END USER EXPECTATIONS
Bill Inmon
Founder, Chairman, CEO, Best-Selling Author, University of Denver & Scalefree Advisory Board Member
END USER EXPECTATIONS
By W H Inmon
Users don’t know much about computers, for a lot of reasons. Computers are complex to build upon and to use. Computer courses have been taught in school for only a short time now. Computers require a certain mindset. And in the haste to study other disciplines, for most students there just isn’t much time for studying computers.
For these reasons and other reasons as well, end users end up in accounting, marketing, sales or finance with only a scant or limited knowledge of computers.
So what happens if you start to talk about the building of a new computer system with your end user? What is going on in the mind of the end user? What are the end users assumptions and expectations?
Assume that your end user knows little or nothing about technology (which is a typical state of mind).
You start talking with your end user and what are the end user expectations for the system that is being discussed?
The presumptive end user expectations for the system that is to be built can be described by the following simple chart.
The first thing your end user expects is that the system is automated. It doesn’t do the end user much good to expect anything else. This is the most basic of the end users assumptions.
The second very basic thing the end user expects is to be able to access whatever data is being presented. Again, this basic expectation is so fundamental that a computerized system would be worthless without meeting this expectation.
领英推荐
The third basic thing the end user expects is to be able to access the data in a reasonable amount of time. Now this issue gets to be complex because the waiting time for access is very different for different kinds of data. If you are at an ATM machine, you expect sub second response time. But if you are being audited by the IRS it may take you a week to get back checks that you wrote ten years ago. So the speed of access to data is a very relative thing.
Another thing the end user expects is that data across the enterprise is integrated. When the end user refers to monthly revenue the end user wants the corporate revenue, not the sales forecast or the marketing revenue.
Finally, the end user expects the data that is requested to be accurate. Data from a spreadsheet that someone has concocted a half hour ago at their desk is not what the end user expects. The spreadsheet could say anything and have no basis to reality in any case.
The end user has these expectations whether or not he/she as ever seen a computer before. These are just what the end user assumes that all data and all automated systems provide.
Now let’s match up the end users expectations with some of the popular architectures that are out there. In particular let’s look at a data mesh architecture and a data warehouse architecture.
Data mesh meets some of the basic issues of end user expectations. When it comes to automation, access and speed of access, data mesh is a good fit with end user expectations. However, let’s?look at integration and integrity and a data mesh architecture. Data mesh does not meet those expectations at all. Or it is just an accident if those qualities are found in a data mesh architecture. Now let’s compare the end users expectations and data warehousing. Data warehousing fulfills ALL of the end users expectations.
It is true that a data warehouse architecture takes a lot more effort to build than a data mesh architecture. But the data warehouse fulfills ALL of what the end user simply expects to be in a system. The data mesh architecture does not.
So why does a data warehouse architecture take more of an effort to build than the building of a data mesh architecture? The answer is simple. In order to build a data warehouse you have to integrate data. With data mesh there is no need to integrate data. And – however you do it – integration takes time and work.
So when you start to compare architectures, you have to ask yourself the question – do I want something that is cheap and fast to build, but fulfills only some of my end users expectations, or do I want something that is more difficult to build that will fulfill all of my end users expectations?
The choice is yours.
Bill Inmon lives in Denver with his wife and his dog Jeb. Jeb goes for at least one walk a day unless it is snowing and/or icy. Jeb gets to be cranky if he doesn’t have his daily walk, where sometimes he meets his girlfriend Penelope, a lhasa apso. Jeb is always happy when he meets Penelope.
Solution Data Architect, Enterprise Architect, and Senior Data Modeler, applying an end-to-end view of the enterprise to solve the right problem in context.
2 年My understanding of data mesh is a little different. Due to the distributed nature of the data mesh paradigm, it becomes the ‘responsibility’ of some who cares about integration to take the initial data product from the source systems, and then do the integration. My concern is that this distributed nature would naturally lead to lots of little islands of integration; but then you need some serious data governance to avoid duplication (or slightly different integration algorithms). I see it as a reaction/backlash to heavily centralized DW groups where it takes months to make a small change. That doesn’t mean the answer is to overly decentralize either. (See above about DG.) I expect, like most new ideas, there is some value in there, but is initially overblown. In a couple of years each organization will need to identify the ‘right’ balance between centralization and decentralization; or the next big idea/fad will come along.
Principal Product Owner | Product Manager | Strategist for Data Governance, Data Catalog, Data Privacy
2 年Well said. Companies have been looking for a short cut to the same problem for many years, when the answer has been right in front of them for decades, thanks to pioneer and thought leader #billinmon.
Supporting ambitious CXOs growth with AI and Data | Fractional CTO/CIO | Multiple Patents |Follow for posts about improving your business with AI and Data
2 年Without integration and integrity, each data set could be different from the other. You will lose your end user trust in the data that you are giving them. Also, any results from AI algorithms that you run on the created data sets will not align with each other. Not doing a data warehouse is a disservice to your business, if you are anyway doing data projects.
Director, Data and Analytics Center of Excellence, GSA
2 年Bill Inmon you are spot on with the observations and I will also say the user scenario is changing. There are various groups of users depending on when they joined the knowledge work force and how well they kept with computing in general. Recently I met a MD who is whole heartedly serving as a Data Architect. I was taken aback by their competence and dedication. I also notice that end users have their lens (day time job) on when they are interacting with the systems and not necessarily think through how those systems are architected. Computing Literacy along with Data Literacy (most fundamental) should be an on-going effort for an organization if they want to survive. As far as Data Warehouse is concerned I have seen integration requirements vary by users. Its just impossible to pre build integrated data that will satisfy all users' needs. Due to the investments involved in building Data Warehouse it's not possible to justify a data warehouse for just a select group of users. With the availability of modern data technologies these unserved users are willing to roll-up their sleeves and do their integration as per their unique needs from the data available through Data Mesh etc. that is most reliable, current etc.
Living at the intersection of people, systems & data
2 年Hi Bill Inmon, I've been pondering a similar thought lately, curious on your take - a "what if" scenario, but hopefully not too hypothetical, I mean this as a practical question. "What if" treating data as a product in a competitive environment leads to survival of the fittest data products? Can a network of data producers and consumers independently evolve data products where the best survive and multiply, especially with all the competitive pressures? I don't have an answer yet (but a curiousity), and based on your comment, I suspect you might answer no. And I have been a practioner and long studied your work as well as that of the founding fathers of data (much respect), so I do feel I have at least some understanding of the real world implications and challenges of working with data. Yet, I can't shake the possibility that the hand of economics, capitalism, competition, evolution, etc might actually end up playing a hand in the story of human use of data. In other words, the work to model, govern, cleanses, integrate data, etc, whether distributed or centralized or every which way, will be forced by environmental factors and only the strongest data products will survive?