大数据安排的战略布局（英文）

Strategic Placement for Big Data in   Organizations

Posted by Don Philip Faithful
I tend to examine the different roles played by data. For instance, when I work on computer code, I often ask myself what the presence of data is meant to accomplish. Sometimes the analysis is not at all straightforward or simple. In society and organizations, people exist and persist in the records as data. The data survives even as employees come and go. I therefore consider it important to regard the data and its environment as a system in itself, something that has a life all of its own. The data exists beyond the individual sentiments of managers, administrators, and customers. Data contributes to the process of reasoning and the expansion of knowledge. Consequently, I find it a bit disturbing to read about “big data” often in rather vague terms. I notice that discussions tend to lack a theoretical basis. In this blog, I will attempt to provide a conceptual foundation for big data as it relates to organizations. However, the underlying purpose of the blog is really just to stimulate discussion. I expect that some readers will immediately disagree with some of my points and offer their own perspectives. In the end, I am sure that the foundation for any field of study develops over time through collective and collaborative efforts.

Perhaps when the boiler was first marketed, some companies looked at the technology and thought, why would any organization need one of those things? Can the ability to heat fluid and move it to different places be useful? Think about the amazing feats of production made possible by being able to have many people comfortably occupying a shared space full of capital resources regardless of the time of year. Discussions about big data tend to suffer from the irony of being marketed in small terms; this makes it possible to dismiss a rather big idea by its apparent lack of congruence to isolated applications. Yet I notice no shortage of vague generalities. “You say this boiler will change the way we do business, but I don’t understand how it might help me sell more hogs.” I consider the following a worthwhile approach:  start with a conceptual foundation and then build up a case. Trying to sell something complicated without any firm reference points can be problematic: generality plus peculiarity equals implausibility.

In order for me to offer a setting for big data, it is first necessary to examine the role of data in organizations. I have chosen to approach the challenge of establishing placement through systems theory. But rather than chop up the usual liver, I will go ahead and add my spin. Although the systems model seems to explain the flow of materials through organizations, it is also possible to interpret the model in relation to the flow of instructions, information, and data. In this blog, I will formally separate the structural complexities of an organization from its informational complexities. Many people are familiar with the basic components of systems theory. I want to point out that the theory holds the idea of progression or direction: the movement is from inputs to outputs and not the reverse. Direction and flow are particularly evident in organizations once the delegation of authority is assumed: organizational control tends to emanate from “the top.” Just to make my point, I made a minor addition to the systems model diagram

The so-called “top” in relation to my flow diagram is at the left. I call the emanation of authority “projection.” Projection creates data by imposing specific requirements that can then be used to determine degrees of adherence. I call these requirements the metrics of criteria. There is a saying that goes, “What gets measured gets managed.” I suggest the truth is slightly different: “Managers measure what they want to manage.” My sentence is probably rather awkward to say; but anyways this is not a blog on English grammar. A long time ago in simple work environments, oversight and control were probably more direct. Control meant giving instructions and remaining present long enough to ensure adherence. In most contemporary work environments, there are still work orders and instructions. But the evaluation of work occurs through the use of criteria. The implementation of criteria offers the workplace tests or standards that give power to authority. Consequently, ensuring compliance is a matter of applying standards of conduct or behaviour and then generating data, assuming that the organization hopes to eventually manage what it measures. Data is the product of projection - among other things.


Criteria can be found in dashboards, scorecards, and many forms of performance evaluations; these may or may not be well-aligned with the broader business objectives of the organization. I hope most readers would recognize that although there is quite a bit of data involved in management regimes, in fact the need for “big data” is fairly limited. It should come as no surprise how big data might seem baffling and be met with considerable skepticism in particular organizational settings. Specifically, operations where there is little employee participation in decision-making processes do not require sophisticated management approaches. I would go so far as to say that these types of organizations are sometimes specifically designed to minimize the number of managers. Procedures and processes might be rather entrenched. So there is little need for intervention or creative problem-solving. It is not my intention to make any kind of criticism but simply to highlight the limited data requirements of more rigid organizational constructs. When conditions are static, there are maintenance rather than growth opportunities in relation to data products and services. Organizations would probably need protected markets to persist for long in such circumstances.


Although it is probably not an enlightened management strategy, it is possible for an organization to project authority irrespective of circumstances or consequences. For instance, a company can manufacture products that nobody will buy. Managers can order employees to flap their arms like chickens, which I think would be dreadfully comical. Why flap arms? In Canada, a particular province required everybody in its fire department to achieve specific aerobic outcomes as a condition of employment. I am not saying that this is the same as imitating chickens. I am focused on how the criteria can be disconnected from reality. The Supreme Court found the logic of such criteria questionable for somebody maybe working in dispatch. A nation can, as part of a back-to-work program with good intentions, hire people to dig holes and others to fill the same holes, the idea being to man-handle the data to agree with the criteria for progress. We still encounter situations like coal mining accidents and factory fires perhaps in the textile industry when workers remain at their duties despite extremely hazardous conditions; the performance criteria is not necessarily connected to worker safety. Suffice it to say, projection can occur in an entirely desensitized organization; and companies routinely find themselves sliding into competitive oblivion due to their impairments as I will explain in greater detail shortly.

I added some megaphones to the diagram. While the identities of the other parties are probably debatable, I would suggest that at the far left we have “the market” (those that might buy); at the far right “the clients” (those that have bought or tend to buy). This is something of an oversimplification. The market occupies both sides. Those that have bought can easily influence those that might buy. However, those that are most likely to complain have already bought; this is why their megaphone is pointing back at the company. In traditional systems theory, the idea of “feedback” might be compared to a response by the market to the behaviour of an organization. Government regulators can often be found at either side of the organization: at the left to ensure administrative compliance to work standards; at the right to protect consumers and citizens once something unacceptable happens. These are fairly hostile conditions for an organization that has projection as its main operational strategy: thus the “!?!”


Except in the case of a tight military operation with certain specific objectives, projection irrespective of consequence would probably contribute to disappointing results. Also from the standpoint of data, projection offers just a certain type of data. The only real measurement is level of compliance or adherence to expectations. It is not necessary for the organization to understand anything about its environment since there is no need to adapt beyond the original design parameters. I call this a “survivorship tactic” not to be confused with a survival tactic. If we throw seeds over different areas of land, there is limited adaptation to the circumstances. The seeds will grow on fertile soil and exhibit different degrees of success on less fertile soil. That is all they do. It is a “live or die” strategy. Similarly, an organization with a mission to make pencils might do well in a place where pencils are quite scarce but there is an abundance of paper. So there is an information ceiling in this setting that greatly reduces the need for big data. One does not need massive amounts of data to ensure that people follow specific instructions. Nor would following the instructions necessarily lead to organizational success. Even in situations of total compliance, there might be absolute failure. Again, this is a “live or die” setting - natural selection rather than internal adaptation.


I covered projection. I actually have two more major data flows to cover. But I am going to skip the one in the middle and return to it later because it’s easier to do so in terms of the diagrams. I apologize for the confusion. So we return to the organization in its hostile operating environment, generating data through its metrics of criteria. We hear wailing and gnashing of teeth. “This is ridiculous. I need some kind of process to satisfy the market, deal with consumer complaints, the regulators, activists, our suppliers, wholesalers, and retailer distributors.” It is actually common in stable systems for another player to be added to the organizational setting. The term “player” is rather embodied, suggesting perhaps that I actually mean a body; but actually I am referring to the part or role played. I describe playing this role as “articulation.” Specifically, “internal articulation” is where the system attempts to self-regulate or adjust in order to prevent aspects of insulation or isolation (desensitization). Some organizations will attempt to anticipate problems before they occur; or some effort will be made to incorporate experiences into future decision-making.

A simple example of internal articulation can be found in heating systems. A furnace can heat a house even in the summer during the hottest days. A furnace makes use of a special device to ensure comfortable temperatures: a thermostat is an important part of a heating system although it usually generates no heat itself. A thermostat does not “control” a heating system the same way a home owner controls a thermostat; but rather the thermostat fulfills a function that is an integral part of the system. In a production environment, which might be able to produce continuously regardless of demand or what comes out of the system, the role of articulation is sometimes played by those in quality control and assurance; customer service; marketing; and logistics. The extent of articulation depends on the intended scope. If a company tries to sell a product that nobody wants; that is deficient in some manner; that cannot be feasibly brought to its market; then projection actually becomes almost pathological in an organizational sense. I believe that articulators play an important role in adaptation. Articulation can generate much more data than projection because of the different types of metrics that are often needed to make an organization sensitive to its environment.

The data generated by articulation results from the metrics of phenomena. In my own programming projects, these metrics are conveyed through complex data structures. Theoretically as I already pointed out, it is possible for an organization to operate using only methods of projection. However, once sales start to decline, the data system might provide few if any insights on the reasons behind the slide, assuming that the system is not meant to articulate information about the environment. If an egg producer cannot successfully and consistently bring products to market, retailers might find a more predictable and reliable supplier. The data system can give many indications of problems before radical decisions are necessary: for instance, unusually long delivery times using particular routes under different road conditions. These are useful metrics. I believe that articulation is poorly developed from a technological standpoint because it is so much more complicated and difficult to achieve than projection. Simple and perhaps even the simplest organizations can implement a rigid delegation model. Complex enabling structures on the other hand require a high level of organizational sophistication. I feel that articulation offers the best strategic placement for big data.

The data generated by articulation results from the metrics of phenomena. In my own programming projects, these metrics are conveyed through complex data structures. Theoretically as I already pointed out, it is possible for an organization to operate using only methods of projection. However, once sales start to decline, the data system might provide few if any insights on the reasons behind the slide, assuming that the system is not meant to articulate information about the environment. If an egg producer cannot successfully and consistently bring products to market, retailers might find a more predictable and reliable supplier. The data system can give many indications of problems before radical decisions are necessary: for instance, unusually long delivery times using particular routes under different road conditions. These are useful metrics. I believe that articulation is poorly developed from a technological standpoint because it is so much more complicated and difficult to achieve than projection. Simple and perhaps even the simplest organizations can implement a rigid delegation model. Complex enabling structures on the other hand require a high level of organizational sophistication. I feel that articulation offers the best strategic placement for big data.

There are certain common terms that are tossed about in the community, and perhaps I have been responsible for some of the tossing. For the sake of consistency in my own posts, I want to explain that projection and the metrics of criteria are part of the “prescriptive” regime of organizations. In contrast, articulation and the metrics of phenomena help form the “descriptive” regime. Description is ironic by the simplicity that it implies; it suggests an almost mechanical process of describing things. I don’t really cover the “predictive” aspects of organizations; this is to say, I leave out the term in my diagrams. So my colleagues are ahead of me given my ambivalence to prediction. Further, I immediately place myself at a disadvantage by approaching as my main area of discourse something that many people dismiss as rather straightforward. It’s been said that Eden is protected by a burning sword, without which paradise would become quite a tourism destination or condo-development project. However, it would be far too easy to locate Eden if this were true given the rarity of burning swords.  Some people would go just to see this peculiar burning sword.  I therefore study obvious things that are too perplexing to justify further consideration.

I am only just now starting to write about data-embodiment, which is my personal approach to articulation. In criteria, it would be fair to say that measurements tend to be rather specific. It is easy to determine what belongs and doesn't belong. Absenteeism is about lack of presence. So a person missing from the workplace could be recorded as absent. The exchange of goods for money leads to sales figures. In phenomena using data-embodiment, the measurements don’t behave in this manner at all. Everything is related to sales albeit at different levels of relevance. In other words, everything is related and connected but not necessarily well. There never has to be a clear separation. A data-embodiment approach tends to make use of large amounts of data - potentially all of it - all of the data - and possibly more massive amounts of assertions of data. For instance, PERSON and BRIEFCASE might represent specific isolated events. But it is possible to assert the following given the types of people that carry briefcases: SUSPECTED_SALESPERSON or SUSPECTED_TERRORIST. Temporal and spatial issues create challenges because the act of description might be for transient phenomena. Interconnections determine the reasonableness of recognizing things as existent. I have described data-embodiment as an “ecosystemic approach” since no effort is made to examine things in isolation; it is necessary to handle the data in an interconnected and complex state. This is different from focusing on specific aspects of isolated things in standard ways: e.g. monthly sales.

The final informational flow is practically a side-order of fries compared to projection and articulation. I call the third flow “direction.” Projection is an aspect of design. Articulation is an aspect of self-regulation. Direction reflects the intended routine of an organization; it likewise generates data. Good examples of direction can be found in the financial industry: large amounts of data are produced over the course of operations as an aspect of the day-to-day business. So if the megaphones are removed from the left and right of the organizational construct, business can still take place as usual for a period of time under normal conditions. Direction generates accounts, transactions, logs, registries, and records. I feel that when most people discuss big data, they actually mean data from how organizations direct their day-to-day routines. Big data would therefore be big in a literal sense due to the amount of data; but beyond quantity, the data itself creates no unusual demands on existing systems. The illustration below contains all of the points raised so far plus a few that must be set aside. I will try to discuss organizational domains and pathologies in later blogs.

I feel that organizations that are destined to do well with big data are those that direct significant resources at articulation.  This is something of a polemic statement from a management perspective:  it tends to support the rise of complex operations, which creates a need for sophisticated management approaches and therefore a need for managers.  Nonetheless, it hardly makes sense to collect much data if there are few plans to ever use it.  Big data requires complexity.  So do organizations that employ adaptive strategies.  I agree that simple and efficient operations sound lovely as far as frilly preconceptions go, except that other companies will always be able to offer the same products and services faster and less expensively.  Efficiency is no substitute for growth.  Even in completely protected markets, at some point saturation will be reached; beyond this, expansion will depend greatly on environmental sensitivity.  Articulation supports adaptation by making individual organizations more connected to their unique environments and specific circumstances.