Information Systems in the Consumer Industry/General processes

A case study of reengineering of information systems – industry — A case study of reengineering of information systems – retail — Introduction to methodology — General processes — Industrial processes — Retail processes — Conclusion — Appendix A — Appendix B — Bibliography

Let us now analyze the processes which exist in both methodologies.

Ability to perceive and categorize customer needs

In this process, the system is the environment “company – potential customer” where the customer is the owner, more or less conscious, of his needs. The goal of the process is to share this information with the company.

Why should we try to perceive the need? The answer lies in an agnostic approach on other people needs. In my opinion a company role is not to believe to be the truth depositary but to be the instrument for maybe extracting and solving customers requests.

Who acts in the process of need perception? Three kind of people act: who has the need (the customer), who collects the need, who transmits the need.

Feeling the need, the customer is the person who explicitly raises the request (demand driven) or who owns the hidden need who is to be extracted (supply driven).

We can categorize our customers into:

Well known customers, person we know via a history of relations (fidelity cards or so); among these customers we care particularly about customers who have shown to be statistically significant in terms of representation of the whole customer set
Unknown or random customers
Potential customers: among these we could include people belonging to social groups which we consider target for our product, people looking at shop windows or people coming and not buying anything (prospects).

It is obvious that there are different marketing strategies on the three groups: in the first case we want to keep our customer and increase cross-selling or up-selling. In the second situation we want to make our customers loyal, in the third case we want to reduce entry barriers.

Collecting customer needs: I think we can individuate three kinds of people, the first one is someone whose job is to “feel” the market in its wider sense (society, various environments), sometime they are called “cool hunters” or scouts. The second group is formed by the agents who work together with retailers: their view is wider than the single retailer but is narrower and more precise than the one of the scout. The last group is made by sale executives who are actually in contact with specific customers so they have a clear, definite view of a limited request.

Transmitting the information: it would be easy to say that these people are the same that collects it, but it is not true in my experience. This function is often felt as a secondary job so it is delegated to another person, an assistant, done in the spare time or even worse hidden as felt part of the person assets. This attitude implies a great loss of quality in the information transmitted to the company.

What is a need, what does a customer looks for. Let us start by classifying this dimension on three values:

Defined needs: specific requests of a definite object (raincoat, wind jacket) maybe connected in a new environment (e.g. long gloves for snowboarders)
Semi-structured requests: more general than the former one, yet not very precise. One possible example could be “something like that jacket but the fur around the hood” or “slightly longer or shorter”. This kind of requests, mainly regarding the product, are often expressed in a sign, non-verbal language so they are very difficult to collect and transmit
Geographical or context needs: particular colors related to a culture or a “glamour” trend better than an understatement. These are examples of customer needs difficult to receive and use but very rich in informations.

In this dimension we have actually analyzed only product needs, we can use the same detail level also for the other levels of needs (internal esteem, social and self-actualization needs). As far as I know there is no research going on this topic which seems very interesting and largely unknown.

Where: about the place the need occurs. There is definitely a relation between the “where” and the class of needs (defined, semi-structured or context); we can recognize:

Shops: these are the places where there is an explicit interaction with the customer regarding well structured needs. This is a very important information asset
Shop windows or shop routings: we are talking about potential customers who do not interact explicitly, this is an important patrimony of informations fairly unstructured and difficult to catch and categorize
Competitors: we are looking at customer needs through what the competitors are offering; it is like looking at the solution and trying to guess the problem. This becomes important when competitors have access at context we cannot get into, for example far away markets
General geographical distribution: the most general and also the most difficult situation to analyze; as an example you can think about the fact that there is no accepted taxonomy in the fashion business.

As far as the definition of the collection and storage of data, the problem is now shifting form technology problems to the definition of an accepted semantic of data.

How. To understand how we can perceive customer needs let us consider how these needs show out. We can think of three kinds of externalization:

Conscious personal needs like an explicit request to a sale operator
Unconscious personal needs like taking a garment in my hand, looking at it and putting it back
Social needs like a new “fashion”.

How do we perceive these needs: for each kind we must think of a way to define the elementary information (the state) so we need

Instruments for sales support, both real and missed
Instruments for context analysis
Instruments for competitors analysis
Instrument for social trend analysis.

Specially talking about the first process, we must be very careful about the easiness of use and the precision of the instrument as the actual operator sometime works in a “normal” situation and sometime in a changed environment like “sales” or “rush times”. Another topic to be considered is related to the significance of the data we are collecting:

Does an observed customer behave like an ignorant one?
The data I am getting are biased by the collection process?

At the moment I think these can be considered smaller problems compared to the fact that now we have no data.

Last but not least there might be a problem on context bias: if we collect data in a luxury shop we shall have a “luxury” view of the problem, same thing for a low income customer environment. This obviously becomes a problem in inter-class analysis while it is not so bad if the company target is parallel to the sample set.

When do we collect informations and do we use them?

Using the data schema we just defined, we can say that

Geographical or context needs have a longer time horizon than the specific product; could be one or two years
Semi-structured requests are looking mainly a few months in advance, from one season to another; they seems a slightly shorter “time to market” of what is accepted today
Defined needs are typically immediate in their nature; the ideal would be to answer as soon as the need arise.

Data collection time is definitely related to the nature and horizon of the need but we can post it, more or less, in the sale period and consider a large period dependence on the level of the signal.

Usage is again related to the class of need and has a strong period dependence the shorter the horizon is.

How many signals should I collect and how significant must they be?

This is a three sided problem: a problem of method, a problem of measured content and a problem of reference value.

Let us face the method problem: if a customer exclaims “beautiful” in front of a garment the statistical variance of the variable “beauty” is equal to the value of the variable itself so the information is not very reliable. De Moivre’s theorem tells us that the variance of the sample is related to the variance of the population sample as the reciprocal of the square root of the dimension of the sample. Four people saying “beautiful” are twice more reliable than one customer and begin to be statistically significant.

As far as the content of the variables we want to measure, they have to allow you to create a predictive model so they must be objective oriented. Let us make an example: sales turnaround during full price periods and during sales. If we make this analysis in mono-brand shops this becomes an indicator of price correctness between different categories as the offer stays the same and only the price changes so if the trousers percentage sales increases it means that people thought that trousers were too expensive. If the same statistics was run in a multi-brand wholesaler this could be an indicator of “among-brand” competition.

This last example introduce us to the third problem: the value of a measured variable has to be properly fit in its context. A large fur company made a large error some time ago when its sales rose quite lot in one period. This was due mainly to a large competitor’s problem but the signal was interpreted as a growing trend. The fact was that the company did not take into account the fact that the whole fur-coat market was in a steady/declining situation so the raise was to be explained as temporary phenomena.

This dimension can thus be summarized as the capacity of collecting “good” data on significant variables and fit them in the right context.

Methods and instruments: the collection of data is a very difficult process from an operational point of view.

Defined needs, as we said these needs exist only for known customers in well defined places. Almost all sales point have instruments for operational processes (inventory movement, invoices, cash accounting), some of these products also have limited CRM functions mainly for after sale care (fidelity cards, recalls). A few of them also have options for supporting sales planning both a budget point of view and from a space/path analysis. As far as I know there is no common solution for requests and lost sales analysis.
Semi-structured requests related to occasional or unknown customers. In my opinion this is an uncovered field as no specific instrument is available. I read about one experience of customers wearing RFID bracelets to follow their path inside the sale room but it seems to be abandoned (I would not have liked it). Video analysis for figure recognition, apart from privacy problems, is not yet fully available.
Geographical or context needs: This is a huge and extremely interesting information problem. It relates to the capacity of analyzing socially open spaces to get signals of “trends” which might appear. As far as I know there is nothing like this existing at the moment. An interesting approach would take into consideration TV and movie pictures: they are biased by the choice of the costumes director but they can be analyzed more and more, till when the algorithm gives acceptable result.

Ability to understand, rationalize and interpreter customer needs

In this case the context system to be considered is the company and the process goal is to share the data perceived by the marketing department with the design, industrialization, production and logistic departments (industry) or purchase department (retail) which are the ones who have to propose a company realistic solution.

Why: the existence of this process is related to the need of converting general into company useful data; we need to translate context events into an internal useful language.

Who: people can be involved in this process both as process and state definition resources. As we shall discuss later (How) methods can be various and this implies that the relative importance of the various actors is quite different. Going back to the various roles, we can point:

People supplying data are the same collecting them or market search external companies
People rationalizing customer needs: marketing, product or market managers as this is a very delicate operation which needs high technology competence and market knowledge. Maybe a better name could be “product/service” analyst.
People interpreting: in the fashion-product environment they are called “stylists” and they merge context, unstructured, informations with marketing data with a view for a long time horizon.

The relative role of these people is obviously related to the company market positioning: in a luxury company the stylist role is prevalent compared to the same in a large mass market enterprise.

People using these data are people in the design and supply chain of the company.

What. As we said customer need is made up of physical objects (the product),subjective aspects (personal esteem) and social requests.

As far as product requests we can divide between “upgrading”, variations on existing themes, and “innovation” which relate to requests which did not exist or did not have an answer in the former technological context. This last need is the most difficult one to satisfy as it implies the ability to interpreter social sign which could have an impact on future customer desires.

Subjective needs can be divided into “Basic” messages related to the “universal” nature of man, for example ease-of-use, comfort, functionality and “culture related” messages linked to the context the man lives in, for example color choice and matching which are very related to the society the consumer lives in.

Social messages: our aim is to be able to quantify social attention and recognition and social reputation.

At the moment these are very vague parameters.

Where. As far as the geographical location where the process exists we must keep track of two aspects:

where the process actually takes place, mainly the company main site, and where the process results are used, typically we are talking about the localization of the results. At the moment the most common ways to proceed on the topic include either local analysis and central strategic consolidation or central analysis and local adaption. Unfortunately both solutions tend to have a “time to market” too long and this reduces the competitiveness of the system, we shall get back on this.

How. As soon as we start talking about analyzing data, so that they become information, the first thought is about data-warehousing, data-mining and, eventually, knowledge management. In reality we are now trying to rationalize and interpret needs and not hidden or explicit correlations. The problem is thus quite bigger and it must include also informations different from what we have in our standard information systems. From a methodological point of view we can divide analysis into

Rational: taxonomies and abstraction/interpretation methods and instruments. This last group include projections (defined schema and historical data) and previsions (new schemas on historical data)
Semi-quantitative: analysis of competitors, market semantic analysis
Irrational or context analysis.

Usually all methods are taken into consideration as we try to avoid theoretical models which could be not complete or intuition with no relation with reality.

It could be very interesting, at the moment I know of no tries, using heuristic techniques for the solution of interpretation models; I guess model are still not formal enough.

''''When. From the time point of view we must consider two different horizons: the analysis period and the time the results are meaningful.

As a rule of thumb it is correct that the analysis should be done at the same time of the data collection as it might turn out that some data are incomplete or wrong.

The validity of the result, for the physical part, is related to the total lead time of the product we are considering while it is not clear how long our analysis is valid when we talk about customer service, it could be interesting analyzing this.

How much. This dimension is always very difficult to quantify. Let us go back to the Maslow pyramid: how much do I need to satisfy level 2 (product) before moving to level 3 (personal needs) ? We assume that this percentage does not need to be 100% as once the homeostatic equilibrium of the lower level is reached, attention will move to the upper level. The point is: solving the lower at 80% and the upper at 50% (assuming that you succeed in measuring it) is better than 90%-40%?

I think it is a company choice to be made.

Altogether we must estimate the need for

product in terms of which is our reference market, its dimension, our positioning (the quota) and how much does our product satisfy the market in terms of price/performance
personal/social aspects: how important and how much do we satisfy them.

Instruments. Now that we went through all the dimensions of the process of “Ability to understand, rationalize and interpreter customer needs” we can have a look at the state of art as far as available instruments. We can group them into

Analytical methods: based on known and recursive data
- Algorithmical and or statistical: at least statistically precise
- Heuristic: model not totally defined
Contextual methods: using common sensations, even if not measurable and recursive
- Brainstorming
- Top-down or bottom-up methods.
Individual methods: related to a particular man-environment situation, very often give non-reproducible results, e.g. Zen ispiration.

As a personal opinion, even though different methods could be seen as provocative to some people (think of a physicist who should use a zen approach), as long as this process gives usable results is should be accepted anyway. Human beings have being using gravity as a force and apples kept falling long before Newton quantified the mathematical model. Man was using heuristic models to create manufacts.  In field of social science maybe we shall never have a Galilean, recursive model  but this does not mean that we must not get better use of data to be informations.