BIG DATA AS GOLIATH, ANALYTICS AS THE STONE AND DAVID AS HUMANITY A use case for credit management: customer value before customer risk
Table of contents
Executive summary
The meaning of big, data & analytics
Analysing big data analytics (tooling)
Intermezzo: BI without G and Big data analytics as the essence of BI
Three aspects: data mining, analysing trends & data discovery
The fourth aspect: gathering data gathering methods
From human to machine? The human element!
Intermezzo against Darwin
Use case: customer value before customer risk profiling
Big non-data driven conclusion about the terms and conditions
Executive summary
(14-6-2017). This article will show that Big data analytics will not make the role of the credit manager obsolete, but, on the contrary, will increase in importance. Credit management itself will increase in importance becoming a more integrated part of the whole customer journey.
Whether the individual credit manager will survive big data analytics is up to the individual. To survive means to embrace big data analytics and to embrace means to have a certain skill set. That skill set mainly exists in being able to ask the right question to the data and being able to work in cross border teams (united in making the customer journey as best as possible for the right kind of (paying) customers).
To understand this a clarification of what big data analytics is is of the utmost importance. The difference between Big data analytics and Business intelligence is discussed and concepts as ‘data mining’, ‘trend analysis’, ‘data discovery’ etc. are clearly defined and examples are given.
A very concrete example is given why human subject matter expertise to ask the right questions is necessary to get the right answers from big data analytics. This subject matter expertise is from a human. The expertise is further differentiated by cleverness, smartness and intelligence. It is about becoming a clever credit manager or rather a credit marketer.
(17-6-2017). In order to really appropriate the subject ‘Big data analytics’ an analogy is made between a certain hopeful interpretation and the hopeless idea that humans descend from apes (or a common apelike animal ancestor of both apes and humans). Something is at stake here: you are either a talking medium raw steak of a(n ensouled) human being. This may magically transform the subject of ‘Big data analytics’ into a way to think about the riddle of live!
The meaning of big, data & analytics
(14-6-2017). One can say that ‘big’ is a word expressing a subjective perspective on a volume. The fact that all the ‘stuff’ I have to do is big and so big that I cannot finish what I have to do is not what is meant in ‘big’ within ‘big data’. To be more specific ‘big’ in the context of ‘big data’ has nothing to do at all with a subjective perspective. When I cannot finish my to do list it is implied I have an idea what to do and how big my big task is. In big data ‘big’ means we have no idea how big big is.
There are many tribes whose counting goes like this: 1,2,3,4 and many. How big is many? Well many is a lot of ‘manys’, many is really quite big. You might think kind of counting is primitive, but if you think you must conclude that we are primitive too: 1,2,3,4….infinite. (For more insight on ‘counting till infinite go to: ’https://www.dhirubhai.net/pulse/difficulty-clever-person-understand-business-value-james-roolvink). How big is infinite? It is big! In Islamic prayer one does not say God is many or infinite, but one says ‘God is big’. Anyway the ‘big’ in big data is an objective big volume of data.
‘Big data’ is an abbreviation of ‘Big data analytics’. We have discussed the big-part and the big part is identified with ‘volume’. Gartner identifies variety of the data and the velocity of big data processing (the calculation power) in order to legitimately speak of ‘big data (predictive) analytics tooling’. These three ‘V’s’ perfectly match the three words in ‘big data analytics’:
This is not just a conceptual theoretical clarification of the concept ‘big data analytics’ or rather the theory allows you to act very pragmatically in choosing (or selling or innovating or making) a big data tool. You should focus at least on two aspects: velocity and variety.
Analysing big data analytics (tooling)
Big data analytics (‘BDA’ – before dying acceptance) is not only big for the big is just about the volume, but is also really complex, that is rich (and this full of opportunity). To analyse the ‘velocity analytics’ part we need a new Venn diagram.
That Venn diagram cannot be included in the ‘analytics circle’ of the previous Venn diagram for both ‘qualitative’ variety of data and ‘quantitative’ volume of data are addressed.
Volume is simply addressed in ‘data’ and variety is addressed by the ‘theoretical concepts’ for these are criteria to use certain data sources. A theoretical concept may be ‘customer revenue’ or ‘customer profit’. The point is not that we will see that if we look to the data with two different theoretical concepts we will extract different useful information from the data (or none at all), but that the theoretical concepts decide what kind of data (sources) are to be used in the first place! (A priori) theoretical concepts decide what data sets are used that need to be analysed. After analysing the data sets it is quite possible that a priori theoretical concepts need to be refined, abolished or that new theoretical concepts are discovered a posteriori in order to use these newly discovered theoretical concepts a priori in the next gathering and thus analysing of data. ‘Big’ is big intelligence gathering!
A definition of the Greek word ‘logos’ (that is minimalized into ‘logic’, and ‘X–logy’ (X = psycho-, theo- etc.) as an a priori method) is ‘that (divine active force) that gathers’. It is closely aligned to the Greek word ‘krinein’ meaning to actively be able to make useful theoretical distinctions in perception and from which our word ‘critical’ is derived. We have to be critical when using big data analytics tooling’ in order to gather our success stories.
Due the complexity the ‘we’ is an interdisciplinary team consisting of people from different departments within the organization and/or from people external to the organisation.
Intermezzo: BI without G and Big data analytics as the essence of BI
One might wonder what the exact difference is between BI (Business Intelligence) and Big data analytics for BI also has the intention to improve business processes and predict future business. Even if BI can predict, if BI is the analytical tool to predict it can at most say that something will happen and what will happen (to specify the ‘something’), but BI cannot find out the reason why. A familiar distinction between BI as an intelligent reporting tool that reports about the past and Big data analytics being able also to report about the future, that is predicting, does not hold. With BI one can certainly extrapolate from the past to the future, but it is impossible to find reasons why the extrapolation is valid.
However BI and Big data analytics form a continuum. Best to say that Big data analytics is the essence of BI, that Big data analytics was always the ideal of BI and that Big data analytics is the realisation of that initial ideal.
Three aspects: data mining, analysing trends & data discovery
In ‘big data analytics’ we come across three main areas, aspects or terms, namely ‘data mining’, ‘trend analysis’ and ‘data discovery’.
Data mining = researching data from the perspective of a specific question on the basis of an already known causal relation. The relation as relation is researched more thoroughly.
Example: what is the risk of a high DSO in combination with paying via credit card? In this example the emphasis is not so much on the causal relation, but on the concept of ‘risk’. The causal relation is already known.
Trend analysis & predictive analytics = searching for correlations that have a high probability as relations to be causal relations.
Example: are the day's sales outstanding more dependent on the age of the customer or the medium in which the customer communicates pro-actively with the organisation? (One can guess that the medium and age already correlate and that this is not a good question. If that is the case the question really concerns the causal relation between the medium of pro-active communication and the days sales outstanding. If communicating via e-mail (rather than calling the service centre) is a cause of a higher DSO the risk profiles of customers may be adjusted with this new information or a new risk profile may be added: ‘If customer communicates via e-mail risk is X% for a dispute, Y% for a higher DSO’. On the basis of these risk profiles credit managers take affirmative (automated) action).
Data discovery = searching (blindly) for any causal yet unknown causal relations.
Example: connecting all kinds of (external) data resources like an historical data base of the weather, the TV guide etc. and see if (and in what way) the weather influences the DSO.
The fourth aspect: gathering data gathering methods
The idea is that ‘data discovery’ allows you to extract new theoretical concepts from the data. The ideal to start without any theoretical concepts that regulate innovation in data discovery and general big data analytics tooling is really good and although that ideal will be approximated asymptotically it will never be realised.
Of course there is a reciprocal relation between theoretical concepts and the data…
It is an illusion on the basis of a dream on Artificial intelligence. Even all the cleverness build in Artificial intelligence is made by humans and the building stones of this cleverness are theoretical concepts (maybe not so much on the subject matter, but at least ‘second order’ theoretical concepts on learning, on how to learn a subject matter (and that is still dependent on a certain knowledge of a subject matter).
A better visualisation that expresses the idea that in the reciprocity, between (raw) data and general theoretical (cooking) concepts, the theoretical concepts are more primary in order to get specific information (tasteful meals) is this:
With ‘theoretical concepts are more primary’ is ‘paradoxical’ meant that one beginning on a circle with two beginnings begins befor the other beginning begins.
That is the reason (are call to action if we accept the paradox) why a fourth aspect should be taken into account, namely ‘data gathering’. Data gathering on the basis of a priori theoretical concepts. It is an underrated aspect caused by not being clear about what the essence of big data analytics really is (in essence). Not being active in data gathering (within you own organization) and the basis of preconceived a priori theoretical concepts means you will take the quantitative amount and qualitative sort of data for granted as if the data you have for analysing is just an a priori fact. Data gathering is a very active and intelligent work.
From human to machine? The human element!
We luckily see there is ‘balance in Force’ concerning the human ‘factor’! (17-6-2017, Diemen). An aphorism can be that we need more advanced data analytics tooling in proceeding from data gathering to data discovery if we resign the task of discovery solely to the big data tooling. Leaving that solely up to the tooling is believing that a thousand apes behind a typing machine for a billion years will produce a poem with the quality of William Shakespeare (1564-1616) or William Blake (1757-1827).
Intermezzo against Darwin
Dream on dream on. There is just a categorical, an infinite, difference between apes and humans. Even if our genes match for more than 99% that 1% difference would make all the difference for we must take the whole into account and if one part is different the whole is different. Even between humans there are categorical infinite differences even if it seems we are all equal. That equality between apes and humans and between humans is based on general commonalities, the sharing of certain common features, like the fact that our bodies exist of bones, flesh and blood.
Apes transforming into a Williams by typing, that is apes evolve into humans is believing that raw data by sheer machines will be cooked into information (correlations and causal relations). Even if the raw data is cooked by machines one has to realise that raw data is already cooked for the data as data is already chosen by a human (‘cook’) and that these machines are made by humans. To express this differently: it is less untrue to say that an ape is already humane, an ape is a proto-human than to say that a human is in essence a highly advanced ape (on the same page as the latest most advanced smart phone).
The idea of genes does not match the truth that the part and the whole resemble each other and even that the part can steer the whole. Genes are like Lego-blocks and if you have build a castle with Lego-blocks the castle will not change if you swap all the Lego-blocks, all the parts, with ‘other’ Lego-block with the same features (in colour, volume and size). The swap is with impersonal other ‘sames’ making the castle, the whole, impersonal. The idea of genes is a scientific atomistic idea that opens reality in a certain way, but at the same time closes us off reality.
All thought Nietzsche (1844-1900) believed we were sick (reasoning) animals he had a personal argument against the idea of development in history, thus an argument against Darwin ((1809-1882) and himself), and it goes something like this: I am not dung (manure, fertile droppings of excrement) for future generations.
Nietzsche intended and longed for the soul, but his brain could not grasp the idea of the soul and he became mentally ill, not a sick animal, but a sick human. That idea is embedded in our constitution in the idea of an individual having individual inalienable rights. We are not impersonal persons, but really personal persons. Our personality is not a fake illusion.
End of the intermezzo
In resigning the task of discovery to both a human and (big data analytics tool) machine we need a machine, a tool, that is able to verify a the validity of hypothesis’s very quickly. That is a tool that is fairly simple from the perspective of calculating itself what may be useful correlations and causal relations.
Use case: customer value before customer risk profiling
(14-6-2017). Imagine a success story with big data analytics in which you were able to make in a very smart way new risk profiles of customers defining the risk they will become a bad debtor. You may have discovered that people who pay immediately have a higher risk than people who pay just before or after the pay date. You may have discover the reason: honest people with debts want to pay as fast as possible to be sure they paid because they are not sure if they wait a little longer they can still pay.
Now image a much greater success story in which you as a credit manager cooperated with marketing or a customer journey officer. From a credit management perspective certain customers are no risk at all because they pay on time and pay for more than an average amount of products they use from your organization. Having access to the whole customer journey data you may cleverly discover that some of these ‘loyal’ customers with no to low risk are actually not profitable! You may discover they are draining the energy and resources of your organization by calling the service centre to many times with questions that they could have solved themselves if they would have gone to the frequently asked questions (FAQ) page on the website.
As Marcel Wiedenbrugge in his ahead of his time (and for the sake of justice soon to be) future classic ‘Customer profit hacking’ starts with a very basic theoretical (business) concept in credit management, namely the customer relation existing in the added value for the customer from a supplier and the value of a customer (in terms of finance, marketing value etc.) for a supplier. If we had started with that simple very practical a priori theoretical concept in data gathering rather than to start with data mining with the a priori theoretical concept of ‘risk (profile)’ our success would have been bigger. If we had started with the (basic) beginning our continuation from data gathering to data mining it would have been better. The gained risks profiles would be more relevant.
This example shows that human subject matter expertise, or plain common sense, is the basis for successful big data analytics. With cleverness, starting from the beginning, the fundament, you can make smart ‘smarter’. Integrate the customer journey in credit management and become a credit marketer.
Big non-data driven conclusion about the terms and conditions
We clarified ‘big data analytics’ in terms of the terms volume, variety, velocity, mathematics, AI (artificial intelligence), statistics, data, software, theoretical concepts, computer science, data mining, trend analysis and data discovery.
The conditions to use these ‘terms’ are to start at the beginning: data gathering. In order to gather we must have an a priori pre-conception about what not only what we are searching, but what we want to find. We must already have theoretical concepts of a certain domain. We must already have a kind of subject matter expertise. We need the humanity, humans, YOU!
The idea of the materialistic empirical Anglo-Saxon school of philosophy is wrong that we can start knowing, perceiving and investigating without any a priori concepts (or rather knowledge) or skills (that imply knowledge). No tabula rasa! How ironic that the Anglo-Saxon cultural area should learn this via technology (for I would guess big data analytics originated in that culture) and by technology some refute their own cultural heritage. In the next article I would will go a lot deeper into that subject from the perspective of Immanuel Kant (1724-1804).
For now happy hypotheses making!
Kind regards,
James Roolvink member of the Pulse DC team: bringing big data analytiscs to credit management