Modern IT fashion under skeptical scrutiny
Why Big Data doesn’t work in the real sector?
Why it needs no artificial intelligence?
Why it doesn’t embrace blockchains and smart contracts?
Why "neural networks" are a dead end?
Finally, what will really help our real sector leap from the 1980s into the 21st century?
Let’s start our search for prospects with artificial intelligence.
Since it makes no reason to talk about artificial intelligence from a scientific perspective (AI as a term has no scientific definition) we’ll use its colloquial meaning: a functional human mind’s analogue of artificial origin.
True artificial intelligence when, if ever, it appears will be in demand only accompanied by human brain: either to amend evolutional shortcomings of human mental processes or to imitate them. In other words, artificial intelligence is either something similar to a power suit for a feeblish human brain or a recreational exercise: imitating typical behavior patterns of an unreliable biological computer.
AI without "natural" mind interacting with it is essentially meaningless similar to a shoe without a foot. Rational needs of production can be satisfied completely by the progress in robotics: whether conveyor industrial robots or easily re-adjustable built-in algorithms in the intelligent systems to replace the office staff.
That is the intelligence real business needs: not less yet no more. An ideal replacement for a ticket agent in the subway is a ticket vending machine rather than an artificial intelligence of Bender Bending Rodriguez from a well know cartoon.
Put in a good word for Big Data
Real big data success is related mostly to financial sector. Real economy lacks ones. Except for that notorious story when Target had figured out a teen girl to be pregnant before her father did. Frankly the whole story is so much movielike that it will probably weigh heavily on the PR people’s conscience.
Anyway, no solid profs of any beneficial influence of Target's claimed achievements with Big Data on the company’s major business? The reality is not encouraging: Target is out of the Top five of the US retailers. While Walmart has been number one for many years and still is.
Mass media quite rarely mention talking about Big Data power and magic that raw data must be ideally consistent, adjusted, filtered and — eventually — mathematically robust. Unless data does not suit these requirements than — alas! — basic principle of cybernetics in action: garbage in — garbage out.
The success stories mostly addressing laymen either assume there is ideal raw data a priori or such side issue is never discussed whatsoever.
Well, off we go. First, where can a retailer gain high-quality raw data? Straightforward question as it seems. Naturally, it comes from the corporate information system. But... "The devil is in the details".
Typically, a regular store chain uses an ERP system consisting of a set of modules covering the headquarters" major business processes (SCM module, reciprocal payments, CRM, financial module, etc.).
Sometimes with WMS module at the distribution warehouse. In general, the number of the ERP modules and functional diversity are limited by the company’s solvency and the top management’s desire to find IT elixir.
Hundreds of stores are equipped with local systems with data being regularly uploaded to the core ERP system.
Since it is modular ERP is not a "system" in colloquial language. It is nothing more than a set of dependent, semi-dependent or independent separate programs (modules) with their own datasets to be also synchronized continuously: there is a great many of separate and simultaneously functioning mechanisms of synchronizations interacting and mutually responding to changes in a way one can hardly foresee.
So, all those syncing routines spawn mismatches and discords on a daily basis. Or, in more academic language, it grows unidirectional entropy in a poorly ordered system.
The only representative snapshot within the entire chain is data from sales receipts which can be received from the cash-registers. Interestingly, what makes it possible is that cash-registers are read only thus they exist somehow outside the perimeter of the corporate IS.
Information from the receipts — date, time, customer, items with prices — is an extremely limited ground for analysis. Thus, though results might look like exact figures (array of numbers) but their reliability is no better than a rule of thumb.
For instance, useful interpretability of sales data downloaded from cash-registers without accurate and reliable inventory stock data for the reviewed time period equals zero. If no item A was sold last week it might mean anything — low demand, uncompetitive price, or it was out of stock.
Moreover, even if constrained slip data section analyzed shows anything meaningful, apparently, it will be either information about staples (like beer and chips) useful to change the layout or some info to organize a successful personalized email marketing.
End of story. Since receipts do not contain any other relevant data.
One might say something is better than nothing. No way for if you have something really confirmed effective it will be immediately duplicated by competitors. In the end, the grateful ones may be competitors but can’t be shareholders.
What about blockchain?
It is all just the history repeating. Smart-contracts applied to the reality of a typical hang in the air due to unreliable CIS data dead-end whereas the instantaneousness of the response is limited by non-standardized synchronization rate between modules.
Big data technology or smart-contracts are not to be blamed. The very basis is critically behind times: the paradigm of modular ERP has hardly changed for the last 30 years.
It perfectly explains why the financial sector succeeds. The only goods banks work with is money and one US dollar always equals one US dollar. So even primitive CIS in the banks can cope.
In real economy with the competition high all the companies including retailers exists in the environment with such multidimensionality, aggressiveness, stochasticity and growing mutation rate that if compared with the bank’s ecosystem the gap is as huge as the difference between the Sun’s core and a steamer. It goes without saying ERP systems from the 1970's are of no help.
What instead of ERP?
Naturally, a new business automation paradigm requirements are quite obvious: single enterprise information field; guaranteed data consistency, reliability and permanent availability and intelligence in the amount necessary and sufficient for business processes to be executed in automatically and unmanned way.
Therefore, it is necessary to ensure the transactions" execution in real time and within a single information field. In turn, it requires effectively high (moreover, guaranteed) system performance and robustness in general.
So, the output is ideal data to be analyzed with DB methods (machine learning, data mining, etc. to your liking), available online without any preliminary processing.
At the same stage it's all prepared to be connected with the Internet of Things under the umbrella of a new post-ERP paradigm: data is fed directly by RFID tags, electronic scales, barcode scanners, web-sites, automatic warehouse loaders, weather stations and unmanned drones.
Let's think once again on intelligence mentioned above. The one can be optionally implemented as adjustable system scenario responses of any complexity to external events and internal changes.
Dreams on IEM
This post-ERP system will not only be able to plan resources (as we remember, ERP means enterprise resource planning) but to perform immediate and straightforward enterprise managing.
Thus we come to the understanding of a new paradigm — enterprise managing system. Moreover, considering required business process automation all the grounds are there to call it "intelligent". So, please give a warm welcome to IEM (intelligent enterprise managing system).
It is characteristic that well known icons of so called "new economy" — Amazon, Ebay, AliBaba, etc — use anything but not mass produced ERPs. Even more — conservative businesses (such as Walmart) prefer in-house products. Majority of home-made CIS used by leading companies do meet many IEM-related requirements. Though can we consider them as true IEM systems?
In-house development of fully functional corporate management system has the most important advantage: perfect customization. On the negative side — resources (money and time) and risks.
Hundreds of thousands companies have tried to develop their own CISs for the last decades. How many of them managed to make a good one? Everybody knows them (Walmart, Amazon…). These ones steadily remain among the most popular choices for the front pages. What about the rest? You’d better ask some professionals.
That way mass-produced ERP-systems look more appealing indeed: a 30% to 60% chance to have something functional. Any theoretical IEM-related speculations can be valid if two hardly compatible parameters are met: customization potential not inferior to in-house products and, at the same time, deployment costs and failure risks not inferior to mass produced ERPs.
However, free economy is changing much faster than 30 years ago. So, if an ERP-like system goes into action in 5 years (with a 30-60 % probability) it is useless. Any business in the competitive environment constrained by the transition period limitations will die much faster.
Thus, IEM Paradigm can become viable on condition that it is deployed much more rapidly in comparison with how long serial ERP-systems usually take let alone in-house ones.
Reverting to fashionable innovations the case is as follows. Talking about Big Data instead of scarce information from the snapshot with 3 to 5 parameters from the cash registers the IEM System will provide 100% valid data on hundreds or thousands of factors, with tremendous number of analysis-relevant interlinks and all this 24/7.
Multifactor dependencies (human brain is fundamentally unable to imagine), deviation management (in particular, faultless detection of corporate corruption), and so on and so forth.
Talking about opportunities — the sky is the limit for IEM.
Omni channel narrowly known to retailers, which is mainly an insoluble task within ERP paradigm, is implemented in quite natural and obvious manner using IEM platform applying single information field environment with real-time transactions.
Block-chains with smart contracts aggregated with functioning IEM System will turn real economy into something very similar to what has already been on the stock exchange for years: robots trade with robots.
Transaction costs will go down through continuous clearing of millions operations for millions of companies and billions of people daily.
Goods will not travel a long way from the manufacturer to the distributor’s warehouse and then to the store than a(oh, my!) finally to the customer place which is… near the plant.
To sum up these future IEM systems are to become the foundation for digital business revolution coming next decade, the one comparable with telegraph and railways invention.
In other words, implementing IEM in real life is the key to throw open the doors to the XXI century for real economy.