Spend a vision with me

Tuesday, 9 October 2018

Analytical Maturity, why organisations should embed analytics across all depts, so all depts grow together.

Last year, a friend of mine asked me my thoughts on an HBR article “If Your Company Isn’t Good at Analytics, It’s Not Ready for AI” (1). It was an interesting one and caused me to pause. I had started writing a blog post about it but abandoned it half-way. Today this topic is more relevant than ever. So here goes.

If you read some of the thousands of articles, or listen to people’s reactions around AI, you would have thought everything is AI nowadays. But this isn’t the case; why?

Similarly, the same friend was surprised to learn that some organisations such as tencent has a huge team of data scientists; so does Zhong An(2) by the way (at least 54% of their employees are engineers or technicians). That does seem to indicate that the deeper you get into analytics/ml/ai, the more data scientists/machine learning engineers… you need, not less.

Why is this so? Why can’t an organisation “put all the data in one machine and all answers come out” like what one of my bosses wanted? Can’t everyone just adopt AI?

If you build it, they will come (3)

I love the movie “field of dreams”. It’s about a whacky farmer who decides he wants to put all his eggs in one innovation and goes full hog, uprooting his previous business model and establishing a new one driven by his passion (and a voice in his head).

To me, the most memorable line of the movie is “If you build it, they will come”.

This is often the idea behind transformation projects, where the idea is to transplant analytics (let alone AI) into an organisation. Whether it is from a bunch of external consultants, or via hiring a relatively experienced head and build a team internally, the results are most often the same. There are only ghosts on the baseball field.

This is why I consider most Insurance company big data labs a failure; the men and women in white coats/uniforms are playing amongst themselves, and life goes on as per normal for ordinary folks, like 2 distinct worlds. So what is the RoI of these labs? To me, I would call this a failure since I believe the main aim of analytics/”Data Science” is to generate RoI and benefit all parties – grow the pie and everyone gets more.

So why the do many attempts at embedding analytics in organisations end in failure? Why do 85% of data lake projects fail (4)? The technology is there. Sure, there messy implementations, broken pipelines, choked pipelines, clogged processing engines, extremely dirty data where ELT works only sporadically or knowledge disappears with staff attrition…

harder

Well, as the article (4) says “More than 85 percent of respondents report that their firms have started programs to create data-driven cultures, but only 37 percent report success thus far. Big Data technology is not the problem; management understanding, organizational alignment, and general organizational resistance are the culprits. If only people were as malleable as data.”

Even if you build it, they may not come (for longer than the rah-rah show).

Basically, production of analytical pieces is ‘easy’. You drop me or any decent analytical person in a datalake, throw an SMEs and a good data engineer (my claim of analytics as a team sport (5)) and we are bound to catch some fish for the client and clean it, and cook it for him/her; but he/she is unlikely to know how to continuously include fish in his/her diet unless the people from the client team are ready.

What is most often missing is the ability to consume the pieces of analytics in a consistent and on-going manner.

Consumption of analytical/”data science”/AI output s not as obvious as you may think. And this is the part that most failed implementations have in common, (and also why I have consistently refused to join any organisation trying to transform itself in terms of analytics if the role focuses on the IT production side).

There can only be one, can’t there?

You could argue that it is only necessary to have 1 good consumer in the organisation, 1 department adopt analytics/”data science”, show the benefits and drag all other departments along. Afterall, once a piece of analytics is successful, each head of department can choose to adopt analytics and enjoy the benefits at a much lower risk.

There are 2 flaws in this argument. Firstly, we are forgetting the ability to consume, wanting to consume is one thing, but being able to (analytically mature enough) is not a given, Secondly, departments rarely exist in isolation in an organisation. A simple example will illustrate this.

A while ago, I was demonstrating how quickly a selection of customers based on their behavioural similarities can be made and readied for an experiment. I gave up when the customer informed me it usually takes 6 months to run a campaign (even a mini-one) and that was the only way to run experiments. An organisation often moves at the pace of its slowest department.

This brings us to organisational analytical maturity.

I will admit that this is a topic that is very close to my heart and mind at the moment (hence the idea to revive the blog from last year). I fundamentally believe that in order for an organisation to fully benefit from the advantages provided by analytics or eventually becoming data-driven, it is critical for all parts of the organisation to be pulling in the same direction and preferably at the same speed.So how do I define analytical maturity?

To me, the easiest way to understand how mature an organisation is, is to understand the kind of questions that the people within the organisation are trying to answer using data.

The range of questions where analytics can provide a good answer ranges from what has happened to how can we make this happen. For simplicity the analytical maturity can be broken into 4 stages.

Descriptive Stage

The descriptive stage is the first encounter many organisations have with data. It often takes the shaped of backward looking reports: what has happened? How many of item X did I sell last month? This is a stage most organisations will be familiar with.

Diagnostic Stage

After getting the hand of static reports, the next stage is the diagnostic stage, where hypotheses are formed. Questions are asked around “why” and often require further slicing and dicing to find potential answers.

Predictive Stage

The predictive stage is when the questions move from looking backwards, to looking forwards. While concerns about the future may have been implicit in the diagnostic stage, it is in the predictive stage where specific tools, methodologies and algorithms are employed to uncover what is likely to happen, and often how likely it is to happen, what are the drivers of the behaviour.

Pre-emptive/Pro-active stage

At this more advanced stage, instead of taking certain variables/inputs as given and trying to predict the outcome, the idea is to influence the variables and thereby cause a change in the behaviour/status… Nudging, Behavioural Economics, Game Theory are common strategies and approaches.

A simple example can illustrate the difference, the “drain the swamp”(6) example:

· Descriptive Stage: How many people voted against me?

· Diagnostic Stage: Why did these people vote against me?

· Predictive Stage: Who is that person likely to vote for?

· Prescriptive Stage: How do I get that person to vote for me?

It is too easy to underestimate how difficult it can be for people to climb through the stages of analytical maturity, some never get to the pre-emptive/pro-active stage.

I believe that usually people do not want to make their lives harder than it is, hence the best way to make people in various parts of an organisation more analytically mature is by showing them direct benefits to their own selves. It is about change management.

At eternity’s gate (7)

For organisations with people in departments who are only used to static reports or even who are so busy that they don’t look at the reports, making descriptive analytics visual is a natural step. To anyone who is interested in helping make reports relevant to people, creating meaningful dashboards and triggering people to think using numbers, I would recommend books by Stephen Few (8); I had the opportunity to attend a course by the author a few years ago, and would like to think I learnt a lot and I try to follow the guidelines as much as I can.

The great thing about this book is that the principles can be applied using most software, so you can start from today itself.

One of the more logical approaches to (re-)introduce the use of simple reports in an organisation is to take stock of existing reports, gather business requirements, and do a gap analysis. In parallel or even prior to that, it would be good to have special purpose pieces of work answering specific ad hoc business questions. When immediate needs are met, the focus can switch to future needs, the discussion can move easier to dashboard design and ability to drill, slice and dice.

Basically the idea is to use ad hoc analyses and visualisations to encourage people to think about data, and to use data to try solve their problems, moving from the descriptive stage to diagnostic stage.

One of the important aspects of the diagnostic stage is the culture of experimentation. Hypotheses can be formed, may be even theoretically tested, but true learning comes from actual experimentation, and this gets more important in the next phase.

Back to the future (9)

The move from backward looking to forward looking is a very important one. Creating hypotheses (as in diagnostic stage) can still be done without knowledge of statistics for example, but evaluating them and making inferences requires some statistical knowledge, so does the evaluation of the results of experiments. This is even more so when one moves into the realm of predictive analytics.

Why statistics? Well I believe that having working knowledge of maths and stats allows the understanding of many techniques used for predictive analytics. And I will, as usual, place my favourite data science diagram (10):

Advanced Analytics/”Data Science” is concerned about predictions, and as it can be seen above, knowledge of stats/maths is an important characteristic of “data science”.

Once an organisation is comfortable in the world of creating hypotheses and possibly testing them, the next step is to use predictions to guide the ‘best’ course of action. It is important to note that in order to maximise the impact of predictive analytics, the culture of the organisation must have evolved to one of experimentation.

Once the culture of experimentation is established, we have a learning organisation and can become data driven. Again, it is important that experimentation permeates the organisation, it is critical to understand some experiments will not get the expected results, and learning from them is the point, not learning is a failure.

Minority Report: A Beautiful Mind (11)(12)

Predictive analytics assumes that the behaviour variables are given; pre-emptive/pro-active analytics attempts to change the behaviour. This falls in the realm of behavioural economics, game theory, nudging, precogs(11)… Most organisations are not there yet, plus there may be some ethical implications (after all the swamp hasn’t been drained yet, has it?)

In sum, analytical maturity is critical to ensure the successful adoption of the more advanced tools of analytics/”Data Science” (to me AI is a tool); to paraphrase the article quoted earlier (4), people are not ‘malleable’, putty is. So as long as we are dealing with people, change management, bringing people across an organisation up the analytical maturity stages is important.

However, that is not to say that it is not possible for organisation to engage technological leapfrogging. One of the interesting aspects of technology is that you do not need to understand it fully to use it to make decisions. As someone said in the Global Analytics Summit in Bali last year (you can find a piece I presented there in a previous blog post (13)), “managers who know how to use data to make decisions will replace managers who don’t”.

Once a technology gets to the bottom of the through of despair in the hype cycle (14), what brings is back up via the slope of enlightenment is that it starts getting applied beyond the purely technical hype, real life applications are what make technologies reach the plateau of productivity.

In Sum

To me it’s our job as analytics/“data science” practitioners to help organisations go through the analytical maturity. What about new technologies to come you would ask? The answer is that if an organisation is mature enough, has become data-driven, it will naturally seek to adopt new technologies and be competing with data.

So to answer my friend, yes, if an organisation is not doing analytics, it can’t simply adopt AI. However, it is not necessarily take that long to learn and become analytically mature, as long as there is a framework and commitment through-out to do so. And I would like to add, I certainly believe in technology leap-frogging, I am betting on it.

Sunday, 2 September 2018

What is Value? Or how much should a "data science" project cost / "data scientist" be paid?

Singapore has been rocked by the Sing Health hack (1) and the fact that the government has been downplaying it saying the data was not valuable - no state Secret (2) - and the description by the ex-PM that people who do not make a million dollars annually are very very mediocre (3). My question is “so what?”(4)

What is the thing called value? What is the value of the data? Is less than a million dollar salary very very mediocre?

In this blog, I’ll take a stab at “value”.

SingHealth

First let’s take the case of SingHealth. You must understand that Singapore takes defence and security very seriously, the concept of total defence (5) includes cybersecurity, and some people were amused that the CEO of SingHealth is the wife of the minister for defence (6). But that’s not the point.

So what is the value of the data? How do you measure the value of data? To me this is a very subjective area. However, I assume everyone will agree that accuracy of data is important, even (or especially for) AI/ML.

How accurate is SingHealth data?

There was another story recently (7), SingHealth actually tagged someone as HIV positive whereas she was not. So my question is how many such mistakes are there in the data? Saying someone is HIV positive when the person is not is a huge mistake. In fact, the article states that, since the ‘victim’ was pregnant, the husband talked about divorce, abortion...

I am not slagging SingHealth specifically, but value of data is tightly tied to its accuracy, and there are some doubts over the accuracy of SIngHealth data, after all how many of us have looked into what is in our files at the doctor/hospital/clinic (especially those of us who are not medically trained)?

However, the government stated that only basic data such as name, NRIC number, age, gender was compromised. But the NRIC is the key to most (if not all) databases in Singapore. In fact some shops want to use the NRIC as their loyalty card, and I refuse to give that up. With this detail, someone can apply for a loan and make you responsible for it.

SingHealth is aware of the issue and has tried immediate mitigating factors; that kind of shows the depth of the problems (8)(9).

So what is the value of this data?

To me, there are a couple of other criteria that determines the value of the data: the use it will be put to and the skills of the person. So it depends who you are talking about. And this brings us to very very mediocre people.

The ex-PM basically called people who do not make SGD1m a year as being very very mediocre, and needless to say it created some noise in cyberspace, but was it warranted? Many Singapore ministers come from 3 backgrounds: Lawyers, Doctors, and Army officers.

The above charts from payscale.com (10) show the distribution of yearly salaries for these job roles in Singapore. Note that the army numbers are highly skewed since they include people undergoing national service included, and the data has some quality issues (salary of $72 for doctor), but the median is reasonable.

Basically, it is quite clear that people who make $1m or more are definitely above the top 10% in salary for their domain.

The ex-PM also argued (3) that ministers salaries need to compensate the ministers for the salaries they are giving up; and given that Singapore ministers salaries are quite high (11); hence it does make sense that if you aim to compensate people for the salaries they are giving up, to only look at people making S$1m yearly.

I am specifically avoiding questions/discussions around whether this is a good way to remunerate people serving the public. But if the aim is to be able to entice people at the top of their domains as measured by salary drawn, and compensate them similarly, plus given that the ministers’ salaries are close to 7 figures, saying below S$1m is not what they would be looking at, hence “very very mediocre” is acceptable in this context. (You wouldn’t consider an 18 year young man doing his national service as a ministerial candidate for example, hence the ‘segment’ you go for is not the whole spectrum but a small segment at the right of the distribution.)

Surprised that I think so?

I believe that people should be paid based on the value they generate.

If the salary that the minister-to-be is a reflection of his/her value in his/her domain, and if that value can be transferred to how much he/she contributes as a minister, then it is perfectly alright that the salary they receive is similar.

But the real question is how do we measure the value that someone generates?

That is precisely the great thing about being in the Analytics/”Data Science” space. The value you bring in a project can and should be easily measured.

When I took up my first contract more than a decade ago, the business sponsor for me had a choice between employing a new salesperson, or spending the money of an analytics guy, so my contract had targets just like sales people (but no variable income unfortunately). Hence the value I brought to the organisation – indeed that of my fellow analytics guys – was tracked and measured, methods of measurement, metrics all discussed, agreed.

These were extremely exciting times for me. In fact at the time I resigned less than 6 months into a new contract, I told my colleagues to tell my replacement that he/she could rest easy if he/she was paid around the same as I was since I had already justified my existence for the year and therefore his/hers.

This is why, whenever there is a “data science”/analytics project, I insist on having metrics that reflect the impact of the piece of work on the organisation it is being done for, whether it is in terms of savings (say for churn can be decline in number of churners based on past trends, or even monetised by spend – although that comprises of an extra dimension and gives more room to play) or revenue increase or market share increase, whatever is the KPI of the project sponsor; measurable and measured.

When I started in analytics more than a decade ago, we had to prove ourselves to sceptical business, hence we ‘manually’ tracked our impact to justify our existence and gain trust. I spent almost 4 years in that organisation, and we eventually set-up proper campaign tracking. Imagine my shock when I went back and found out that the organisation had stopped tracking campaigns. They did campaigns simply because they had the budget, and “use it or lose it”, without caring whether there we better ways of “using it”.

Some people may like this environment where you get to experiment without risk; but how would you know if the risk paid off, how good your ideas/hypotheses/skills are, if you do not measure the

outcome? How do you know the value you bring to an organisation? How would you know your value?

Value is not a measurement of input, but of output.

Once you have an idea of how much you will be able to contribute to the organisation, then you can apply RoI/break-even rules and determine how much it would be acceptable for you to charge, thereby delivering a win win situation.

The value of a project is a proportion of the value of the benefits the project generates for the client and that proportion is usually based on the typical RoI or Break-even period for projects the client undertakes.

This means that the same effort may generate less value to an SME than an MNC (in $ terms), hence the value you’d bring to an SME is lower than that to an MNC.

I recently had this discussion with someone who works closely with and helps bring innovation to SMEs. I think SMEs have similar problems to MNCs, albeit at a lower scale. While it is true that SMEs are less likely to have a full set of data to start work on, the analytical methods of solving the problems are similar. Furthermore, analytics is not as expensive as many people think it is.

I think SMEs have an advantage over MNCs, they are more flexible. Hence arrangements where the client pays a low base fee and a proportion of the value generated by a project/analytical piece of work can be done with SMEs whereas MNCs may not have that flexibility (neither would large consultancies who would have to account for revenue recognition risk and so on).

Basically to me it is very simple, tie what you are paid to the value you deliver to your customer, this is a very simple way of having win-win situations. And it all starts by knowing the value you bring which is based on measuring the impact of your work.

Similarly, as someone considering paying for the services of a “data scientist” or “data science team”, you should base the payment on the expected returns from the services received, and that starts by looking at the impact delivered in the past.

P.S. As I wrote this blog, a new ruling from the PDPC Personal Data Protection Commission recognises the value of the NRIC, and therefore restricting unwarranted use of it. (12)

Sunday, 26 August 2018

Business Internet Banking, 5 working days to unlock an account? yticoleV. Nudge nudge branch transaction red flag...

I had a very interesting experience last week and I thought it was worth sharing. Basically it is about OCBC and the push towards decreasing an organisation’s costs by either replacing humans by machines within their ranks, or shifting costs onto the customer (one of my pet peeves with supermarket self-checkouts), and especially how that push if being made and the possible impacts on existing staff, customers, and customer experience.

I had managed to lock myself out of my business banking internet account with OCBC. (Hey, what can I say, I love the fact that I have a personalised debit card... have always been a sucker for looks). What is important is that I am/was using self-service internet banking, apparently exactly what the bank wants. When I was locked out, I called the hotline, and was told it only operates during office hours.

The next day, I tried the hotline again, and decided to try and get the problem fixed without having to go to the branch. The superb robot couldn’t understand what I was saying. (I am not complaining about the robot, although I spoke in my version of English throughout (no mid sentence language switch which is horrible to deal with in speech recognition), but the simple fact is I couldn’t get what I wanted after trying a couple of times.)

You can stop laughing at my engriss please. The OCBC robot’s standards are way higher than mine and accent quite posh J

So I decided to take a quick walk to the nearby branch.

Anyway, after a not so small wait, I went to the counter and met a perfectly polite and helpful lady (Ms Lim if I remember correctly), she understood my problem, got the appropriate paper forms, ticket the relevant boxes, got me to sign and all was done!

Just before I left the counter, I asked, “around how long do I have to wait to have my access re-instated please?” and she replied “up to 5 days” and after she realised I was shocked she helpfully explained that they couldn’t do this on-site, and this piece of paper had to be sent to the relevant department for action, so it would take up to 5 days.

Ridiculous, don’t you agree?

Here you have a bank pushing its customers to use internet, but when they are locked out, it can take up to 5 days to reinstate access to the account. Now if like stupid me you thought internet access would mean you can always take advantage of the really fast bank transfers to pay for business expenses, you’d end up like stupid me with an invoice you cannot pay because you have to wait 5 days for the piece of paper to make its way from the branch to the relevant department (are they in internet banking department?) and for someone to take action.

I lost it.

I asked the counter staff to put herself in my shoes, and whether she doesn’t find it ridiculous that a bank that is investing a huge amount of money to replace these counter staff by ‘robots’ (1) didn't instead spend the money on making internet services for customers more seamless. (Unless I am one of the few idiots, if not the only one, who got himself locked out of his internet banking access – in which case I will hang my head in shame).

Anyway, the counter staff said she would talk to her colleague to expedite my case and I left the branch.

A few hours later, I received an SMS from the bank telling me my access had been restored, and an email, followed by a call by a human. Wow, speak of service! OCBC managed to turn an unpleasant experience into a surprisingly pleasant one.

So am I really writing a congratulatory blog?

I started thinking...

If it is possible to resolve the issue so fast, why wait for the customer to complain? Either the bank’s process were short-circuited to expedite my case as an exceptional case (can you do me a favour, I have this customer who is very unhappy...?) or the process does indeed take minutes but it is a deliberate policy on the part of the bank to make branch transactions as painful as possible.

The counter staff did ask me if I had tried calling in, and I said I did, but it didn’t work, the machine couldn’t understand what I was asking for. Was that a factor? So I did try the way the bank preferred, and hence I wasn’t made to pay the full price (5 days locked out).

This reminded me of an interesting article I read on how banks are pushing us towards ‘the cashless society’ (1).

Basically the key to making customers do what you want is “nudging”.

To quote the Wikipedia definition (emphasis mine) “Nudge is a concept in behavioral science, political theory and economics which proposes positive reinforcement and indirect suggestions as ways to influence the behavior and decision making of groups or individuals. Nudging contrasts with other ways to achieve compliance, such as education, legislation or enforcement.”

A classic example of nudging is illustrated below:

For people not familiar, public urinals often now come with a fly attached; the idea is that the urinator takes aim and thus spills less. The urinator is nudged towards doing the right thing. I think most of us would agree than a toilet where there are fewer spills on the floor is a good thing. It’s good for the urinator who is less likely to get splash back on his pants, it is good for the other toilet users who enjoy a cleaner environment, and it is good for the people cleaning the toilet (and the company that employs them) in terms of a less mess to deal with. Wins all around! Some organisations have even cashed in (2)

Nudging is another area that interests me, but again, I’ll hold off for another blog; reserve this one for how it appears to me, as an outsider and based on my personal experiences with and what I have read about OCBC’s latest trends. But for now what I am saying is that nudging is in itself and by itself not a bad thing; it can create win-win situations all round.

Let’s assume OCBC is nudging me away from branches to using internet banking.

You could argue that I can’t complain about the fact that the hotline was not operating is a non-starter since I wouldn’t be getting 24 hour service if I was using a branch.

But that ignores the issue of expectations. We have been sold the idea that using internet banking is superior because it is 24/7. But if support is not 24/7 is internet banking for businesses really 24/7? I don’t think so.

Plus, personally, I much prefer dealing with a human than a robot, face to face rather than via phone. But that’s just me. So if I am moved away from this personal touch, there needs to be a compelling value proposition.

What else is OCBC doing at the branches?

OCBC is planning to replace half of the teller staff (3) by machines by 2020 (4). But it is not firing people thankfully, but retraining them as highlighted (5).

There are a couple of things I’d like to highlight from the articles.

I object to tellers being called relatively low-value added services: “The bank said the tellers would move into roles that allow them to take on "higher value-added" tasks that require decision-making or physical verification.” What is the measure of value? That’s a blog post in itself, but to me being able to get my issues resolved and queries answered is very valuable, and it’s a definite plus if it comes from a human. I am a human, and I relate better to humans; I ‘feel’ a brand if I am dealing with a human representative of that brand, I do not feel anything if I am dealing with a machine from that brand. So to me there is a tangible (issues that are not out of the box, specific preferences, services that require more stringent verification as pointed out by OCBC themselves) and intangible (the human factor).

Another fact that I find interesting is that, even as the new generation machines developed by OCBC at the cost of $14m and over which they have 5 year registered design license are ‘future-proof”, they are retraining tellers into advisory (which is great) and where verification is required. This latter piece puzzles me.

To me, anything that is used multiple times is something that can be automated. For example, if to release funds requires verifying the documentation and signatures, there are basic steps that are always followed, and even for signature verification, technology enables you to score how different a signature if from another. And if there is anything you want when deciding whether signatures are similar enough is some objectivity to decide the “enough”.

How about advisory then?

Well I am sure that, especially given the evolution of the role of counter staff from pure customer servicing and transactional role to having some sales component, many tellers will be able to make the switch. However, OCBC has also debuted ‘robo-advisors’ a few days ago (6). And this is not a one-off; it is a direction that OCBC has taken for at least a year (7).

They recently launched service targets younger investors; well heeled clients already have relationship managers, so what market will the newly trained advisors target?

I applaud the efforts by OCBC to re-train its staff, i think it is too easy to replace people without a thought, and putting resources into improving the skills of existing employees is great. Furthermore this is in-line with the efforts by the government to upgrade the skills of the workforce, for example skillsfuture (8).

However, I believe that an organisation should not cry victory at the beginning of a programme, but should only do so once there is a victory. OCBC claims that all tellers who are being replaced by the new machines will be re-employed in different roles. As the articles point out (4) and (5), OCBC did not disclose the current number of tellers, neither was the number of affected people. I believe that OCBC should follow up this PR exercise to put its money where its mouth is and disclose the number of tellers affected now, and the number from this group still employed in 2020. That would be a true win-win.

I hope OCBC can truly create a win-win-win situation where customers get better service, employees remain gainfully employed and enjoy their roles, and the organisation naturally makes more profits. Automation, Analytics have the power to create all around win situations, and organisations who create such situations would certainly emerge as winners in the medium to long run.

I believe that automation, finding better ways to serve customers, embedding analytics at all levels of processes of an organisation (effectively making organisations data-driven) can enhance the ability of people to do their jobs, make them more productive, decreasing the costs of the services and thus allowing more people to enjoy them. It’s about decreasing costs to decrease prices and make things more affordable to more people, not decreasing costs to increase profits.

Monday, 6 August 2018

"If you don't have a PhD, don't call yourself a data scientist"

“If you don’t have a PhD, don’t call yourself a data scientist”; with these remarks the government linked person set the stage to explain his views via a presentation on AI, and the problems implementation of AI suffers from today.

To flesh out his argument, he argued that only a PhD gives the rigour and access to large enough data to play with to become a data scientist.

A very interesting point of view from someone linked to the government.

Not everything he said was that controversial, at least to me.

The presenter used my favourite diagram for data science, the ‘Drew Conway’ diagram (1), acknowledging the importance of subject matter expertise. Data Science is a balanced combination of “Substantive Expertise” or subject matter expertise, “Maths and Statistics Knowledge” and “Hacking Skills” or IT skills.

Furthermore, the presenter also mentioned how hard it was to find all 3 skills at a required level in 1 person and also spoke of data science teams; or like what I say: “Data Science is a team Sport”.

Also the presenter was at pains to point out that a 3 month course in data science does not make you a data scientist, so even if you are an English Literature PhD, or hold a PhD in Astro Physics, a 3 months data science course does not make you a data scientist; it takes years.

I am on the wall on this one. I think “data science” like every subject needs practice, and while a 3 month course will most likely not give you enough experience, it doesn’t have to take years and years. Any expertise is gained through practice.

Furthermore, the presenter is a proponent of open source, and advises everyone to eschew classes and learn online instead, pay tens of dollars rather than hundreds. I am all for learning online, have taken classes from Data Camp (2) where I learnt a lot, as well as from Coursera (3).

But where it gets really weird, and please remember that the presenter is linked to the government, he then went on to “sell” is classroom courses, of around 3 months, and hopes he can provide some practical experience.

Unless he is targeting only PhDs as students, I find what he is saying quite contradictory...

The reason I mentioned English Literature and AstroPhysics is the presenter further mentioned that one of the reasons why the country may be finding it hard to find “data scientists” is the fault of HR departments. They are looking for a unicorn with degrees in computer science (let alone PhDs). The advice was that they should loosen the criteria and accept people from different disciplines and who have taken the online courses...

My view is not that dissimilar. I believe in passion and without knowing anything about a person, I would say that an engineer is more likely to make a good “data scientist” than a Statistician or a Computer Scientist. The reason is that to me, “data science” is about delivering value and the passion should be to solve problems, the end, not the means – AI/ML/Stats...

Then this goes back to the PhD question. Do I believe you can’t be a “data scientist” without a PhD? Well, it may be self-serving since my profile states “data scientist”, but no, I do not believe a PhD is required.

In fact, quite a few organisations have found this. Basically, people with PhDs are great at their own domain, but “data science: requires a multitude of skills that they may not have (for example subject matter expertise, or statistics for computer scientists, or IT skills for Statisticians) or may not want to engage in: the ‘dirty’ work of cleaning and preparing the data. Hence the organisations whose “data science” department is staffed purely by PhDs find it very difficult to get a decent RoI. (results, results and results).

While I am at it, I will also mention that another way that organisations get their staffing wrong (hey, may be that deserves a separate blog, but here goes) is in the fact that some “data scientists” delegate the data cleaning and preparation to “data preparation” or “data engineers”. It gets worse when the latter do not have a clear career path to the former, like sous-chefs becoming chefs... Data preparation should be done with a purpose, and unless the high and mighty “data scientist” can communicate the purpose effectively and in great detail (probably also requires some EQ), there is a risk that the data preparation will not be that fit for purpose.

Basically I believe that data cleaning and preparation is part of the role of a “data scientist” especially since “data science” is by nature iterative and iterations may involve obtaining and preparing data that was not included initially.

Quite a while ago I did an easy to understand view of the work of a unicorn (“data scientist”); as you can see, data preparation and transformation is part of the process. I can understand that someone who is good a solving business problems may not be very good at getting data in the most efficient way from various systems, or writing production ready code, but surely preparing data is part of the role after all, most people will tell you that this is 70%-80% of the work...(4)(5)

So why I am upset enough to write this blog?

Basically I believe analytics/”Data Science” has the power to unlock enough value to create win-win (win) situations (organisation, customer/society, and (consultancy/ vendor)), and getting the framework for data science is critical in that regard.

From the presentation I attended, it would seem that the government has got some things right, some wrong, and some contradicting each other. I do hope they sort things out; unfortunately, the Peter principle may be at work.(6), or may be it’s HiPPOs (7) or both since there often is a high correlation between the two (A hippo named Peter...)

Actually ya, this might be the topic of my next blog, although I am also itching to write about AI/Automation and retraining...

P.S. Did I mention that the presenter said that SLR is part of AI?
(See it pays to read all the way to the end... now please clean the coffee from your device)

1. http://drewconway.com/zia/2013/3/26/the-data-science-venn-diagram

2. https://www.datacamp.com/

3. https://www.coursera.org

4. https://www.infoworld.com/article/3228245/data-science/the-80-20-data-science-dilemma.html

5. https://www.xenonstack.com/blog/data-science/preparation-wrangling-machine-learning-deep/

6. https://www.forbes.com/sites/roddwagner/2018/04/10/new-evidence-the-peter-principle-is-real-and-what-to-do-about-it/#681be0a81809

7. https://whatis.techtarget.com/definition/HiPPOs-highest-paid-persons-opinions

Wednesday, 25 July 2018

Prudential Singapore (Insurance), Agents and Digitisation, some thoughts

A recent article (1) highlights Prudential’s agents’ unhappiness with ‘digitisation’ of their business. The 2 main questions are whether this unhappiness is founded or not and what impact this has for the future of the insurance industry, especially in line with the emergence on Insurtech.

The complaints centre on the fact that Prudential is making available to customers via online channels the same products that agents offer to their prospects, including the basic/workhorse/most popular ones. Prudential management has been quick to state that the agents returns will not be affected.

So what is the most likely scenario?

Just think of it this way, if, without premiums going up, Pru is able to finance a digital platform, maintain it, and still pay agents full commission on sales they have nothing to do with, then it must mean that the premiums Pru customers are paying today are too high since they can pay for the extra costs of the platform...

Assuming that is not the case, so why would Pru be losing money? It is possible that Pru believes that agents do a terrible job of cross-selling products, and therefore the platform would pay for itself based in sales of products they wouldn’t otherwise have sold.

I worked in the insurance industry for a while, and of course there are customers who are more or less abandoned by agents and left to their own devices. Agents most often focus their efforts on a few customers whom they believe are more likely to buy or people they are closer to. But this doesn’t mean that simply having a platform where these less served customers could reach out to pru and make purchases would generate enough business (minus agents’ commissions) to fund the platform.

I would think that what agents are really worried about is the ‘customer ownership’. Again from my experience in the industry, agents believe it is them who own the customer relationship, not the insurance company. When an agent leaves, there is a process to deal with these ‘orphan’ accounts; they are basically reassigned to another agent, usually from the same agency.

It is possible that this may change, with the insurance company choosing to serve these customers via the platform rather than reassigning to agents, effectively having the insurance company now own the customer.

Furthermore, the company could enrol customers directly via the platform, removing agents from the picture for simple products, and using a pool of trained agents for products that require it.

In fact agents could fear that their role in the insurance industry will change drastically, and there may be a need for fewer of them, to act more as product specialists than engaging in the whole range of activities they cover today.

Does that mean that digitisation is not a good thing for agents?

The answer is no.

Digitisation is important, but it must be part of a transformation of the organisation towards being data driven. I believe that, in order for an organisation to become truly data driven, all its component parts have to be more or less equally data driven – like a chain whose weakest link determined its strength, the least analytically mature component of an organisation determines its overall maturity and how far it is from being data driven.

I am an outsider, and I have no clue how analytically mature Pru is. However, as a customer of Pru, I get to see some of it. For example, I have received calls from Pru agents/agencies I have had no prior contact with, asking me about my policies with them; this should not happen. Usually only my agent or his agency should have access to what I purchased, not other agents/agencies. To me it means their databases are not that secure. The best part is that when I told my agent to complain to HQ, he basically said it happened all the time... So I am a bit skeptical about how data driven the organisation is.

Anyway, that being said, I think it would have made more sense for Pru to work together with their agents, rather than take the approach they have, which leads to some legitimate fears among their agency force.

Pru itself has said that Singaporeans are under insured (2), so if despite the personal attention that agents can give to their customers, these are still under-insured, what makes Pru think that the customer would by himself/herself buy from the platform?

Singapore is a relatively mature economy, insurers have been here for ages, and even then people are still under-insured. People do not decide to under-insure themselves, it’s most likely a matter of education, and that’s where the agent, if properly trained, can bridge the gap. People usually learn better in small settings; just like another Singaporean institution, the private tutor helps students in 1 on 1 or small class settings rather than in giant classrooms (or even worse on youtube videos...)

Basically, to me, agencies like Pru, still need their agency force to be on-board, they have a crucial role to play. (And my friendly agent would hit me with his hockey stick if I said otherwise). Education, and really doing proper financial planning for their customers is key. Sure there are stuff that can be bought ‘over the counter’ like travel insurance and are hugely successful (4). But not every insurance product is in that category.

Does that mean then that the platform is useless?

Of course not. I have even written how the whole insurance process could potentially be run on the cloud (3), so I get the importance of digitisation. To me the platform and analytics should support the agent and enable him/her to serve his/her customers better.

Properly built analytical models should give an idea of not only the potential need of a customer, but also the right timing. The simple way of working with agents to serve customers would be to:

Inform the agents of whom, among their customers, is a good target for a specific offer now
Allow the agent to personally contact some of the customers he/she wants to, get a commitment accordingly and a simple easy to use feedback loop.
Contact the rest making reference to the agent if customers want to take action.

As a Pru customer, I have received SMSes that make me offers, sometimes they include my agent’s contact details in case I want to follow up. Well, none of their offers interested me; and the best part is that my agent knows that. We meet up once a year of so, and go over the policies, life… and he knows I do not need more coverage for now.

Is that all there is to digitisation or becoming data driven?

No. Far from it.

Insurance companies are made up of many parts, and the selling is only a small part of it. For example, once the customer decides to buy a policy (whether it is via a platform or an agent) how quickly is the policy sent back to the customer for take up?

Ideally this should all be straight through processing, especially for existing customers whose KYC (Know Your Customer) is still valid, all that’s needed is to get the customer to confirm the validity or make necessary updates. Then the insurer should have the validated data of teh customer and can proceed with the application proper.

Electronic forms are the best; while the platform should have this feature by default, it takes little to provide agents with the equipment necessary, a tablet for example (issues like online or offline are dependent on the specific market) with in-built checks to ensure all information necessary for an application is available and in correct form.

Then with correct use of technology, standard cases can be approved almost instantly; I have reproduced a diagram from my blog (3) below:

In such a case, underwriters need only work on exceptions. Using technology and analytics in this way makes the policy issuance process much more efficient, customers get their coverage faster and agents can focus on educating and selling rather than to have re-works, and the organisation gets policies in faster and or good enough quality. A win-win-win situation.

Similarly the claims process can be automated, again I would refer you to my earlier blog (3).

So the question is, has Pru done all this?

Well it has tried (5), but while this is a good beginning, this is far from customer centricity. Basically Pru gives discounts on premium if no claims are made. The interesting thing is that customers might choose the game the system, pay for small ailments rather than claim, enjoy lower premiums, and hit the insurer on the big ticket if any. I wonder whether behavioural changes have been taken into account in pricing: no claims doesn’t necessarily mean healthy, and incentivising gaming of the system is not usually a good idea. People are not stupid. Instead of piecemeal attempts, customers should be engaged, the organisation become customer centric rather than organisation or product centric as seems the case above.

And I didn’t even get into data driven customer centric product design...

Sorry if this sounds like Pru bashing, it wasn;t my intention. But I see this episode as a case of an attempt at digitisation without looking at the big picture of being data driven and customer centric. I am sure Pru is not the only insurer in this situation.

As long as there is no effort by insurers to become truly data driven and customer centric (and this is a process, not a single big bang), they will be vulnerable to more nimble technology driven players, picking off specific profitable segments, and that would be a double loss to the traditional insurers.

1. http://www.marketing-interactive.com/prudential-faces-unhappy-agents-as-it-pushes-digital-transformation-agenda/

2. https://www.prudential.com.sg/en/our-company/press/2016/20160811/

3. http://thegatesofbabylon.blogspot.com/2018/03/re-imagining-insurance-company-do-you.html

4. https://www.todayonline.com/business/singaporeans-more-picky-travel-insurance-options

5. https://www.straitstimes.com/singapore/prudential-to-tie-rider-renewal-premiums-to-claims-made