To Centralize Or Not To Centralize [analytics], That Is The Question

The structure of analytics in large organizations can take many forms—from having a gazillion analytics micro-teams embedded in each function or BU, to completely centralized analytics at the corporate level. What is the right strategy? What should your organization do?

Well, in that respect, the title of this post is misleading. To centralize or not to centralize, is actually NOT the question. If you think of centralization on a scale going from ‘not at all’ to ‘fully centralized’, the real question is what is the right level for you?

To answer that question you must be aware of the pros and cons of moving one way or the other on that scale. Having been a part of multiple “reorgs” and that have gone up and down on the scale, and having influenced some of those movements some of the time, I have some first hand insight into this.

So here are the top 5 key trade-offs when faced with organizational structure of analytics.

1. Consultant Mindset vs. Deep Personal Investment: God bless consultants, I have nothing against them. But one thing they cannot claim is deep emotional investment in and sense of alignment with the organization they are working for. This is what high degree of centralization does. Analysts are assigned to BU’s or functions based on prioritization of the project and resource constraints. Their mindset is like that of a consultant, where you work on a project, crunch the numbers, deliver the insights and you job is done… time to move on to the next one. With analytics embedded within the function, there can be full integration of analytics with the project right from its conception. The alignment of purpose this creates, produces very non-linear synergistic effects with respect to the value derived from analytics. This alignment/ownership, of course could be a problem by itself, which brings us to the next point

2. Objectivity (or at least the perception of it): If the analytics team reports into the owner of the domain, and their rewards are aligned with the success of the projects being analyzed, the objectivity of the analysis could be in question. The analyst could potentially introduce a bias to make the project/initiative look better than it actually is. With analytics, credibility is everything. The perception of lack of objectivity could be devastating for the entire group/organization. If you believe that numbers cannot lie, you are either not in the field of analytics or are deluded. Read How To Lie With Statistics for starters.

3. Beaurocracy vs. Efficiency: Centralization brings beaurocracy, sometimes copious amounts of beaurocracy, depending on who is the heading analytics. Everything needs to get into the pipeline, and get prioritized, and get resources allocated against it. There are protocols for communication, to ensure the BUs are not side stepping the process (this seems like paranoia but I have experienced this first hand). It could suck the excitement out of a very creative job (i am talking about analytics of course), and turn analysts into full time project managers (God bless project managers, I have nothing against them).

4. Redundancy vs. Effectiveness: With the “embedded” model, it is easy for different analytics teams to get redundant in their analyses and continually reinvent the proverbial wheel. Centralization dramatically reduces redundancy, thus making the analytics team more effective. There is more knowledge sharing, a better sense of community of like-minded people, and more flexibility in leveraging a wide range of skill sets among analysts. This improves the throughput by improving the utilization of resources, thus also making the team lean.

5. Silos vs. Big Picture:Small teams of analysts embedded within the BU end up working in silos. While they become experts in their own domain, they run the risk of losing the big picture. This can be detrimental not only to the quality and relevance of the insights generated, but also to the career growth prospects and job satisfaction of the members of analytics team.

So that brings us the decision point—what is the right level of centralization. BUs or functional teams will always resist centralization of analytics because they dont get dedicated capacity anymore. Analysts, on the other hand, would likely (but not always) resist decentralization. The holy grail is to find the level at which both the stakeholders are equally happy (or equally unhappy!), such that analysts get some opportunity to move around, cross-train and gain breadth of domain, and at the same time, have the chance to develop deep domain knowledge in a specific part of the organization and to influence/drive the strategy for the BU as opposed to reporting out data. Finding that sweet spot is not easy, but this hopefully gives you a sense of what you are looking for in the first place.

Posted in Uncategorized | Tagged , , , , , , , | Leave a comment

mukulpatki:

This is an interesting post in Forbes, with an excellent critique from 1Click Analytics. Conversion is a problem that is front-and-center for me at paypal 7 days a week. Found this entire discussion very useful.

Originally posted on 1 Click Analytics:

So here is an article recently published by Forbes.com where they relate things websites should do to improve Conversion Rates.

http://www.forbes.com/sites/quora/2012/03/28/what-are-some-top-strategies-for-conversion-optimization/

1 Click Analytics, as a company that is dedicated to the practice of conversion optimization, as soon as the Google alert popped up in my email I gave the article are read. Surprised by the content or lack there of i read it again.

Firstly what the article touches on is not bad or incorrect information, some of the ideas touched upon are standard best practices and I would never want to infer that they hurt conversion rates because I dislike the article.

The reason for my unwillingness to embrace and article with some good information in it, is because in my opinion the article is fundamentally flawed in the approach it suggests to increase conversions. Mostly because it fails to directly address the truly most important thing in any conversion scenario, the consumers.

What the article talks about is window dressing and…

View original 255 more words

Posted in From Other Bloggers | Tagged , , , | 1 Comment

5 BE Principles All Marketers Must Understand

 As discussed in my previous posts, understanding the consumer psyche and the irrationality of the human decision-making process is key to developing winning value propositions or product features to test in the market. Here is a discussion of 5 Key Bhavioral Economics (BE) principles (among dozens) that all marketers should not only understand but internalize. 

Power Of Free: Can reducing the price of two commodities by the exact same amount, compleletely reverse consumer preference of one over the other? Traditional economics says NO. But indeed it is possible.

A group of researchers offered participants of a study a choice between purchasing a Hershey’s Kisses chocolate for 1-cent ($0.01) or Lindt Lindor chocolate truffle for 15 cents ($0.15). The participants, recognizing this as a good deal since the price differencial in a supermarket would be larger than 14 cents between the two options, overwhelmingly chose the latter. However, when the price of both was reduced by 1 cent, thus making Kisses free and the Lindt Lindor for $0.14, the preference completely reversed with an overwhelming majority choosing Kisses!

What happened here? Nothing had changed–consumers would still ge t the same amount of incremental joy (consuming an exotic truffle vs. a regular candy) to the same amount of incremental pain (spending $0.14 more). The preference should not have changed. So why did it? Well, our response to price reduction becomes very non-linear when the price reaches “free”. We just love the word “FREE”. It evokes unreasonably positive feelings in the brain. Just the sight of the word “free” releases large quantities of dopamine in our brain to make us feel happy, and we end up responding irrationally.

So how does this play out in the real world. We get inundated with “free” offers every day and may believe that this does not affect us. But consider two economically identical deals– one messaged as ‘buy 1 get 1 free'; the other messaged as a volume discount deal as ‘get 50% off if you buy two’. Which one are you more likely to respond to?

Dominated Alternatives: Can introducing a third decoy option make you more likely to choose the option I secretly want you to choose?

Consider this scenario at the Economist. Potential customers were given the two subscription offers shown below– essentially an ‘online only’ subscription for $56, and and ‘online + print’ subscription for $125.

 

A large majority of people chose the first option ($56), although the second option ($125)was preferable to the publishers. They then introduced a third decoy option, that they knew nobody would prefer–$125 for print only. As expected no one chose the third option, but something magical happened! An overwhelming majority now chose the second option ($125 for online+print)! The mere introduction of this third option, made option #2 look very attractive–you were getting online version for free now!

What happened here? Well, this goes back to the idea that consumers have a very poor understanding of what a commodity is truly worth. They had no idea what a print or online subscription of the Economist is truly worth in $ terms. The first scenario with two options they had nothing to compare either option to. But with the introduction of the third option, option #2 and #3 are comparable and #2 wins handsdown (you are getting online version for free after all !) . Option #1 has no comparable so it gets left out.

This principle has been demonstrated successfully in many different scenarios. The most bizzarre, according to me, is one of dating. Pariticipants of this study were shown pictures of 3 individuals of the opposite sex and asked which one would they prefer to go out on a date with. Only, there were only two individuals in the pictures, the third was a digitally altered slightly inferior version of one of the two. So think of it as A, B, and inferior B (say B’). An overwhelming majority chose B in this scenario! The idea is the same–no comparable for A, so A gets left out; B and B’ look similar, B being more attractive. Hence B wins in a large majority of cases.

Next time you are evaluating vacation packages, or buying a home, pay attention to how different options are being positioned. These professionals have figured this stuff out through experience, even if they dont articulate it this way.

Irrational Value Assessment: Are you more likely to admire a $5 bottle of wine, if I lied to you and told you that it costs $45? Research says you are. Members of the Stanford Wine Club were invited to taste 5 bottles of wine and rate them based on their liking. Only, there were actually only 3 different wines in those bottles– two wines had two bottles each. Each bottle was marked only with the price tag and nothing else. Some of the same wines were marked at significantly different prices. For example, the $5 wine and the $45 wine were actually the same, the true cost being $5. There was a clear correlation between the rating of the wine and the price tag — more expensive wines got systematically higher ratings. So the $45 bottle of wine got a significantly higher rating than the $5 bottle, although they were the exact same wine!

In another experiment, the same group was asked to rate the same wines again. Only this time even the price tags were absent. The cheapest wine was ranked the highest in this case!

Now, before we start calling these wine-experts snobs, consider this. Prozac was tested against a placebo. Only, the placebo was sold at a higher price ($2.50 per pill) than Prozac ($2.00 per pill). Placebo outperformed Prozac!

Consider another experiment, where students were given a cafeine+sugar rich drink that was supposed to improve their alertness and focus in the short term. Their task was to solve as many puzzles as they can. Half of the group was asked to pay the full price of the drink, an the other half was given a significant discount on the price. The group that got the discounted drink, solved 30% fewer puzzles! This result has been consistent in multiple such studies over time.

So what is going on here? Well, turns out that we inherently expect cheaper stuff to be inferior. This feeling runs so deep, and its effect so profound on our brain, that the cheaper stuff truly ends up having inferior performance. It becomes a self-fulfilling prophecy. So the folks from Standford Wine Club, were not being snobs when they rated the ostensibly more expensive wines as tasting better. They truly did enjoy the wines with higher price tags more. This was demonstrated by the increased activity the pre-frontal cortex of the brain, when a the same experiment was done under an fMRI machine. Consumers of Prozac, deep within, expected a poorer performance compared to the more expensive Placebo. This expectation and conviction was so strong that it did create inferior performance in the body.

Decision Paralysis: Can reducing the number of options available to consumers, actually increase sales? Turns out it can!

In a study to prove this point, researchers sat down in a supermarket with bottles of Jam on display. The expectation was some users would stop by, fewer would taste, and yet fewer would purchase. One group sat with 6 varieties on display, and the other with 24 varieties on display. While more people stopped by in case of the 24-jar display, the number that bought was 10-times less than the 6-jar scenario (3% vs. 30%).

Jars Frank Cooper's jam

Jars Frank Cooper's jam (Photo credit: Wikipedia)

What is going on here? We thought more choice is what consumers want. Turns out, when faced with too many options, we are unable to evaluate them all, and end up deciding not to buy at all. This has been demonstrated in many different situations. In a company with a voluntary savings program the pariticipation in the program fell by 2% for every 10 mutual funds added to it.

At the heart of this finding is our inability to process too much information. This concept is known as cognitive load, which incidentaly does have a magic number — 7 (+/- 2). Consider the study where participants were told (falsely) that they were participating in a study on long-term memory. They were asked to memorize a number, walk down the hall, wait for sometime, and repeat the number from memory to a different researcher in a different room. Half the group was given a 2-digit number, and the other half was given a 7-digit number. But as the participants walked down the hallway, there were refreshments available with a choice of a decadant chocolate cake, or a cupu of fresh fruit. This was the real test — exercising self control when you mind is occupied. The study found that a majority of participans in the 7-digit group chose the cake, while a majorit in the 2-digit group chose fruit.

What is going on here? Turns out that the part of the brain that is occupied with memorizing irrelevant illogical information such as random digits, is the same (pre-frontal cortex) part that is charged with exercising self control. Remembering 7 digits is a tough task–it is approaching our cognitive limit. The brain is so preoccupied with trying to remember those numbers, that it literally does not have the ‘bandwidth’ to exercise self-control.

Attribute Priming: Can just talking to customers about a certain attribute of the product, make them desire that attribute more? Research says YES!

Consider the following study. Researchers approached customers planning to buy laptop computers at an electronics store. Half of them were asked about their memory needs, and the other half were asked about their processor-speed needs. This was not steering or leading by any strech. Simply asking them what their needs were in those areas. Turns out, that the group that was asked about the memory needs ended up buying computers with higher memory, and those in the other group ended up buying computers with higher processor speeds. Just getting them to think about certain attributes of the product affected their decision in favor of that attribute.

In a different study, where people were in line to pick up either yogurt or fruit, half of them were asked how they felt about yogurt, and the other half were asked how they felt about fruit. Later it was found that just talking to them this way, greatly biased their decision about what to eat.

Posted in Uncategorized | Tagged , , , , , , , , , | 2 Comments

Can Your SSN Determine How Much You Pay For A Commodity?

What a preposterous idea! Or is it? Consider the following experiment. The audience in one of my talks recently was asked to pretend they are bidding on a product (the projector being used for my presentation). But the process of bidding was a little unusual. Participants were supposed to answer the following thee questions on a piece of paper:

1. What are the last 3 digits of your SSN (social security number) … say, 123

2. Are you willing to pay this amount (i.e., $123) for this product?

3. What is the maximum amount you are willing to pay for this product?

When the results were analyzed, a huge correlation was found between the the last 3 of SSN of a participant, and the maximum amount he/she was willing to pay! Specifically, participants with SSN in the highest 25 percentile (i.e., 750 to 999) were willing to pay 5 times more than the participants whose SSN ended between 000 and 250. How could that be?

This phenomenon is known as Anchoring. The underlying idea is that consumers have a very poor sense of what a commodity is truly worth. So they are desperate to anchor on to something. The moment we forced them to think if they were willing to pay a certain price for the product (namely the last 3 of their SSN in USD), they started gravitating towards that number and the amount they were eventually willing to pay ended up being closer to the last 3 of their SSN than it would have been otherwise. This idea was first introduced by Dan Ariely in his book Predictably Irrational, and has since been tested and confirmed over and over, under many different scenarios. I myself have confirmed this multiple times by testing it on my unsuspecting audience!

How does this manifest itself in the real world? Well, if you are buying a TV set, do you think the maximum amount you are willing to pay for it would go up considerably if the first TV you see in the store is a $3000 plasma? Where do you think the store keeps its most expensive TV sets? Do you think the amount you are willing to pay for a meal would be dramatically different if the first entree you see is a $60 steak?

Traditional Economics Vs. Behavioral Economics

The example above illustrates an important point (beyond the fact that we are gullible suckers constantly falling for the tricks of smart retailers!) — that our decision-making process is not as rational as we would like to believe. This is the premise of the field of behavioral economics.

Traditional economics assumes that all parties, given a choice, would choose the economically optimal option for themselves. And as long as everybody behaves this way consistently, the “invisible hand” with take care of the rest, and everybody would win. This is the idea behind the efficient market, the free market. It is based on the assumption of a homo-economicus being — a person with the brain of einstein, and the memory and efficiency of a super computer so that she can take all relevant information and comptue the most economically viable decision in real time. If this sounds like a fantasy, that is because it is!

Most ordinary people, such as myself and most of you reading this blog, make decisions based on emotions, fears and irrational considerations, which is the assumption that Behavioral Economics relies on. Such decisions are often economically sub-optimal.

As irrational as our decisions are, the good news is that there is a method to the madness. The decisions we make may often be irrational, but they are predictable. Hence the apt title of the pioneering book in this field by Dan Ariely— Predictably Irrational.

There are dozens of principles of behavioral economics, with scores of examples based on  studies done by social scientists that prove not only the irrationality of our behavior, but also the predictability of it.

 Next Post: More mind-bending examples of human irrationality and how businesses are using this body of knowledge to steer us, influence us, and sometimes downright cheat us!

Posted in Uncategorized | Tagged , , , , | Leave a comment

A Test Is Only As Good As The Hypothesis It Tests

Test-and-Learn is the new mantra in the tech world.  Be it new product features, new policy ideas, marketing messaging, or email content, test-and-learn is gaining steam, graduating from simple A/B tests to complex multi-variate tests. However, when it comes to fundamentally understanding consumer behavior and preference this approach has not lived up to its promise in the industry. The primary reason for this is that due the very nature of this approach, you have to develop hypotheses upfront. Even the most diligently designed test will give you the best option among the ones you tested. It is incapable of telling you if you missed a great idea.

So developing the right hypotheses to test is key. To understand how to develop creative, effective hypotheses we must first be clear on why we are testing in the first place.

Three Reasons For Testing

1. We are not our customers. A senior executive at a large payments company asked a group of about 100 employees a simple question — if you were at a physical store buying something for less than $50, how many of you would prefer to pay using cash. About 10% hands went up. Then he went on to share a new market research that suggests 80% of American consumers (potential customers of this payments company) would prefer to pay using cash for offline transactions of less than $50.

This huge gap in the fundamental consumer mindset runs home the point that we are not our customers, i.e., we are not a representative sample of the base that is our customer.

We must let go of the belief that we can sit in a conference room with decades of collective experience among us, and dream up what customer would want, like, and prefer. So if we cannot trust ourselves to know what customers want, why not ask them?

2. Customers are unable to articulate what they want. Customers are particularly poor at articulating what they want, primarily because they don’t know what is possible and what isn’t. This is the reason why customer-focused companies such as Proctor and Gamble and Intuit like to passively observe customers using their products. It gives them invaluable insights for improving their products and services.

A large producer of consumer goods released a new laundry detergent in a country in Latin America. Sales were not doing good despite big advertising budget and objective evidence of superiority of their product over the competition. They followed their customers home to observe how their product was being used. It was nothing short of a revelation! This was a third world country where women washing clothes had to fill buckets of water and bring it to the place of where clothes were to be washed. The company ended up creating a “high-efficiency” detergent that required half the amount of water as before. This reduced the labor of these women by half, and the product was an instant success!

There is a great talk by Malcolm Gladwell illustrating the same point in the context of spaghetti sauce.

There is no way these women could have asked for a high-efficiency detergent because they did not know it was a possibility. Nor could the customers that Gladwell talks about have known that what they really wanted was extra chunky sauce. Customers are often unable to articulate what they want primarily because they don’t know whats possible and they have not experienced it yet.

So, if we cannot predict what customers want, and customers are unable to articulate what they want, what is the solution? Why not create a few options and ask them to choose? Here in lies the holy grail of experimentation — developing the right hypotheses to test, ensuring that one of the ideas we are testing is the “magic answer”. Development of creative and effective hypotheses relies on our understanding of consumer behavior — the understanding that response of people is often irrational, but that there is a method to the madness.

This brings us to the third reason why we need to test in the first place.

3. Customers may respond to our offers and propositions irrationally. The field of behavioral economics sheds ample light on how irrational the response of most people is when faced with certain kinds of situations. While traditional economic theory (based on the homo economicus being) says that all parties will act in their own economic interest and the invisible hand of the free market will take care of the rest, behavioral economics (based on decades of social science research) shows that people often make economically sub-optimal decisions based on emotions and irrational considerations. Being keenly aware of this body of knowledge allows you to develop hypotheses (and treatments addressing them) that are effective and have a high probability of producing a disproportionate impact.

Whats Coming: In the next post I will discuss some principles of behavioral economics that highlight the irrationality of decisions, illustrate the contrast between a typical consumer and the homo-economicus being, and yet impress upon you that it is all predictable.

Questions for readers: Do you have examples/experiences relevant to this discussion? Have you observed something similar to the “laundry detergent” example above, where customers did not know what they really wanted.

Posted in Uncategorized | Tagged , , , , , , , , , | 8 Comments