While you would get a bunch of knowledge by reading my blog posts, I strongly encourage you to actually attend conferences, interact with experts in the field, grow your network and get more knowledge. It's a different experience being present there. You see their expressions, the way they talk, the jokes they crack, the extra wisdom they share. There's also the sense of belonging.
You'd have heard that "data is the future"? Yesterday, I had been to the conference on Analytics at IIM-B, and through the conference, I found that analytics is at a very nascent stage. The panelists themselves, although experienced, were able to share only a rather generalized view of analytics; not just because it is a very vast field, but also because there's not enough of study done about it.
The introduction to analytics:
Professor Dinesh is a person students would love to have as a professor. His entire lecture was riddled with jokes and wit which kept the audience giggling away.
We were introduced to three different situations of decision making, via three stories:
1. Dosa King. The story of Mr.Narayanan, who did not have much data when he started off. One example was of how although machines for preparing dosas were eventually set up, they realized that the taste of the dosa changed, depending on every hour that the dosa batter was left uncooked. Decision making with not much data.
2. Tylenol having cyanide: When a couple of people died of cyanide poisoning after consuming Tylenol (the equivalent of Crocin), the CEO of Johnson & Johnson had just a few hours to take a decision of what message to send to the public, because of the innumerable number of capsules already in medical shops.
Apparently he consulted his people of whether there was any cyanide in their plant, and they said no. Later, they said yes. This was a time when companies didn't widely use ERP software. Prof. Dinesh emphasized that if ERP was used effectively, the information the CEO wanted, would have been available at his fingertips in no time. This was an example of decision making with incomplete data.
3. Captain Peter Burkill of British Airways: Close to the airport, when the passenger plane was still airborne, both engines of the plane failed. The pilot had 30 seconds to take a decision on what to do, and the professor told us, that he had at his disposal, around half a terabyte of flight data to sift through for taking a decision. The case of too much data.
When you enter the field of analytics, you'll notice that 80 to 90% of an analytics project, is spent on just cleaning the data. Making it useful for analysis.
[ Nav's note: So it's very important to find people who enjoy sifting through data and making sense of it. Most people would find it boring ]
So how do you define big data? : The non-profit Akshaya Patra tries to provide mid-day meals for children, keeping the cost per meal at 7.50 rupees. As part of optimizing costs, they also have to find the shortest paths for their transport vehicles, to minimize on fuel. Now the number of paths to be calculated here, is twenty factorial. Even if you had a computer that could do 1 million computations per second, it'd take around 76000 years to compute the data.
When you have a situation where our existing IT technology isn't enough to solve a problem, you have at your hands, a big data problem.
Types of analytics
Framework for decision making
This section went very fast. Not much info to share here, but these were the main points:
Conclusion:
For a company that uses analytics, it's important to build the right talent and build the right infrastructure.
As an example of how hospitals work, the first day the patient is admitted, is the most profitable for the hospital. The series of (many times un-necessary) tests the patient pays for and the profits made on treatment during the duration of stay. But when it comes to discharging the patient, it would be most economically viable for the hospital to get the patient out as fast as possible and to get the next patient admitted (this often becomes a problem, as doctors aren't available to sign the discharge slips). Data analytics can help by mapping this trend and maximizing profit for the hospital.
As competition increases, analytics becomes more important. For example, if you're setting up a food court, you'd need to predict how many people are going to eat there. One creative way of predicting it is by calculating the number of cars parked in the surrounding area (at offices) and the time at which people come out to eat and the availability of eateries nearby.
[ Nav's note: An excellent opportunity to attract customers with discount offers and a cleaner and less-noisy environment than the competition. Adding music would be another perk ]
Another example where predictive analytics is used, is in the sports industry. Get a load of this: While the sports industry is said to be at 300 billion dollars, the sports-betting industry, is said to be at 400 billion dollars worth!!!
You might also want to have a look at the Analytics Society of India.
The panel discussion
Following Prof.Dinesh's lecture, the panel shared their experience with the audience. Some key points in random:
Also happened to meet a software architect-turned-entrepreneur who very interestingly, has a startup which helps athletes find grounds in Bangalore. He, like me, was already up-to-date about the latest trends in databases and was all for the JavaScript stack of technologies (which btw, is the future; just like mobile devices are).
You'd have heard that "data is the future"? Yesterday, I had been to the conference on Analytics at IIM-B, and through the conference, I found that analytics is at a very nascent stage. The panelists themselves, although experienced, were able to share only a rather generalized view of analytics; not just because it is a very vast field, but also because there's not enough of study done about it.
The panel:
- Prof. Dinesh Kumar (moderator)
- Mr. Vivek Subramanyam, Co-founder & CEO of iCreate
- Mr. Shantanu, Co-Founder, Meshlabs
- Mr. Mahesh R, Co-Founder & CEO of NanoBI analytics
The introduction to analytics:
Professor Dinesh is a person students would love to have as a professor. His entire lecture was riddled with jokes and wit which kept the audience giggling away.
We were introduced to three different situations of decision making, via three stories:
1. Dosa King. The story of Mr.Narayanan, who did not have much data when he started off. One example was of how although machines for preparing dosas were eventually set up, they realized that the taste of the dosa changed, depending on every hour that the dosa batter was left uncooked. Decision making with not much data.
2. Tylenol having cyanide: When a couple of people died of cyanide poisoning after consuming Tylenol (the equivalent of Crocin), the CEO of Johnson & Johnson had just a few hours to take a decision of what message to send to the public, because of the innumerable number of capsules already in medical shops.
Apparently he consulted his people of whether there was any cyanide in their plant, and they said no. Later, they said yes. This was a time when companies didn't widely use ERP software. Prof. Dinesh emphasized that if ERP was used effectively, the information the CEO wanted, would have been available at his fingertips in no time. This was an example of decision making with incomplete data.
3. Captain Peter Burkill of British Airways: Close to the airport, when the passenger plane was still airborne, both engines of the plane failed. The pilot had 30 seconds to take a decision on what to do, and the professor told us, that he had at his disposal, around half a terabyte of flight data to sift through for taking a decision. The case of too much data.
When you enter the field of analytics, you'll notice that 80 to 90% of an analytics project, is spent on just cleaning the data. Making it useful for analysis.
[ Nav's note: So it's very important to find people who enjoy sifting through data and making sense of it. Most people would find it boring ]
So how do you define big data? : The non-profit Akshaya Patra tries to provide mid-day meals for children, keeping the cost per meal at 7.50 rupees. As part of optimizing costs, they also have to find the shortest paths for their transport vehicles, to minimize on fuel. Now the number of paths to be calculated here, is twenty factorial. Even if you had a computer that could do 1 million computations per second, it'd take around 76000 years to compute the data.
When you have a situation where our existing IT technology isn't enough to solve a problem, you have at your hands, a big data problem.
Types of analytics
- Descriptive analytics: This is where you want to visualize your data. Quick View and Tabulo help with this. Data synthesis and visualization. Gramener in Bangalore does this. One other example of descriptive analytics was the use of spot maps. During the cholera outbreak in 1854, there were around 400 theories of how the cholera was spread. Most people thought it spread through air. But John Snow. A doctor, mapped the pattern of the outbreak and correctly located a water pump which seemed to be the source of it. (anecdotally, when he approached the authorities, they asked if that was the case, then why weren't any of the people in the nearby brewery falling sick of cholera, and John found that it was because the workers weren't drinking water from the pump. They were drinking only beer from the brewery :-) )
- Predictive analytics: When you want to use data and trends to predict consumption patterns in the future for example. Or like how Google and the CDC predict diseases based on data they receive and capture from people. Milk consumption is also an example where people need data to predict how much of milk would be consumed before the expiry date and how much of it they'd have to sell to restaurants who are waiting for a discount a few hours before the expiry. Also about how much of the remaining milk beyond expiry date, would go for preparing paneer, cheese and to sweet shops (apparently many sweet shops use milk that's crossed the expiry date).
- Prescriptive analytics: This is the toughest of them all. To be able to prescribe a solution, based on the data you have. Of course, this also involves using predictive analytics.
Framework for decision making
This section went very fast. Not much info to share here, but these were the main points:
- Problem identification: As an example of what Target retail chain did: What would you do if you wanted to find out which of your store customers were pregnant? If you identify these people, you could advertise and sell a slew of products for baby care.
- Ask the right questions: Asking, requires good domain knowledge.
- Collection of relevant data: The data has to be in a certain format. One of the reasons companies aren't able to make good use of the data they have in their ERP systems, is because is not built on analytics. Such data is useless to them. The solution to this, is to redesign the ERP systems, with analytics at its core.
- Pre-processing: Decide how you're going to use the data.
- Model building: Create data models that will be useful for analysis.
Conclusion:
For a company that uses analytics, it's important to build the right talent and build the right infrastructure.
As an example of how hospitals work, the first day the patient is admitted, is the most profitable for the hospital. The series of (many times un-necessary) tests the patient pays for and the profits made on treatment during the duration of stay. But when it comes to discharging the patient, it would be most economically viable for the hospital to get the patient out as fast as possible and to get the next patient admitted (this often becomes a problem, as doctors aren't available to sign the discharge slips). Data analytics can help by mapping this trend and maximizing profit for the hospital.
As competition increases, analytics becomes more important. For example, if you're setting up a food court, you'd need to predict how many people are going to eat there. One creative way of predicting it is by calculating the number of cars parked in the surrounding area (at offices) and the time at which people come out to eat and the availability of eateries nearby.
[ Nav's note: An excellent opportunity to attract customers with discount offers and a cleaner and less-noisy environment than the competition. Adding music would be another perk ]
Another example where predictive analytics is used, is in the sports industry. Get a load of this: While the sports industry is said to be at 300 billion dollars, the sports-betting industry, is said to be at 400 billion dollars worth!!!
You might also want to have a look at the Analytics Society of India.
The panel discussion
Following Prof.Dinesh's lecture, the panel shared their experience with the audience. Some key points in random:
- The world is going SMAC: Social Mobile Analytics Cloud
- Machine learning, Bayesian calculations, Hadoop and DevOps are the 'in-thing' for analytics right now.
- Opportunities in analytics: Anything that generates a lot of data as part of its functioning, would be good for analytics.
- Analytics for a small business: One of the challenges faced is in how to explain analytics to business people, in Hindi or Kannada?
- Learning analytics: Apparently Bangalore University offers courses in analytics. There are five colleges in Bangalore offering a course. Word of caution though, is that no matter what tools you learn through these courses, what really matters is the experience of actually working with data, statistical data, models and open data movement.
- Surviving and succeeding: Ask the right questions, solve the right problems.
- Advice: Don't venture into a field in analytics where people are already doing a lot, and you don't have anything different to offer or you don't know how to do it better.
- How analytics helps: Many times, business people tend to rely on intuition and despise analytics and data. But analytics helps better than intuition: Like for example, a retail chain located at a place where hurricanes were frequent, wanted to know what people tend to purchase just before a hurricane. Employees guessed torches, raincoats, beer etc. But when they looked at the data, they saw that what sold the most just before and during a hurricane, were strawberry pop-tarts. Amazing, isn't it? Would intuition ever have given you that info?
- Three most important things for a startup: 1. Talent 2. Talent 3. Talent. All other things can be purchased with money. You have to find the right talent for analytics, and the supply is scarce. Because you have to know what problem to solve, and who will solve it for you. Find people who are at least good at linear programming problems. They'll have to scale their skill more for bigger data problems.
- The go-to-market is more important than decisions on what technology to use.
Also happened to meet a software architect-turned-entrepreneur who very interestingly, has a startup which helps athletes find grounds in Bangalore. He, like me, was already up-to-date about the latest trends in databases and was all for the JavaScript stack of technologies (which btw, is the future; just like mobile devices are).
No comments:
Post a Comment