differences in storage, memory, display and camera, all iPhone 12 models come with 5G capability, a new design, facial recognition, and a next-generation A-series chip.
Exhibit 1: Details of the 4 upcoming iPhone 12 models
As illustrated in the simplified timeline of events (Exhibit 2), Apple must generate a demand forecast for the upcoming iPhone and submit orders to its network of suppliers (in June) before it is able to observe the actual market demand following the launch in September / October. Mass production then begins approximately one month prior to the launch date and continues throughout the selling season. At the end of the selling season, leftover units are either discounted, as is the case for base models, or discontinued, as is typically the case for high-end models.
Exhibit 2: Apple’s Timeline of Events[efn_note]This is a rough approximation based on several sources of data. We do not have internal access to Apple’s operations processes.[/efn_note]
The single-period Newsvendor Model
As a starting point, we will assume that Apple must commit to a production quantity with its suppliers before the selling season and is unable to launch a second wave of production during the selling season. We will then move to the more realistic Reactive Capacity model.
The first step in setting up the Newsvendor Model is constructing a demand forecast. To this end, we surveyed a group of 20 industry experts that either work in the smartphone industry or consider themselves experts on the subject. Each of these experts provided a specific point estimate for the iPhone 12 demand in the US along with their rationale. In the survey, we provided the respondents with historical sales data for previous models, which allowed us to then assume a normal distribution. The survey yielded a mean (µ) of 24,557,500 units and a standard deviation (σ) of 3,292,786, thereby giving a coefficient of variation (CV) of 3,292,786 / 24,557,500 = 0.13. It is worthwhile noting that the mean is below the historical data provided for the last generation of iPhones (iPhone 11) and that the forecast variability is rather low meaning most experts agree on softening demand. The key reason provided is the impact that Covid-19 will have on consumer spending and price elasticity.
Furthermore, each expert provided a breakdown of their estimate by model, which allowed us to run the Newsvendor Model for each of the four product derivatives. For purposes of brevity, we will focus our analysis here on the high-end iPhone 12 Pro Max. The reason for this is the higher volatility in demand estimates, which makes for a particularly interesting case. The mean (µ) demand estimate for the iPhone 12 Pro Max is 4,159,000 units with a standard deviation (σ) of 2,131,392, yielding a coefficient of variation (CV) of 2,131,392 / 4,159,000 = 0.51. Contrary to the CV for the total iPhone 12 demand, the forecast variability for the iPhone 12 Pro Max is much higher signalling an interesting divergence in opinion regarding the success of this higher priced model.
The economics of the iPhone 12 Pro Max
As depicted in Exhibit 1, the assumed selling price of the iPhone 12 Pro Max 256GB is $1,199. On the other hand, the cost of production is slightly more difficult to obtain. A recent investigative report from Tech Insight and NBC[efn_note]https://www.nbcnews.com/now/video/how-much-does-it-cost-to-make-an-iphone-11-70289989901[/efn_note] estimates the total cost of goods sold per unit for last year’s iPhone 11 Pro Max to be $490.50. We have estimated other variable costs, including labour for manufacturing and assembly and expenses for shipping to be $50 per unit. Furthermore, we are assuming a 0% retailer margin, given the tendency for carriers to subsidize iPhones in the US. The final component is the holding cost, which is a function of Apple’s annual cost of inventory and inventory turn. While we were unable to obtain the exact cost of inventory for Apple, we have estimated it to be 40%. This figure reflects the high cost of obsolescence, the risk of theft, and the need for warehousing, security, and insurance. As for Apple’s inventory turns[efn_note]This is an approximation given that Apple does not provide specific business line financials.[/efn_note], we used Apple’s 2019 financial statements (Yahoo! Finance 2019) to calculate as follows: Inventory Turns = COGS / Inventory = $161,782,000 / $4,106,000 = 39.4 turns . Therefore, per-unit inventory costs = Annual inventory costs / Inventory turns = 40% / 39.4 = 1.01%. This implies that each iPhone 12 Max Pro has a $490.50 * 1.01% = $4.95 inventory cost. This brings the total cost per unit to $490.50 + $50 + $4.95 = $545.45. This results in a contribution margin of approximately 55%, which is in line with what has been reported[efn_note]https://www.phonearena.com/news/Profit-margins-on-the-iPhone-have-fallen-to-60_id111023[/efn_note]. Finally, the high-end models (such as the iPhone 12 Pro Max) are generally discontinued at the end of the selling season. Therefore, for the purposes of this model, we will assume a $0 salvage value.
Running the Newsvendor Model
The first step in running the Newsvendor Model is to calculate the critical ratio, which is a function of the underage (Cu) and overage costs (Co). The underage cost is equal to the foregone margin on a lost sale = $1,199 – $545.45 = $653.55 and the overage cost is the difference between the cost of production and the salvage value = $545.45 – $0 = $545.45. The critical ratio is then Cu / (Co + Cu) = 653.55 / (545.45 + 653.55) = 0.55. This means that the optimal quantity (Q) of iPhone 12 Pro Max’s to produce is such that there is a 55% probability that actual demand is Q or lower. After finding the z-statistic, of 0.11, we can calculate the optimal production quantity, as follows: Q* = µ + z * σ = 4,159,000 + 0.11 * 2,131,392 = 4,400,355. This amount is higher than the mean, which is intuitive given the risk profile that pits the slightly higher cost of underage ($653.55) against the slightly lower cost of overage ($545.45). This means that Apple should produce 4,400,355 iPhone 12 Pro Max’s for the upcoming selling season. The model also yields a host of performance measures, as summarized in Exhibit 4. It is interesting to note an expected profit of $1,705,117,754 and a significant mismatch cost of $1,012,996,740 mainly due to the relatively high coefficient of variation for this model (0.51).
Exhibit 4: Single Period Newsvendor Model
The extended Newsvendor Model: quick response with reactive capacity solution
The reactive capacity approach involves the possibility of Apple launching a second production wave during the selling season once actual demand is observed and the forecast readjusted. This solution is most likely the one currently employed by Apple and is of great importance given the relatively high mismatch costs that are observed in our Newsvendor Model (over $1 billion). Apple would benefit by being able to produce fewer units for the beginning of the season and then deciding on any additional production once initial demand is observed. However, Apple’s suppliers would most likely only be willing to agree to this reactive model for a price premium. After all, they could face idle capacity (lower than expected demand) or the need to scramble for additional capacity (higher than expected demand). I will assume that Apple is able to negotiate a 20% premium fee – that is, every iPhone 12 Pro Max ordered once the selling season begins costs Apple $545.45 * 20% = $109.09 more to produce.
We can now use the Newsvendor Model to determine the optimal number of units to produce before the selling season starts. Apple should still produce a certain amount before the selling season because they are cheaper to produce than the ones during the selling season. Under this model, the overage cost (Co) remains the same ($545.45), but the underage cost (Cu) is now the premium Apple must pay ($109.09). Therefore, the critical ratio becomes 109.09 / (545.45 + 109.09) = 0.17, with a z-statistic of -0.97, which yields an optimal pre-season production quantity of 4,159,000 – 0.97 * 2,131,392 = 2,097,046. In other words, Apple should commit to producing 2,097,046 units before the season starts if a subsequent production wave is possible.
This approach significantly increases expected profits and reduces mismatch costs. As illustrated in Exhibit 5, the expected sales for the pre-season production wave are 1,908,175, the expected lost sales are 2,250,825 (which under this model are the expected second order production quantity during the selling season) and the expected leftover inventory is 188,871. Expected profit in this case = Maximum profit – (Co * expected leftover inventory) – (Cu * expected second order quantity) = $2,369,552,167.
Exhibit 5: Reactive Capacity Newsvendor Model
In-stock probability and potential customer reactions
But what if Apple was not able to adopt a Reactive Capacity approach? This is particularly relevant for firms that face demand uncertainty (who doesn’t in 2020?) along with an even shorter selling season or a more rigid supply chain. In this case, what other solutions exist to reduce mismatch costs and increase profit? To illustrate a potential solution, let us return to the iPhone 12 case and revisit the initial Newsvendor Model. Under this original model, and as shown in Exhibit 4, Apple faces a relatively high probability (45%) of stocking out during the season. To be clear, this does not imply that 45% of demand will not be met, but rather there is a 45% chance that not all of the demand will be met. From a marketing perspective, it is critical to understand what might happen if potential customers face a stockout during the season. With this objective in mind, we ran a survey in late September 2020 amongst a representative sample of n = 1,000 consumers aged 16 – 69 in the US to determine their appetite for the new iPhone 12 Pro Max and what they would do if faced with a stockout. In order to look at potential differences in reactions, we first segmented the consumers into 6 distinct clusters based on their proximity to the Apple brand (Exhibit 6): “Die-hard Apple Fans” (16%), “Active Apple Customers” (33%), “Passive Apple Customers” (25%), “Apple Abandoners” (6%), “Apple Haters” (11%) and “Apple Distant” (9%). For the purposes of brevity, we will focus our analysis on the following two segments: the “Die-hard Apple Fans” and the “Passive Apple customers”.
Exhibit 6: Six key consumer segments based on proximity to the Apple brand
Nearly all the Die-Hard Fans (96%) currently own and use an iPhone as their main mobile device, and 81% of them also used an iPhone as their last mobile device. Two-thirds of them (66%) have heard that Apple is launching the iPhone 12 and 67% declare they are likely to purchase it (22% very likely). Furthermore, they are most likely to purchase it quickly upon release and 94% are either willing to wait or accept another model (trade-up or trade-down) if faced with a stock-out. Their average acceptable wait time is 8 weeks. In other words, the point of entry for this segment is the Apple brand.
Contrast this to the Passive Apple Customers, of which only 44% currently own an iPhone (30% own a Samsung). Only 37% have heard of the new model and 41% would be open to purchasing it (only 2% very likely). Only 40% would be willing to wait if faced with a stock-out and many would simply switch brands. In other words, this segment contains customers that we need to be much more careful with: they are more likely to be purchasing an iPhone for their first time, and if there is a stockout, they are more likely to opt for a competing brand.
Exhibit 7: Correspondence Analysis based on opinion of the different brands
Exhibit 7 is a Correspondence Analysis that provides another way to visualize the differences between customer segments. The Die-Hard Fans and Active Customers are found in the smaller concentric circle around the Apple brand. They are much more likely to only consider this brand for their next smartphone purchase. The Passive Apple Customers segment is found in the next circle, where there are many brands vying for their attention. Finally, the more distant segments are found in the outer circle, with a clear preference for brands other than Apple.
Despite our original Newsvendor Model being helpful, it does not quite capture the differences inherent to these different customer segments, and in particular, it does not account for the potential reactions from these different customer segments in the event of a stockout. On the one hand, if a Die-hard Fan is faced with a stockout, that customer would be more likely to purchase another model, meaning Apple is still able to benefit from the margin on another product. On the other hand, if a Passive Customer is faced with a stockout, that customer is much more likely to switch to another brand, meaning that Apple foregoes the margin. To complicate things further, the Die-hard Fans are more likely to purchase towards the beginning of the selling season, thereby increasing the probability that the Passive Customers are the ones facing the stockout!
A yield management solution at the crossroads of Operations and Marketing
A yield management solution to this dilemma involves setting protection levels for the Passive Customers, similar to what is done for high-worth customers in the hotel and airlines industries. The logic is that we wish to maximize profits from these two customer segments given the limited supply of iPhones we committed to (4,400,355) and given what we know about their reactions to a stockout. We know the more loyal Die-hard Fans are likely to purchase another Apple product, while the Passive Customers are likely to switch to a competing brand. This implies that, if possible, Apple should serve the Passive Customer segment as much as possible with their limited supply of iPhone 12 Pro Max’s. However, since the Die-hard Fans are more likely to purchase towards the beginning of the selling season, Apple will need to protect some of its stock for the late-coming Passive Customers. The question is then: how many iPhones should Apple set aside for the Passive Apple Customers? Again, Apple faces another trade-off that can be solved by determining a critical ratio with and underage and an overage cost.
First, the consumer survey showed that 10% of consumers open to purchasing the iPhone 12 Pro Max belong to the Passive Customer segment. This translates to a mean demand of 415,900 units and a standard deviation of 213,139 for this segment. Moreover, let us assume that, if faced with a stockout, the Die-Hard Fans would opt for the iPhone 12 Max whose margin is estimated to be $410 (versus the initial $653.55 margin). This is akin to a Salvage Value for the Die-Hard Fan segment. On the one hand, if Apple under-protects (sets aside too few units for the Passive Customers), it would turn away Passive Customers and forego the margin on the iPhone 12 Max ($410). On the other hand, if Apple over-protects (sets aside too many units for the Passive Customers), it would end the season with excess stock. The overage cost is the cost of production of the iPhone 12 Max Pro ($545.45) minus the margin of the iPhone 12 Max ($410), because for each unsold iPhone 12 Max Pro, a Die-Hard Fan was turned away and purchased the iPhone 12 Max. Therefore, the true overage cost is $545.45 – $410 = $135.45, yielding a critical ratio of 410 / (135.45 + 410) = 0.75, a z-statistic of 0.68 and an optimal Passive Customer protection level (Q*) of 415,900 + 0.68 * 213,139 = 560,784. This means that of the 4,400,355 iPhone 12 Pro Max units that Apple should produce prior to the season, 560,784 of them should be set aside for the Passive Customer segment in the (rather likely) event of a stockout. In other words, 13% of the units (560,784 / 4,400,355) should be protected to serve 10% of the potential customers, given the risk associated with having this particular segment face a stockout.
Exhibit 8: Protection level analysis for the Passive Customer segment
In fact, returning to our initial Newsvendor Model, we can calculate the (maximum) potential profit increase this approach could have. The expected lost sales in our initial model is 735,070. Without the protection in place, these lost sales could be mostly Passive Customers resulting in the initial mismatch cost of 735,070 * $653.55 = $480,405,266. With the protection in place, we could increase the likelihood that these customers are Die-Hard Fans, which implies a much smaller foregone profit of $179,026,398 ($480,405,266 – $301,378,868, which is the iPhone 12 Max margin of 735,070 * $410). Thus, by protecting the right customer segment, Apple could increase profits by up to 18% (maximum recouped lost profits of $301,378,868) / expected profits of $1,705,117,754).
A portfolio management approach
We have seen that setting protection levels for the high-end iPhone 12 Max Pro amongst the Passive Apple Customer segment has the potential to reduce mismatch costs and increase profits, given the availability of other products in the portfolio for the more loyal Die-Hard Fans to trade-down to. We will now extend this logic to adopt a broader portfolio-management approach. To this end, let us look at what might happen for the base model, the iPhone 12, with an estimated retail price of $649.
The experts projected a mean demand of 9,159,375 and a standard deviation of 4,104,304 for the iPhone 12 base model, which provides a coefficient of variation (CV) of 0.45. There is slightly less variability than the iPhone 12 Pro Max, but still a significant amount. For this model, we are assuming a price of $649, a total cost per unit of $390, and a salvage value of $349 (base models are generally discounted the following year). This provides a critical ratio of 0.86, a z-statistic of 1.10 and an optimal production quantity of 13,655,306. The question is how would our two segments react to a stockout of the base model?
As illustrated, the Passive Apple Customer segment would be more likely to switch to another brand (resulting in a lost sale for Apple). On the other hand, the Die-Hard Fans, who use the Apple brand as an entry into their purchasing journey, would be much more likely to trade-up to the next model (the iPhone Max), which means an even higher contribution margin for Apple! This implies that Apple would want to protect some of their iPhone 12 stock for the Passive Customer segment (just like for the high-end model), but unlike for the other model, Apple is actually benefiting from a stockout affecting the Die-Hard Fan segment. The higher contribution margin for Apple results in a double incentive to protect as much stock as needed for the Passive Customer segment. So how much stock should be protected for the Passive Customer segment? The consumer survey showed that 12% of consumers open to purchasing the iPhone 12 belong to the Passive Customer segment. This translates to a mean demand of 1,099,125 units and a standard deviation of 492,516 for this segment. To maximize profits, Apple would want to ensure with a very high probability that the Passive Customer segment would not face a stockout. For example, should Apple wish to be 99.9% certain that the Passive Customer segment will not face a stockout of iPhone 12s, it would need to set aside approximately 2,650,000 units for this segment. This means setting aside 19% of the stock (2,650,000 / 13,655,306) for a segment that represents 12% of the potential customer base, given the high trade-off cost of missing out on a Passive Customer sale.
Extensions to the model and other considerations
This model could be enriched in several ways. One possibility is to determine the Customer Lifetime Value (CLV) for specific customer segments. This involves calculating the Net Present Value (NPV) of all future cash flows that a specific type of customer generates over the life of business with the firm minus the cost of acquisition. For example, the consumer survey shows that the iPhone is the biggest “gateway product” to the Apple brand, with 56% of Apple customers having purchased an iPhone as their first Apple product. It also shows that Apple benefits from a very strong customer retention rate. Therefore, a CLV model would consider these elements (along with others) and project the potential cash flow from future purchases of this type of customer.
An alternative model is also required for single-product portfolios, where trading-up or trading-down is not possible. In these cases, we would benefit from looking at the willingness to pay of our customer segments or the impact stockouts might have on their goodwill. This would lead us to protect supply for the segments with higher willingness to pay or those whose goodwill is more fragile.
Furthermore, Apple could estimate the willingness to pay of the Die-Hard Fans for locking in a guaranteed model of their choice via an early access scheme, and determine the trade-off with running the risk of stocking out amongst the Passive Customers. This multi-tiered tariffing approach would allow Apple to maximize the value capture amongst the segment with the highest willingness to pay. Perhaps as part of the Apple One offering?
The challenges to this approach
While this type of approach can reduce supply and demand mismatch costs and optimize profits, it does require overcoming several challenges. First, marketeers must have a solid understanding of their customer segments, their willingness to pay and their potential reactions to stockouts. Second, encouraging marketeers to disregard conventional wisdom of “rewarding loyal customers” and reserving supply for “light” or “peripheral” buyers is a tough pill to swallow for some. While this can be counter-intuitive, the Marketing literature has come to terms with the importance of “light buyers” over the past 10 years and with their role in growing sales[efn_note]This school of thought is most associated with the Ehrenberg-Bass Institute for Marketing Science.[/efn_note]. However, marketeers must also be careful in determining the impact on customer goodwill this approach could have. This is a challenge inherent to any price discrimination program based on implicit or explicit segmentation. Third, this approach requires marketeers to adopt a much more complex portfolio strategy that considers cannibalization and cross-selling potential. Fourth, this approach requires the sales force to be able to identify the specific segments at point of sale. Fifth, this approach requires supply chain coordination, and in particular the ability to align third party retailers on this approach. In many cases, this might involve modeling the data to account for supply chain profit maximization, rather than at a specific link in the value chain. And finally, this approach requires both Operations and Marketing to work more collaboratively.
As is the case every year, Apple (like many of us) must make production decisions in the face of uncertain demand. The level of uncertainty is exacerbated this year given the COVID-19 crisis, which most experts agree will adversely affect the sales of the upcoming iPhone 12, and in particular, the high-end models, that faces even steeper price elasticity of demand. This case illustrates that reactive capacity solves both stockout issues and significantly reduces, though does not eliminate, mismatch costs associated with matching supply and demand under highly uncertain conditions. However, when such a lean supply chain solution is not available, it is imperative that the Operations and Marketing functions work closely together to understand potential differences in customer reactions to cost / service trade-off decisions and put forth a customer-centric strategy to maximize profit.