Lately, I’ve been hearing a lot about how governments should regulate supply chains and enforce more transparency as a way of solving supply chain issues. But what if governments are not the best option when it comes to supply chain management and, contrary to their intentions, they inflict even more harm on the already fragile and volatile supply chains?
Great post Gad. Very helpful explanation for me. I wish I was the person who planned ahead and bought a few tests during periods of calm but I am not. Happy new year!
Nice post! My favorite part was the discussion of how rationing can cause people to inflate their demand. But I do have a minor complaint. When converting desired service levels to safety stock, you use standard normal tables. In this context, I don't know how I would estimate expected demand, let alone it's standard deviation. In addition, what suggests that demand (which is driven by factors such as timing of new variants, holiday travel, and social restrictions) is normally distributed? I'm scarred from years of MBAs assuming that everything is normally distributed, and think that stocking to achieve a specified service level would actually be quite difficult!
That's a fair point. When I assume that it's a normal distribution, I am clearly underestimating the amount of safety stock needed. But note that you are talking about two different things. You are talking about sequential correlation (i.e. trend and seasonality, for example) and variance. The data seems to indicate that there are significant trends, which means that at the beginning of a new "variant" I should be able to estimate the demand for these tests. On top of that, there will be variance. I think you are conflating the two. Once I accounted for these. I am sure the demand is not normal. But I am not sure it's not a good approximation. Note that this is very different than, say, internet traffic of viral tweets.
I agree that I was conflating two things. After accounting for seasonality, current case rates, etc, I could imagine that it's not so crazy to approximate demand for the next month with a normal distribution. But then I am uncertain about how to estimate the standard deviation (residual uncertainty) of this month's forecast: it's very different (likely less than?) the empirical standard deviation of demand across months. I guess this also relates to the question of how much lead time one needs when making forecasts: demand next month might reasonably be modeled by a normal distribution, whereas demand in 6 months is subject to the sort of macro-level uncertainties that might cause heavy tails.
Anyway, I know this isn't the main point of your post! I just think I'm especially sensitive when people quickly assume normality, after having taught a course where I was supposed to link "6 sigma quality" to "2 in a billion error rate."
I completely agree and really appreciate the depth of response.
Regarding the normal distribution. I have gone a full circle. I hated the rush to assume that everything is normal. But I find it easy to articulate things. Only to then break the assumption if there is evidence to the contrary.
Great insights on the supply chain origins of the Covid test kit shortages. Wishing you a happy, healthy and more obscure 2022!!
Great post Gad. Very helpful explanation for me. I wish I was the person who planned ahead and bought a few tests during periods of calm but I am not. Happy new year!
Nice post! My favorite part was the discussion of how rationing can cause people to inflate their demand. But I do have a minor complaint. When converting desired service levels to safety stock, you use standard normal tables. In this context, I don't know how I would estimate expected demand, let alone it's standard deviation. In addition, what suggests that demand (which is driven by factors such as timing of new variants, holiday travel, and social restrictions) is normally distributed? I'm scarred from years of MBAs assuming that everything is normally distributed, and think that stocking to achieve a specified service level would actually be quite difficult!
That's a fair point. When I assume that it's a normal distribution, I am clearly underestimating the amount of safety stock needed. But note that you are talking about two different things. You are talking about sequential correlation (i.e. trend and seasonality, for example) and variance. The data seems to indicate that there are significant trends, which means that at the beginning of a new "variant" I should be able to estimate the demand for these tests. On top of that, there will be variance. I think you are conflating the two. Once I accounted for these. I am sure the demand is not normal. But I am not sure it's not a good approximation. Note that this is very different than, say, internet traffic of viral tweets.
I agree that I was conflating two things. After accounting for seasonality, current case rates, etc, I could imagine that it's not so crazy to approximate demand for the next month with a normal distribution. But then I am uncertain about how to estimate the standard deviation (residual uncertainty) of this month's forecast: it's very different (likely less than?) the empirical standard deviation of demand across months. I guess this also relates to the question of how much lead time one needs when making forecasts: demand next month might reasonably be modeled by a normal distribution, whereas demand in 6 months is subject to the sort of macro-level uncertainties that might cause heavy tails.
Anyway, I know this isn't the main point of your post! I just think I'm especially sensitive when people quickly assume normality, after having taught a course where I was supposed to link "6 sigma quality" to "2 in a billion error rate."
I completely agree and really appreciate the depth of response.
Regarding the normal distribution. I have gone a full circle. I hated the rush to assume that everything is normal. But I find it easy to articulate things. Only to then break the assumption if there is evidence to the contrary.
Really interesting explanation! I like the use of Google Trends — a constantly fascinating pulse checker on anything at any time.