Do you really need to fear AI?


As I open my Facebook profile, I see offers by Amazon on the side with the title, “Here are some black dresses especially recommended for you”. Wait what? How did Facebook know I was looking for a black dress? And how does it know what styles I like? How does Saavn (a music streaming service) know what songs I’d like to hear? How does Google Now know what I want even though I muttered something to it groggily? The answer is simple, and it’s everywhere! From the Fitbit on your wrist to the social networking sites you surf to the customer service offered by major companies, Artificial Intelligence has taken our lives by storm.
Machines outperforming humans is a tale as old as the Industrial Revolution. But as this process takes place in the exponentially evolving Information Age, many are beginning to question if human workers will be necessary at all. They believe that Artificial Intelligence has ushered in an era of a fourth Industrial Revolution with technologies like 3D printing, robotics and nanotechnology advancing at a faster pace than ever before. However, recent developments have shown that there is one significant difference between the two. Unlike the Industrial revolution phase, where machines were used to displace muscle power for mechanical labor, the algorithms used in the machines today are starting to pick up cognitive tasks. In a limited sense, they’re starting to think like people. They’re starting to encroach on that fundamental capability that sets us apart as a species – the ability to think.
There is constant debate as to whether artificial intelligence will replace human workers in the future. Many believe that it has the power to spread to every employment sector in the economy, which would not leave any safety net for the people. It is going to make virtually every industry, less labor intensive. On the contrary, there are also many people who opine that Artificial Intelligence is well, artificial and that it should be treated as such. They believe that however advanced the technology and artificial intelligence becomes, it would still be under the control of the human operating it.


The antagonistic view
“We are being afflicted with a new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come—namely, technological unemployment.”
This prediction was made in 1930 by world renowned economist, J.M Keynes in his essay titled ‘Economic possibilities for our grandchildren’. Had he predicted at that time what we are experiencing now? The phenomenon of fast-paced technological advancement displacing human jobs in different sectors of the economy. Many believe so, including famous theoretical physicist Stephen hawking who mentioned in his interview that the rise in artificial intelligence could potentially wipe out mankind. He is not alone in thinking so, and is joined by Elon Musk and Bill Gates who also believe that if this advancement is not brought about in a controlled manner, it could negatively impact the human race.
It is believed that with rapidly moving innovations and technological improvements, artificial intelligence could reach a phase where it is able to outsmart its creators. In this case, the Luddite fallacy could eventually have an expiration date. We have seen rising replacement of human labor with machines in the manufacturing sector, but what will happen when machines replace humans in the services sector, the one sector they call their home? There is no doubt that artificial intelligence will also lead to creation of jobs, but these jobs are majorly technologically advanced high end jobs. Middle skill jobs (secretaries, librarians, customer service and call centers etc.) are most likely to be fully replaced by innovative machines with cognitive computing capabilities. This has developed a fear among workers which has made them hesitant in fully embracing and adopting artificial intelligence.


The optimistic view
While there are some people against the full scale adoption of artificial intelligence, there are also many proponents of the technology. Many innovationists and developers are of the opinion that although artificial intelligence has the potential to become cleverer than their creators or replace them fully, we are nowhere close to that period. They put forth several points in favor of artificial intelligence.
The advancement can be utilized as a substitute for human professions that are dirty, dangerous or dull. Machines can be used to replace routine information processing tasks or repetitive tasks in a factory line which a human worker may find dull or boring. As a result, people could work only because they want to, and not because they need to. Humans have now decided that they were meant to be ballerinas, full-time musicians, athletes, fashion designers, yoga masters, fan-fiction authors, and folks with one-of-a kind titles on their business cards. With the help of our machines, we could take up these roles.
Since the technologies develop individually, they will hasten the development of other segments (for example, artificial intelligence might program 3D printers to create the next generation of robots, which in turn will build even better 3D printers). It’s what has recently been identified as the Law of Accelerating Returns: Everything is getting faster—faster
Furthermore (if not somewhat ironically), improving technologies can create novel opportunities by lowering the bar to positions that previously required years of training/experience; people without medical degrees might be able to handle preliminary emergency room diagnoses with the aid of an AI-enabled device, for example.
The machines are not replacing human jobs. In most cases, they are performing tasks that humans cannot do or couldn’t even think of. We are doing, and are sometimes paid for getting involved in, a number of new activities that would have stunned and amazed the farmers of 1850. These new accomplishments are not merely chores that had been difficult before. Rather they are really dreams that are created mainly by the capabilities of the machines that can do them. They are jobs the machines make up and for this reason, we are in need of a Watson or a Cleverbot or a MetaFore.

Co – existence
Society has always adjusted/adapted to technological changes in the past. This time is no different. Time and again it has been proven that human-computer teams beat all solely human or solely computer competitors. It’s time to work not against but alongside the machines.
Every successful bit of automation generates new occupations—occupations we would not have fantasized about without the driving of the automation. We all know how great it is when technology works, and just how frustrating it is when it doesn’t. Even large reputed technology companies haven’t completely eliminated their human customer support teams, because when something goes wrong, it is usually a human who needs to repair it.
Generally, there will always exist a need for on-site, human expertise when we deal with machines. Robots will have glitches, need revisions and require new parts. As we rely increasingly more on mechanized systems and automation, we will require more people with specialized skills to revise, update, fix and take care of these systems and hardware. This will lead to new jobs being created in science, technology, engineering and mathematics (STEM) fields like nanotechnology and robotics.
We are at an interesting transition point where we are moving from using tools to perform work that we were unable to do, to using them as active partners in our decision making processes. For example, in a pharmacy setting, machines could automate the process of filling up prescriptions while the pharmacist can concentrate on more interesting work such as advising the patient. This combination of AI and HI has the potential to dawn upon us an era of unprecedented creativity, innovation and intelligence in every decision making process. Human intelligence collaborating with artificial intelligence will help us in leveraging the best of both worlds. Where machines can perform repetitive, dangerous or delicate tasks for us, we can help machines with soft skills such as asking questions, planning, creative problem solving, and empathy where they lag behind.


What do we, at MaFoi feel about AI?
To find out how soon MaFoi employees feel they would be able to witness advanced artificial intelligence becoming a part of day-to-day life, we conducted a survey. The results are as follows:


The results clearly show that we are absolutely ready for AI to take over our daily functions (except major life-altering ones). In the areas of finance, marketing and other business functions, we are already there.
Just like any other technological advancement, AI has also been met with uncertainty and mistrust, but there can be no doubt regarding the potential of AI together with HI. The power of human creativity working with artificial intelligence can be applied to any field, be it cooking, transportation, athletics or business. The results will always be better than a fully manual or fully automated approach. Instead of thinking about ways in which artificial intelligence can replace our jobs, we should think about the ways in which we can work together with artificial intelligence to make this world a much better and much advanced place to live in.
At MaFoi, we embrace every novel technology with open arms. We constantly look for ways to keep ourselves updated with the latest technology, and therefore have made significant advances in the field of artificial intelligence with our cognitive computing expert, MetaFore. MetaFore uses AI, pattern recognition and Natural Language Processing to find hidden patterns and insights in different forms of data.
It’s time to buckle up and hop on the Artificial Intelligence Bandwagon. We already have our Jarvis (Read: MetaFore), what are you waiting for?

Data Science Driven Recruitment

Availability of data and information in recent times is changing the recruitment process massively, starting from sourcing to authenticating candidate credentials to final placements. When it comes to hiring talent to full fill key positions in various organizations from entry level to top level, it is very important to select the right candidate. Particularly for top level positions (Directors, Group Managers, Chief Executives) extra precaution need to be taken as these positions carry a lot of responsibility-they are partners in profit, hold significant amount of shares of the company and also act as torch bearers for the company. At the same time for mid-level or entry level positions, when an employee leaves the organization or it is discovered that the employee is not fit for the role that he was hired for, by that time company would have spent a bomb on training, compensation and reimbursements including the cost for back fill the position. The cost of a ‘bad hire’ to an organization is five times the bad hire’s annual salary and hence companies should focus on hiring the right talent to mitigate business risks (Times of India, May 25, 2015).

There are three major challenges that the recruitment industry is currently facing:

  • Pressure and urgency to fill up open positions in different companies, having a variety   of open positions
  • Higher number of applications every year, fresher’s and competitors applying for same positions
  • Lack of proper reference checks and background verification

In order to restrict the wasteful expenses and non-value added activities and minimize business risk from recruitment process, can Analytics, Data Science, Machine Learning and Artificial Intelligence be used to make a recruiter a Smart Recruiter?


Let’s look at how data science can help a recruiter to excel the science driven recruitment process.

  • Sourcing and Matching:

Access to digital profiles now days, is not a challenge, thanks to LinkedIn and other job sites where a pool of profiles reside. For a recruiter how to search the best fit from the pool is a real challenge. Can big data come to the rescue? Yes, unstructured data analysis along with application of natural language processing frameworks can establish similarity between a desired profile and the profiles available. One of the leading companies in recruitment space, CIEL HR recently adopted a product developed by Ma Foi Analytics for resume matching and subsequent scoring. The Resume Relevance Algorithm developed by Ma Foi Analytics parses the CVs to extract its context – skills & experience. Statistical procedures are applied on the derived context to rank the CVs against the context of the Project/ Assignment of the User Company and show top matches to the user. The product is believed to increase the fill rate from current industry benchmark of 20% to somewhere near 60%. Therefore the challenge now can be addressed. The product not only matches the best profiles from the pool, it also suggests top N matches based on key search criteria. As a result a significant amount of recruiter’s time that is spent on searching and matching is reduced and recruiter can spend that time on some other meaningful activity.

  • Evaluations:

As the number of applications have been increasing at an exponential rate over the years, screening of candidates becomes very important. The first level of evaluation starts when a recruiter compares different profiles that are being selected by the algorithm. The second level of evaluation is required by interacting with the candidate. The big data analytics approach can solve the second level of evaluation as well. The following process explains the scientific way of evaluation.


The fake interviews e where there is no authentication that the person being interviewed over telephonic interview is the same person being hired given the above process, when the telephonic interview is recorded the same should be conveyed to the candidate before taking interview. This can reduce such instances drastically. The current industry practice of recruiter’s interaction with the candidate is either captured in notebooks or not captured at all. The entire conversation can be captured and converted to text, so that the same can be stored in structured data bases. In future the consistency of statements made by the candidates can be tracked in order to avoid debates and discussions. Based on the conversation sometimes the recruiter needs to take a call on whom to select, at that time the insights would be readily available in a summarized form, so that the recruiter does not have to recall what happened over the previous discussion.

  • Background verification:

The pressure and urgency to fill up the open positions and higher number of applications, somehow gives room for the candidates to ‘game the system’. Though there are various background verification agencies that do it for big companies, there is a possibility of creating an alternative way to do this. Check prospective candidate’s LinkedIn profile to cross verify the information, send emailers to colleagues from past organization. Pull the social media posts, like twitter and Facebook posts and mine the text data to understand the candidate better.

In the recruitment space, if the three problems as mentioned above can be controlled and improved by even a certain percentage adopting big data analytics and machine learning, a lot of money can be saved. Indian firms are estimated to have lost at least Rs 2,460 crore in bad hiring in 2012. The figures were Rs 2,270 crore in 2011 and Rs 2,120 crore in 2010, says a study on Bad Hiring Activity in India by recruitment tendering platform ( The figure for 2013 is still date not available, but guess I can say it would be around 2000 crore plus. Introduction of big data, data science and machine learning methods and approaches as discussed above are only directional, there are a lot more possibilities given the data and features that are captured.


Indian Predictable League

Cricket is a funny game, especially its limited overs version, you cannot actually predict what a game of cricket is hiding for you until you watch it full. Indian Premier League is a perfect cocktail for this gentlemen’s game. It has all the masala (x-factor) which you would desire to watch in a cricket match. Individual performances like Chris Gayle’s 66 ball 175 knock in IPL 2013 or Adam Zampa’s 6 wickets for 19 Runs in IPL 2016 makes it even tougher to predict.

There are many factors which can affect an outcome of a cricket match like playing eleven’s current form, pitch conditions, venue, type of opposition, toss, past performances, team composition to name a few.

Having said that, what I am about to say might seem quite the opposite, what I have here is a list of factors which I have used to predict IPL 2016 matches.


Objective: Predicting an outcome of a cricket match for all the IPL Matches

Factors that affect an outcome of a cricket match:

  1. The Playing Eleven

First of all, I have allotted a batting score and a bowling score to each and every player. The score assigned is based on the past performance of the players in IPL. For a particular match, total team batting score and total team bowling score is calculated as a summation of player scores. Team Batting Score and Team Bowling Score are the two most important factors for this prediction approach.playing eleven

  1. Ground Type

Each cricket ground brings up some twists and turns. Based on the past matches in a particular cricket ground, each ground has been assigned a tag of either Batting Friendly or Bowling Friendly. As the name suggests, a Batting Friendly ground gives some extra points to the Team Batting Score whereas a Bowling Friendly ground earns extra points for the Team Bowling Score. Apart from that, Ground dimensions are also taken into account. All the grounds are divided into 3 categories: Small Ground Dimensions, Medium Ground Dimensions and Large Ground Dimensions. For Small and Medium grounds extra points are given to players who fall in Powerhitter category (Strike Rate >130 and >50% Runs in Boundaries). For Large ground dimensions, spinners are given extra points.Ground Type

  1. Type of Players

It is important to check the team composition to get some meaningful insights. Four categories are taken into account for the analysis: Batsman, Bowler, All Rounder, and Powerhitter (Strike Rate >130 and >50% Runs in Boundaries). Each player can fall into more than one category. Based on the team composition (Number of Batsmen, Bowlers, All Rounders and Powerhitters) extra points are added to Team Batting Score and Team Bowling Score.
Player Types

  1. Home Advantage

Some teams have an added advantage while playing at their home ground. Some teams understand their home conditions better than their opponents. This factor takes into account Number of Wins in the last 5 Home matches. Extra points are given if a team has won more than 2 matches out of the last 5 Home matches. Some points are deducted if a team has won less than 3 matches out of the last 5 Home matches.Home Advantage

  1. Recent Performance of the team

Current form of the team is one of the most important factors in cricket. If you are continuously winning matches, your confidence will be high and the probability of your winning will also be high. This factor takes into account Number of Wins in the last 5 matches played. Extra points are given if a team has won more than 2 matches out of the last 5. Some points are deducted if a team has won less than 3 matches out of the last 5.Recent Performance

  1. Decision of Batting First at the Venue

If you are playing a night match in Wankhede Stadium (Mumbai), you have to bat second because of the dew which makes it very difficult to grip the bowl in the second half of the match. Conditions of this type make the decision of Batting First/Second at a given venue a very important factor. Extra points are given to the team taking a fair call of Batting First/Second at a given venue based on the previous match outcomes.Decision of Batting First

After combining these 6 factors, final Probability of Winning is calculated for both the teams and the team with higher win probability is our predicted winner of the match.

Yes, it is so much easy to predict a cricket match. Now let us see how these factors actually predicted an outcome of an IPL match between Royal Challengers Bangalore (RCB) and Mumbai Indians (MI) played on 11th May, 2016 at M. Chinnaswamy Stadium (Bengaluru).


Factor 1: Playing Eleven

 Virat Kohli 79.92 0.00
 AB de Villiers 98.45 0.00
 Shane Watson 68.36 65.21
 Sachin Baby 0.00 0.00
 Chris Gayle 80.69 0.00
 Lokesh Rahul 32.99 0.00
 Stuart Binny 26.67 49.37
 Varun Aaron 0.00 66.01
 Sreenath Aravind 0.00 0.00
 Chris Jordan 0.00 0.00
 Yuzvendra Chahal 0.00 89.18


 Rohit Sharma 83.70 0.00
 Parthiv Patel 60.58 0.00
 Jos Buttler 0.00 0.00
 Ambati Rayudu 61.03 0.00
 Kieron Pollard 74.29 3.05
 Harbhajan Singh 37.09 91.05
 Nitish Rana 0.00 0.00
 Mitchell McClenaghan 0.00 64.40
 Jasprit Bumrah 0.00 44.79
 Tim Southee 0.00 42.01
 Krunal Pandya 0.00 0.00


Total Batting Score for RCB          = Sum of Individual BAT_RAT      = 387.08

Total Bowling Score for RCB         = Sum of Individual BOWL_RAT = 269.77

Total Batting Score for MI             = Sum of Individual BAT_RAT      = 316.68

Total Bowling Score for MI           = Sum of Individual BOWL_RAT = 245.29


Factor 2: Ground Type

Playing Venue: M. Chinnaswamy Stadium (Bengaluru)

Ground Type (Based on Research): Batting Friendly

Ground Dimensions (Based on Research): Small

Factor 3: Player Type

Player Name Bat Spin Fast Power
 Virat Kohli 1 0 0 1
 AB de Villiers 1 0 0 1
 Shane Watson 1 0 1 1
 Sachin Baby 1 0 0 0
 Chris Gayle 1 0 0 1
 Lokesh Rahul 1 0 0 0
 Stuart Binny 1 0 1 0
 Varun Aaron 0 0 1 0
 Sreenath Aravind 0 0 1 0
 Chris Jordan 1 0 1 0
 Yuzvendra Chahal 0 1 0 0


Player Name Bat Spin Fast Power
 Rohit Sharma 1 0 0 1
 Parthiv Patel 1 0 0 1
 Jos Buttler 1 0 0 0
 Ambati Rayudu 1 0 0 1
 Kieron Pollard 1 0 1 1
 Harbhajan Singh 0 1 0 1
 Nitish Rana 1 0 0 0
 Mitchell McClenaghan 0 0 1 0
 Jasprit Bumrah 0 0 1 0
 Tim Southee 0 0 1 0
 Krunal Pandya 1 1 0 0


Number of Batsmen for RCB (Bat_RCB)                = 8

Number of Spin Bowlers for RCB (SBowl_RCB)    = 1

Number of Fast Bowlers for RCB (FBowl_RCB)    = 5

Number of Powerhitters for RCB (PH_RCB)         = 4

Number of Batsmen for MI (Bat_MI)                    = 7

Number of Spin Bowlers for MI (SBowl_MI)        = 2

Number of Fast Bowlers for MI (FBowl_MI)        = 4

Number of Powerhitters for MI (PH_MI)             = 5

Bat_Factor for RCB          = ( Bat_RCB / 10 ) + ( PH_RCB / 10 ) + 1                   = 2.2

Bowl_Factor for RCB       = ( SBowl_RCB / 10 ) + ( FBowl_RCB / 10 ) + 1       = 1.6

Bat_Factor for MI            = ( Bat_MI / 10 ) + ( PH_MI / 10 ) + 1                        = 2.2

Bowl_Factor for MI         = ( SBowl_MI / 10 ) + ( FBowl_MI / 10 ) + 1           = 1.6


Factor 4: Home Advantage

Home Team                                                                    = RCB

Wins of RCB in last 5 Home matches (Home_Win)    = 2

Home_Factor                                                               = 0.5 + ( Home_Win / 5 )             = 0.9

Factor 5: Recent Performance of the team

Wins in Last 5 matches for RCB (Recent_RCB)          = 2

Recent5_Factor for RCB                                              = 0.5 + ( Recent_RCB / 5 )             = 0.9

Wins in Last 5 matches for MI (Recent_MI)              = 3

Recent5_Factor for MI                                               = 0.5 + ( Recent_MI / 5 )               = 1.1

Factor 6: Decision of Batting First at the Venue

Batting First Team                                                            = RCB

Wins of Team Batting First at M. Chinnaswamy Stadium in last 5 matches (Bat_First_Win)                                                                                               = 1

Bat_First_Factor                                                               = 0.5 + ( Bat_First_Win / 5 )         = 0.7

Combining the Factors:

Net Batting Score for RCB (Bat_A)             = 965.68

Net Bowling Score for RCB (Bowl_A)         = 152.96

Net Batting Score for MI (Bat_B)               = 1532.73

Net Bowling Score for MI (Bowl_B)          = 269.82


Win Probability:

Winning Probability of RCB

= ( ( Bat_A / ( Bat_A + Bat_B ) ) + ( Bowl_A / ( Bowl_A + Bowl_B ) ) ) / 2

= 37%

Winning Probability of MI

= ( ( Bat_B / ( Bat_A + Bat_B ) ) + ( Bowl_B / ( Bowl_A + Bowl_B ) ) ) / 2

= 63%

Where, Bat_A    = Net Batting Score for RCB

                Bat_B    = Net Batting Score for MI

                Bowl_A = Net Bowling Score for RCB

                Bowl_B = Net Bowling Score for MI

Hence, the Predicted Winner comes out to be Mumbai Indians (MI) with Win probability 63% and the Actual Winner of the match was also Mumbai Indians (MI).

To Spend Where It Matters….

Why does the allocation of marketing budget a dilemma for any CMO? Why does it stress them so much?



Here are few challenges that the marketers are currently facing:

  • 98% of the marketers affirm that digital marketing has moved into the mainstream. But only one-third of them have digital techniques incorporated into their marketing operation
  • Two-third of the marketers expect their marketing budgets will continue to grow by 10% in 2016 though the increase in ROI from them is not even close
  • B2B & B2C marketers dedicate nearly the same portion of their marketing budgets to digital commerce

(Source: Gartner CMO Spend Survey 2015-16)


Not only do they struggle with the budget allocation but they also have a whole mountain of real world constraints for budget optimization:

  • Running parallel campaigns may affect Budget Optimization – for instance, a high ROI campaign and a low ROI campaign that need to be run together may reduce revenue
  • Order of running two or more campaigns might affect their ROI; for instance, a radio ad followed by a TV ad about the same product might not add any value as the audience is already well informed
  • Audience reach might get compromised by abandoning the campaign with low ROIs and running only the one with the highest ROI
  • Time period in which campaigns are run might affect their ROI. Campaign A might provide higher ROI in summers while Campaign B might work better in monsoons


As much as allocating budget is a challenge, optimizing the same while solving for real world constraints makes it more complex.


So what’s the breakthrough?


An Attribution Model can provide a helping hand in the complex situation of multiple campaigns or other marketing activities to the CMOs. The attribution model helps in identifying the activities with high ROI which enables optimal allocation of marketing budget. It is developed using the historical sales data for the past campaigns. Attribution quantifies the influence of each advertisement impression on a consumer’s decision to make a purchase decision



The ROIs calculated from the attribution modelling can be applied to optimize budget when solving for different objectives with different constraints:

  • For instance, for revenue maximization, then I might use a combination of email marketing and SMS campaigns providing discounts to regular customers and social media marketing on LinkedIn to target working professionals.
  • For reach maximization, I would lead with a TV advertisement, a huge Bill-board and a free product sample to newspaper subscribers. Here my marketing activities would not be restricted to a target audience.



A vacation ownership company was burning cash and fast, in an effort to maximize its reach in India through TV ads, newspaper ads, lucky draws, banners, social media ads, you name it they  were trying it. It’s ROI on these campaigns was close to nothing. With Attribution modelling, they were able to identify campaigns with higher ROI at specific customer segment levels. While social media ads worked well with the digital savvy audience driving 30% lift in membership, lucky draws and newspaper ads helped maximize reach amongst non-tech native business and trader community further increasing the membership by 15% in two quarters. The attribution model not only helped in targeting audience in the most effective manner but also increased marketing budget ROI without any additional spend.


Am I spending enough on Marketing? Am I spending a lot?

Are my campaigns working? Are they getting in extra sales?

Well, if such questions are haunting you, marketing attribution might provide you the answer you seek!


For more information on Marketing Analytics service click here.

Robotically Yours…!!!


Will robots take our jobs? That’s not even a question any more, the answer is a resounding YES, the more pertinent question and one I think a lot of people have chosen to take an ostrich head in the sand approach to (though I think the English language has given ostriches a bad rap, they don’t really do that…anyway I digress and I have just started)

Most global think tanks have pretty similar conclusions- things ain’t looking that bright for the so called superior beings on the planet., in an interesting case of the biting the hand the feeds (creates) you, anywhere between 25 on the conservative to 50+ on the gloomy,  is the percentage of jobs to be lost for good  to automation /robotics etc.

I keep thinking am I being a Luddite…as in the “Luddite fallacy” – people who believed that mechanisation would be death of employability-while they were proven wrong and somewhere I am hoping that I will be too, because isn’t the bright side of this story supposed to be that automating routine work would mean more creative work for me to focus on, but there seems to be a robotic spanner in the works…

There is a school of thought, one that I would currently subscribe to, that says the concept might not hold true any more as, if jobs go from across sectors and people can’t find jobs in other sectors with  technology only growing faster and more engrained in every aspect of our lives, (the fact that you reading this on a smart machine in the palm of your hand more powerful than the computer that launched the Apollo 11, is proof enough) and what if society isn’t able to evolve and adapt faster than technology then what happens?
What does this mean for a country and its people, we already have serious youth-unemployment problems today.

What  happens when a large percentage of unemployed youths think they don’t have a future, if you  take away the element of hope and the silver lining that things will get better – the newspaper headlines of rioting as a result of civil instability seems to the first visuals that come to mind,
Here’s what happens if 45+% of a country is unemployed  – 1)  They are not paying taxes so where is the government going to have money to support the 100% of the population and 2)  They are not consuming – so if they aren’t consuming that would stop the cycle of production…the wheels come off…you get the picture.

There is an urgent need for massive amounts of retooling, forget retooling…I would go as far as to say reimagining (65% of children entering grade school this year will assume careers that don’t yet exist) .

While individuals and families need to think about this with a degree of urgency; Policy makers, economist and planners need to look at this with a state of PANIC (of course they need to get their heads out of the sand first)

Can we at least start a conversation somewhere…Here perhaps or at our local and state legislators, maybe we need to nudge more than a few people…


Who drives CX now? CX Strategists or Customer?


Wondering what your goals for exceptional consumer service be this year?? Well that’s the ubiquitous question on everyone’s mind- especially with the ever increasing tech savvy consumers, who have good knowledge of how this whole experience cycle works. All CX evangelists are now constantly pushed to be one step ahead in providing the best experience to their customers.  So, what should your focus this year? Can you have a well defined CX strategy in place to counter the ever changing business landscape and be at the top of your game too?Let us see what Customer Experience Trends need to be tracked, especially by customer service departments, in order to stay ahead of the pack!

  • Consumers  are set to be participating across multiple channels- When it comes to experience, today’s customer is smart enough to understand what is good and bad by keeping themselves up to  date on advertising or marketing promotions and tend to pit one brand against the otherConsumers also know that they have several options to raise an issue- and the ability to quickly switch to a channel that is the most popular and provides instant gratification–like a tweet complaint- staying away from from the longer process of complaint handling. This is likely to gain more and more traction in times to come . Now that customers are familiar and have become more active in sharing reviews and feedback, the firms are taking it as a challenge and trying to meet  expectations by not leaving a single instance to chance. This, in turn, has given a chance to organizations to analyze their wide customer base and possible causes of attrition
  • questions
  • More active investments in CX initiatives – especially collecting feedback and suggestions– Gone are the days when businesses looked for loyalty without investing in the experience. The whole scenario is changing,and by 2017, it is expected that 50% of product investment projects will be focused on CX innovations. Interestingly, according to a recent survey by Gartner, the firms who implemented CX projects concentrated on improving collection and analysis of consumer feedback. This is a great way and the best time to adjust, transform and start investing in CX
  • Customer loyalty
  • Self Service Platforms gaining traction– Most of the customers would like to lead in resolving issues by themselves and they expect companies to provide them with a self-service platform to speak out and participate and collaborate. Gartner predicts that by 2020, the consumers will manage 80% of their relationship with the vendor without interacting with a human. This means, organization need to be well equipped with technology to enable such interactions seamlessly.
  • images (1)
  • “Mobile” will be the customer experience trend setters. With the increase of mobile usage among different age-groups, mobile apps will play an active role in customer interaction and customer support. On top of that, social media is going to be your best buddy as the voice of the customer for both bouquets  and brick-bats equally. Building effective communication and tracking and monitoring these channels should be the top most priority for any consumer facing organization.
  • inAppBugReport

Customer service and support is done being the auxillary node to CX operations . Its time to make it core to the new customer experience strategy by enabling it with new tools and technologies.

Want to revolutionalize your customer service?? Try ServeSmart – Putting the “ME” back in Customer Engagement!!!

To BE or not to BE!



As I browse through my LinkedIn profile, I see a range of innovations and technologies adopted by various companies round the world. The purpose is same, to derive maximum return on their investment, to secure the maximum possible customer range. As more and more companies enter into the race for serving the maximum number of customers, the stakes are also getting higher and higher. Companies continue to find out ways to gain that competitive edge. However, the answer has been here all along. A simple tool, in my opinion, for a better Customer Relationship Management strategy, is understanding the principles of Behavioral Economics (BE).

Behavioral Economics in simple terms, may be defined as “a method of economic analysis that applies psychological insights into human behavior to explain economic decision-making”. The answer is simple. To maximize their ROI/Profitability, companies would have to take a deep dive into human behavior (read: Behavioral Economics).

Every aspect of marketing tactics and strategy—from advertising, branding, messaging, promotions, and placement to pricing, product features, distribution, packaging, and distribution— ultimately is aimed at influencing the purchase behavior of consumers based on behavioral economics. I doubt very much that Adam Smith would disagree.

In real world, we can see numerous instances where a theory of BE affects customer decision making.

Does it make sense to buy new, unproven technology products or cars with no track record; to shop in expensive stores for products which are available for less elsewhere; to buy early in the season rather than wait for sales; to pay more for a specific brand instead of buying unbranded stuff? It all comes down to the individual consumers’ set of values: If they are an early adopter technophile or a “car person”; prefer the status/service/whatever of more expensive stores; want to enjoy something now rather than wait; or appreciate being identified with a particular brand. All of these decisions are perfectly rational reflections of the trade-offs between economic maximizing and other values that consumers have.

It is thus clear that for a better CRM strategy, companies have to look for ways to a better understanding of human behavior. But how will this be possible? Companies have to indulge in deep research using various methodologies which would help them not only in analyzing the trends and patterns in consumer behavior but also in learning from the previous mistakes of other companies. This is the reason there is a huge need for Market Researchers in this space. The three step strategy for profit maximization would then be:

  • Putting yourself in your customer’s shoes (having a deep insight into the customer’s mind and predicting their behavior to different situations)
  • Using data to understand your customers (analyse previous consumption patterns and make predictions based on that)
  • Asking your customers what they think (through a number of surveys, quizzes, feedback forms etc.)

The examples discussed above prove that more often than not, what a company might consider as a fool-proof strategy for profit maximization, may not turn out to be one. They have to consider the possibility of customer decisions that are based on certain values and are not, in recognized terms ‘rational’. The failure to do so, may lead to failures of even those products, which you thought were your ‘Brahmastras’. Examples are many: Apple’s Newton and Lisa, Motorola’s smartphones, HP’s Touch-pad. The time has arrived when all the companies should sit back and try to find answers to one simple question: What do customers really want?

Loyal customers are the reason for that “feel good factor”

“Customer is the king” is something I have been hearing since my childhood and it’s the truth indeed. What is that ‘something’ that mesmerizes a customer to come to your store time and again?
In traditional retail business, customer experience was often neglected and was difficult to measure-and not much has changed even today. But this traditional thought is on a verge of becoming a history, soon, as the retailers are becoming more customer-oriented and are trying hard to channelize most of their efforts in providing exceptional services and better customer experience. After all, the most satisfied customer is the one who often relates himself to your brand and ultimately becomes a loyal customer.

Customer experience management plays an important role in retail sector. For example, in case of retail Jewelry store, a customer who walks in is spontaneous and always is ready to compare every single detail starting from the price tags to the design, pattern or quality of the gems used. Today’s consumer is more advanced and educated savvy about the product information and its need. If the staff at the store is not well informed about their offering, the consumer will not be willing to make a purchase.

It is evident that a great customer experience drives loyalty and revenue as it encourages them to look up to you every time they have a matching need, ultimately resulting in greater turnover and profit for your business. This cycle continues-A business is richer with more loyal customers, more referrals and more brand advocates.

Taking time to talk to customers about their individual needs rather than just processing a bill is an effective way of discovering opportunities to engage with them better. A visitor to a satisfied customer to a loyal customer to more dollars in the bag – ‘Isn’t it a good feeling’? :)

Customer loyalty

Analytics as a game changer in retail sector.

“Being a girl I should be more shopaholic, but I am not. One day, me and my friend went to central mall and saw that the accessories section which was there on first floor has been shifted next to ethnic wear section, which made us buy earrings matching my kurti right away, otherwise I would have delayed it. I was wondering, why they have reallocated the sections”… there is nothing called coincidence in this, here is where analytics plays a critical role in retail sector.

With customers becoming more techno savvy, retail sector is experiencing growth both at online and offline fronts, opening up new stream for forward-thinking retailers. With the wealth of data, analytics holds the potential to drive real front line differentiation for the particular retailer. Retail Analytics can be applied on various business functions be it customer management, pricing, strategies, Fraud detection, supply chain management, workforce analysis, cross selling  and up selling analysis etc.

Here are some of the business functions on which analytics can be applied:

Customer management: With informed and challenging customers in this highly competitive environment, traditional methods of customer engagement have become redundant. Customers’ preferences and behavior keep on varying over time. Therefore, meeting customer expectations has become priority for retailers. Analytics deliver insights on customer engagement assessment, churn rate, share of wallet (identifying customers with the propensity to increase spending and opportunities for the same). Using predictive analytics, retailers can increase their sales and reduce potential losses by proper targeting.

Pricing strategies: Pricing is a key to survive competition as well as the growth for any entity. Analytics provides valuable insights on the demand for a particular good, competition in market, customer segments and their shopping behavior, with which price can be set up accordingly without eroding the customer base.

Fraud detection: Analytics plays a role for evaluating processes within the organization by analyzing the unusual patterns encountered; this reduces the risk for the organization. More complex the operations,  higher the chances of fraud. With periodic analyses, analytics captures where things are going wrong.

Workforce analytics: Quality matters more than quantity, the same holds for workforce employed. By increasing their efficiency, one can better handle the customer base. Analyzing store traffic, required tasks etc. increases workforce efficiency and productivity. Analytics also captures productivity of the employees which influences their hiring and retention decisions.

Supply chain Management: With increasing customer base, maintaining inventory, reducing transportation cost, increasing collaboration with suppliers becomes crucial for the retailer. With the help of data, predictive analytics can be applied to draw insights related to amount of inventory that will be required and at what time.

Cross-selling and up selling : This is one of the strategies used by retail outlets to increase their revenue. By analyzing the shopping pattern of customers, retailers can suggest the complementary products to the customers. With appropriate product mix it can make offers that can be mutually beneficial for both customers and outlet. Likewise, up selling to the target customer requires analytics to identify them.

These are some of the focused areas where analytics has proved to be invigorating with respect to overall growth of a retail store and there are many more areas to explore in retail analytics.


As a Data Analyst, we flaunt our skills for analysing the data using different data tools, for creating some funky charts from the data and for predicting the future with the data. To my surprise, we all miss out on one step which is a must for making any sense out of data. Yes, THE ONE BIG STEP before analysing which takes the major chunk of our time. DATA CLEANING!!!!

I have never seen anyone flaunting about his/her Data Cleaning skills. According to research, every Data Analyst follows an 80-20 Rule which means 80% of his time is spent on Data Cleaning and the rest 20% is spent on Data Analysing, Data Visualization, Data Mining, Data Warehousing, Data integration and all other data related processes.


One Step before Analysis

In a data analyst’s life, the biggest hurdle is Data Cleaning because it is the only thing on earth which can neither be generalized nor automated. Every piece of data is unique in itself and if by any luck the data is collected from “the common man” then it becomes even more unique. Don’t get surprised, I can prove that to you. You just have to play a small game: Ask any 5 of your friends to fill up the following details about themselves individually without seeing how the other person has filled it.

Name Email Id Phone Company Name State

I am pretty sure you’ll get responses like the following.

Name Email Id Phone Company Name State
Anant 0900****999 Godrej UP
Barinder Singh 7777****77 Flipkart Chandigarh(UT)
Anurag 7676****76 Maruti Suzuki Uttar Pradesh
Chandan Rishi +91 9666****66 Bajaj Chd
Simarjeet Singh 9788****88 Amul Punjab

You can observe how different people mention their names, phone numbers and states . These are just basic questions and still have so many things to clean. Think about a full questionnaire having 50 such fields to be filled by over 200000  candidates. That sounds like a big task, isn’t it? That’s why 80% of our time is spent on cleaning the data.

Some of you might be thinking, what’s the point spending so much time on something which is not so cool? What if we skip this part and directly start applying our algorithms on the data? Will it make any difference? The answer is a big YES. It makes a lot of difference. It’s an absolute myth that you can run an algorithm over raw data and have any useless insights pop-up.

So, if we cannot skip this part then we should atleast try to tackle it in a systematic manner. No two data preparation techniques are the same so automation is hard. Here I have attempted to jot down few techniques which are more frequently used in data cleaning. I would like to call it a checklist for data cleaning.

Checklist for Data Cleaning:

  • Store your data in a data frame with suitable column names
  • Each column of your data should be consistent with one data type (like numeric, integer, character)
  • Converting variables (columns) to a suitable data type(like if age variable is stored as a character type, convert it to a numeric type so that you can apply relevant mathematical operations on it)
  • Apply Date Conversions as per the requirement
  • Apply Character manipulations:
  • Removing prepending or trailing white spaces
  • Trim strings to a certain width
  • Transform to upper/lower/proper case
  • Search for strings containing simple patterns (substrings)
  • Approximate matching procedures based on string distances
  • String Normalization(transforming strings to a smaller set of string values which are more easily processed)
  • Removing the missing values/NA’s from the dataset
  • Removing/Trimming irrelevant variables
  • Adding new variables into the dataset using variables which are already present
  • Detection of Inconsistency which means checking for information which violates the basic logic
    (like Phone number containing alphabets)
  • Selection of field causing inconsistency
    (like for State of Texas Country cannot be Canada. In this case, it is not very clear whether State is wrong or Country is wrong)
  • Simple transformations which means finding out patterns in data and transforming them accordingly
    (like a simple Phone Number can be written in many formats (172)-64**82 or 0172647**82 or +92-172-647**82 but they all are same so we need to find out patterns and replace them to one standard format)
  • Converting to a single unit
    (like if a Height variable has heights in centimetres, feet and inches then we should first convert them to one single unit and then proceed with any computations)
  • Deterministic Imputation which means sometimes it is possible to find a missing value of one variable with the help of other variables
    (like if we know for a particular entry the State is Karnataka and Pin Code is 560102 then definitely City will be Bangalore)
  • Checking for Outliers

This checklist is definitely not exhaustive and can have many more points in it. If you have any other points for data cleaning, please mention them in your comments. Let’s make this checklist bigger and better so that we all can save our precious time.

data cleaning

Happy Cleaning!!!