Analytics - a tale of two movies

The first movie is of course 'Moneyball' and the second is 'Trouble with the curve'. Both movies are about recruitment of players in baseball. But both movies take a completely different route for this. This article seeks to point out the extreme difference and I seek views of all of you readers.

In Moneyball, Billy Beane (Brad Pitt) is a the General Manager of a poor Oakland Athletics team. He develops a statistical model for player selection with a Yale economics graduate. The focus of selection is numbers and not intuition and experience of the scounts. Gus (Clint Eastwood) is an aging scout for Atlanta Braves in Trouble with the Curve. He has a very weak eyesight. He rejects a star prospect based on the sound of the bat hitting the ball. This is in spite of his amazing statistics. In the end Gus is proved right.

The debate is between using analytics and individual insights for decision making. In the movies both seem to work. In real life a lot of importance is given to the brilliance of an individual manager who seems to be able to make better decisions. Books by Kahneman and Taleb negate the role of intuition and seem to suggest a role for random luck. They support the decision making with numbers view.

What would you do? How much importance would you give to decisions taken on the basis of analytics versus the gut based decisions of an experienced manager?

The Mumbai dabbawallah myth

The Mumbai dabbawallahs seem to be a stuff of fantasy. Every supply chain conference and training seems to be incomplete without a mention of their supposedly fantastic performance. Their par six sigma performance is a dream achievement for every supply chain professional. They have seemed to have achieved an iconic level of excellence with 'uneducated' operators and no technology.

Let me make my position clear. Yes, the dabbawallahs have surely achieved operations excellence. But, they have been able to do so because of some benign business conditions. Most other business are not blessed with the simple and stable pattern existing in the dabbawallahs business. Thus the dabbawallahs are able to excel in Operations without strategy, technology or any type of documented processes.

The same dabbawallah visits the same houses every day at the same fixed time. He collects the lunch box (dabba) and relays it through the same train to the same office at the exact same address. The dimensions of the products are also more or less the same. This 'same' based operation continues every single day. The only variation is that a customer may not have a lunch to deliver on some days. No business enjoys such luxury of stability.

The arrival of customers, the time taken to serve the customers and the type of service or products demanded by the customers all vary in 'normal' businesses. Without variation, or with low variation, business is easier to plan. With stable conditions a business can easily avoid all excess inventory, excess resources and product stock outs.

Let us assume that a beauty salon has exactly one customer coming every ten minutes for exactly the same style of hair treatment and that the treatment takes exactly 30 minutes. In this case the salon can easily employ a staff of 3 people and deliver excellent customer service with very high level of resource utilisation - leading to six sigma operations with high profits. 

An average FMCG firm would have hundreds of SKUs for customers to randomly choose from. The distribution network with multiple independent profit making entities would also have their own buying pattern and behaviour. And, of course, the customers have no compulsion to order in a stable pattern. Thus these businesses are considerably more difficult. The dabbawallahs do the same thing every day. Thus they can easily work from memory. Since businesses have a high number of variables, they need the support of technology to make decisions.

The dabbalwallahs would be very praiseworthy if they were picking up lunch from different houses every day as per some random pattern and also if the delivery addresses were different. To add to the fun it would be wonderful to have the trains running at varying time tables. What the dabbawallahs have achieved is good but nothing incredible. Their model of operation may be a benchmark for abnormal conditions of stability, but is absolutely useless in normal business conditions.

Stampede - an Operations failure

Operations failures that cause a loss of lives and are avoidable without much difficulty are very painful. Here lives are at stake and not merely excess stocks or breakdowns or some other performance metrics. Take the lastest stampede in Madhya Pradesh: http://bit.ly/1p7f9yf. 

As of now 10 people have been reportedly dead.

The number of temple stampede deaths in India are unprecedented. Take this data:
October 2013: Madhya Pradesh, 115 dead
February 2013: Allahabad, 36 dead
November 2011: Haridwar, 16 dead
January 2011: Sabrimala, 102 dead
March 2010: Kunda, 71 dead
August 2008: Himachal Pradesh, 162 dead

The list can go on. These are all stampede events with relation to some religious activity. The data has been sourced from wikipedia.

All places of religious importance have a significantly higher than average rate of human density. This makes the system highly susceptible to failures. An electric fire at home may at most cause a fire and in a worst case death of a few people. In a dense area, an exactly similar event can lead to very grave outcome. A small spark may ignite a chain of events leading to an enormous tragedy. The first point being made here is that tragedies of similar nature in domestic and public locations have a much greater impact in public places.

At homes, or even in offices the people in the system are more or less constant. All those people generally have a long term stake in the well being of that place. Thus, in a majority of cases, minimum safety standards are ensured. In public places, in spite of the religious connection, the situation is very different. Pilgrims may not always be as careful in ensuring a risk minimizing behaviour. That there are so many other pilgrims may create a condition where no individual takes the onus.

A third point is about the concept of risk probability. Say an office has 100 people and each of them is capable of causing a catastrophe with a probability of 0.01%. The chance the there will be a catastrophe in the office is still lower than 1%. In a gathering of 10,000 people the chance of a similar tragedy would be more than 60%. Simple statistical calculations point to the fact that the probabilities of tragedies caused by human failure are significantly higher in public places.

Some religious places have a spurt of pilgrims on some specific days of a year only. This spurt can increase the number of pilgrims by more than a thousand times. With pilgrims the entourage also includes shopkeepers, beggars and similar people. Most places are not designed to handle this sudden increase in the number of people. Thus the system is in a shock and very susceptible to failures. 

There can be more reasons. But, the point remains, that in places with high human density, like the places of worship, the probability and the impact of a tragedy are both very high. It thus makes sense that temple management and the staff of such public places give more than average attention to safety issues. There has to be some sort of paranoia for tragedy prevention.

One way to create such systems is of course by the rule of law. Temples and other religious places could be forced to have a set of safety based policies and the temple management could be made directly liable for any deviations or failures. The fear of punitive action might force attention on this ignored issue. However, it is common knowledge that forced compliance based system rarely result in intended outcomes. People would always find ways to subvert the rules and thus there would be no major improvement in the behavior.

I am not clear to the exact nature of the solution. But, I am clear that every place of religious gathering would have to evolve their own ways of safety procedures. Temple management and staff must be educated and thus motivated to create safety policies. A temple would never have a deity that is unclean. Something should be done so that unsafe practices are a similar anathema. Every human is an incarnation of the divine. Thus, ensuring safety of humans is probably as big as any other offering made to the almighty. 

Imagining the future of Retailing

Sunil Chopra and Aditya Jain had this article in the Economic times: http://articles.economictimes.indiatimes.com/2014-08-13/news/52768197_1_online-retailer-blue-nile-zales

They talk about merging the concepts of online and offline retailing with overall benefits for everyone, the consumers, retailers and etailers. As a concept no one can doubt on such win win scenarios. But, as a matter of practice, the article says nothing, and uses completely wrong examples to prove their point.

There is a comparisons of the costs of an online diamond firm Blue Nile with an offline firm Zales. Of course their cost structures are going to be different. For that matter the cost structures of Toyota and Mercedes Benz would be different, that of Walmart and Macy's would be different. To compare and then try and derive some relationships is perverse academics.

Online and offline are very different businesses. So an online business is bound to have higher revenue per dollar invested in infrastructure. But, the aim of business is surely not to maximise the revenue per infrastructure dollar. For all our enthusiasm for etailing, Amazon is still losing money.

In the Costco and Amazon example, Amazon's high transportation costs cannot be used to justify the better performance of Costco. The low variety model of Costco is great, but the profit margins of Wal-Mart are significantly higher than Costco. Here it seems that the authors have been selective in giving their pre-meditated examples so that they can prove a point.

Online and offline are very different models. As of now, 'touch' is an important part of shopping for many items. Who knows, the evolved customer may not need need to feel. Also, this concept may vary with the category of material. Last mile is surely a major cost for online models, but collaboration with offline for this may not be the only way. Online businesses may create mini warehouses inside cities. Similarly, offline businesses may not be limited to being a demonstration center for products that are ultimately bough online. They will have their own needs and create their own purpose. There also might be some form of collaboration, but not necessarily in the model that Chopra has proposed.

When I read an article by stalwarts like Chopra I expect articles with more imagination and not such mundane discussion that is part of every run of mill SCM seminar. But I guess the pressure to get something in print seems to be taking a toll of even the best among us.

Using Operations to improve customer service

Way back in 1969, Wickham Skinner had written an article in the Harvard Business Review saying that if designed, manufacturing can be a potent strategic weapon for firms. The key was in making manufacturing decisions, or in other words operations decisions, a part of corporate strategy making. Almost 45 years after the article was written, I was sad to see that a firm like Shoppers Stop had completely ignored basic operations principles in ground level operations. So, while operations parameters surely do not seem to be a part of strategy, they are not even a part of routine operations.

July 2nd, 2014 was the first day of supposedly a special sale for their First Citizen members of Shoppers Stop. Obviously they must have expected a higher number of customers as the sale proposition mentioned in the advertising was fabulous. As expected, the customers were surely higher and the number increased towards the evening. There were no major problems in the store, the culprit were the checkout counters.

When the number of customers are expected to be higher, the checkout, which would usually be the constraint, should have been designed to be as fast as possible. The number of customers would be more, and the number of SKUs per customer would also he higher. This one single factor could be a huge factor in realising the sale, as with larger queues, customers could just walk off.

And this was precisely what was happening. It was not a hoard of customers walking off, but a few surely were. The checkout time at around 5:30 pm was more than 25 minutes and it was getting worse. There were loud, irritating discussions in the queue about this.

The long checkout time had many reasons:

1. The discounts were being manually entered. The problem was not just the entry, but since it was manual, the customers had lots of queries. Many were checking the calculations with the counter staff to confirm the discount amount. It is normal for customers to trust manual calculations lesser than automated ones.
2. There was a confusion with the fine print of discounts. There was a 5% additional discount for something and as per the new policy the discount would not be given as cash, but would be added as points. Customers were inquiring about the timing of this policy change, the reason and lots of other things that were in no way helping increase the speed of checkout. 
3. Because the discounts were not cash (but points) a few customers were spending a lot of time at the counter debating if it was worth buying the material. Thus they were using up valuable time at the bottleneck process. 
4. The concept of cash back on credit cards is clear. But, many customers did not know about this and again had loads of queries. 
5. A few items had the bar codes missing. Though this was rare, even one item with a missing bar code could take up around 5 - 10 minutes to locate and clear.
6. The invoice was another classic document. Though it was correct, it was cryptic. You needed a good head in mathematics to match the discounts mentioned to the actual value. Customers would obviously ask questions on this.

There were many other minor reasons. Like one customer debated about the VAT. Another did not trust the manually written product code and insisted to see a similar product. One genius customer was calling someone from her mobile phone to decide on the purchase.

My point is that none of these reasons are radically new. Shoppers Stop has been in this business for around 20 years now. They should have seen through this. Designing for quick check outs is no rocket science. For some reason however businesses routinely ignore the operational issues that arise out of some marketing actions and do not take adequate actions to manage performance. I could hear a lot of talk among sales people in the Shoppers Stop store about their huge targets during the sale period. There was no talk about the queue at the check out counters. Yes, the counter staff were very courteous and patient, almost all the counters were open, but the clear lack of attention to the customer flow through was clear. Come on Shoppers Stop, you can surely do better!!

The other side of large retail in India

I am amazed by the vociferous opposition in the media to the AAP's proposal to ban FDI in multibrand retailing in Delhi (http://timesofindia.indiatimes.com/business/india-business/AAPs-decision-on-FDI-in-retail-irresponsible-Commerce-minister/articleshow/28785884.cms). Newspapers are full of articles of potential losses of billions and also of more than 10 million job loss (http://timesofindia.indiatimes.com/business/india-business/Blocking-retail-FDI-to-rob-India-of-1-crore-jobs/articleshow/28807283.cms?). There is also a lot of talk of how big box retail will improve everything wrong in our country. The opposition may not be completely valid.

The reason that we have high wastage in agriculture is not the lack of systems. It is in fact the existing infrastructure bottlenecks. The cold chains can't operate profitably because of the lack of reliable electricity supply. The truck speeds are slow because of pathetic road conditions. There are high prices and variable quality in the produce because of the fragmented nature of the farm ownerships. All these are crucial issues which if solved would automatically impact the efficiencies of the retail business. The FDI is for direct retailing business. It will have no impact on the infrastructure. As it is the FDI route for cold chains has been open, without many takers, for quite some time now.

There is a difference between the type of jobs created by large retail and the small mom and pop stores. With Mom and pop stores there are entrepreneurs who employ one to four low wage helpers. They are served by entrepreneur distributors and entrepreneur vehicle owners (sometimes with bicycles). Thus the job quality and the level of income is higher for a majority of people. In organised retail, the salary structures and job quality are highly skewed. The super rich but small number of top management would have an extremely high salary as compared to an army of shop floor drones who would barely be able to survive. The salaries of Target and Walmart CEOs are more than 500 times the average salary of their employees (http://money.cnn.com/magazines/fortune/fortune500/2012/ceo-pay-ratios/).  

Such type of income distribution is never good for any economy. High growth and increase of consumption can only happen in markets where the income structures are relatively level. The type of jobs in multi brand retail are thus not likely to create a favourable impact on the economy. Plus of course there is a possibility that the large retail may cause a job loss in the existing set up.

Proponents of large retail in India are demanding proof to show that if the existing large stores have caused any job loss as of now. However they avoid questions on the issue of the supposed benefits of large retail. Have large retail in India benefited farmers in any way? Have they led a significant price decrease for customers?

Efficiency does not always mean low price. Low prices of cigarettes could possibly cause havoc in the economy. Efficiency calculations with a smaller set in considerations are usually wrong and cause something what is called as 'local optimum'. For an alcoholic (or a cigarette smoker) low prices of his vice would be a great proposition. So, if we limit are efficiency calculations to the immediate satisfaction of the person, the prices must be as low as possible. However, if we extend the calculation to his lifetime, higher prices that lead to lower consumption could be better. (Am assuming that consumption is related to price and in cases of alcohol and cigarettes this could be wrong, but the the point is the nature of example and not the exact relationship).

Thus the low buying prices for the consumer could surely create a short term advantage for her. However in the long term, when the focus of efficiency is shifted from the buying transaction to the entire country, the effects of large retail could be very different. The documentary 'Wal-Mart: The high cost of low price' clearly shows that the public health care benefits that the Wal-Mart employees use are a huge drain on the tax revenue. Thus the low buying prices could be nullified by higher taxes.

I am of course assuming here that the large retailers are honest citizens. There have been multiple cases of policy manipulation (though lobbying) and corruption against most large retailers. Being super large they have a clout to conduct such activities unscathed, but this could be the issue of another article. 

It is disturbing that the media savy and educated section of the Indian population are not even willing to consider the other side of the debate. News papers and TV media cannot be relied on so much to frame and base our opinions on such important issues. They have their own agendas. My appeal is for people to think on their own, to read alternate views, to go beyond 'popular' opinion and then  decide. A society that is led by opinions manufactured by mainstream media can be a very dangerous place to live in.