It looks like nothing was found at this location. Maybe try a search or browse one of our posts below.

There are days I curse Disraeli.

Benjamin Disraeli, one of Great Britain’s most distinguished prime ministers, uttered the over-quoted “lies, damned lies and statistics” more than a century ago. It’s been used as an excuse to ignore statistical evidence ever since.

I guess it’s time for dueling quotations, so here’s another: “A witty saying proves nothing” (Voltaire).

The ability to see the world in statistical terms is part of every good manager’s mental toolkit. To understand why, start with the strange case of Judge Roy Pearson. Pearson, a Washington, DC administrative law judge, sued Custom Cleaners two years ago for losing a pair of his pants. For more than $65 million in damages.

To be fair, Custom Cleaners did post signs saying, “Same Day Service” and “Satisfaction Guaranteed.”

Judge Pearson doesn’t, it appears, understand statistical concepts — like, for example, service levels. You’ll recall that these are two part measures, describing a service standard and how often a service provider meets that standard.

If Judge Pearson thought in statistical terms, he would recognize that every time a dry cleaner handles a garment it represents a statistical sample. The probability of it being properly handled is a number between 0 and 1. Were a dry cleaner to run its operations as IT does, it would define the percentage of garments per day it must handle properly — maybe 99.9%. It would next define how many days a year it must meet or exceed that standard — maybe all but two (99.5%).

The dry cleaner would then post that service level as its commitment to its customers. Assuming enough IT professionals frequented the shop to keep it in business (or that customers like Judge Pearson understood statistical concepts), everything would be great.

In business, no matter what you do, the only accurate description of your results is, in some way or another, statistical. If you manufacture, each item varies a bit from every other item. The important question is how often they vary beyond acceptable limits.

If you manage customer service, you’ll never make every customer happy, or even satisfy them. You can’t. Your influence over customer states of mind is less than your influence over manufacturing tolerances, so once again outcomes are statistical events.

The challenge when you don’t think statistically is the risk that you’ll invest in goals that are no more achievable than perpetual motion machines, and complain when others don’t achieve them.

Think you’re the exception? If you’re a baseball fan, have you never griped when a high-dollar batter failed to make a clutch hit? Statistically speaking, even the best hitters will fail in the clutch two out of three at bats.

But thinking statistically is a mixed blessing: It can turn into complacency in no time flat. After all, if perfection is unattainable, what are a few errors here and there after all? Nothing to worry about — they’re inevitable.

The statistical nature of things doesn’t have to make you a helpless spectator, trapped in the randomness of things. What it does is provide a more useful framework for making decisions than thinking perfection is possible.

The statistical nature of things is why business decisions must be framed in terms of the law of diminishing returns rather than the desire for invariant outcomes. It’s the law that tells you each additional increment of improvement will be more expensive than the last one. It’s a statistical thing: You can only predict events in terms of probabilities, not certainties. The closer you are to perfection, the more variable are the errors that remain.

So if you currently backorder 20% of all customer purchases and want to improve that to 15% (a 25% improvement), it will cost you less than if you want to improve from 15% to 10% (a 33% improvement).

Looked at through the other end of the telescope, if improving from 20% to 15% costs a million dollars in additional inventory, the next million dollars of inventory will only improve backorders from 15% to 11.25%. And if you keep on spending until you only backorder 1% of the time, the next million dollars of investment in inventory will only reduce your backorder rate to 0.75% — a barely perceptible nudge.

Does all this mean a business can’t post “Satisfaction Guaranteed” and “Same Day Service” signs without prevaricating? If the law fails to take statistics into account it could happen. We’ll all have to replace simple declarative sentences with elaborate contractual phrasing filled with weasel words, and the world will be the worse for it.

What’s the difference between a “Digital Twin” and a simulation? Or a model?

Not much, except maybe Digital Twins have a more robust connection between production data and the simulation’s behavior.

Or, as explained in a worth-your-while-if-you’re-interested-in-the-subject article titled “How to tell the difference between a model and a Digital Twin,” (Louise Wright & Stuart Davidson, SpringerOpen.com¸ 3/11,2020), “… a Digital Twin without a physical twin is a model.”

Which leaves open the question of what to call a modeled or simulated physical thingie.

Anyway, like models, simulations, and, for that matter, data mining, “Digital Twins” can become little more than a more expensive and cumbersome alternative to the Excel-Based Gaslighting (EBG) already practiced in many businesses.

If you aren’t familiar with the term EBG that isn’t surprising as I just made it up. What it is:

Gaslighting is someone trying to persuade you that up is the same as down, black is the same as white, and in is the same as out only smaller. EBG is what politically-oriented managers do when they tweak and twiddle an Excel model’s parameters to “prove” their plan’s business case.

Count on less-than-fully-scrupulous managers fiddling with the data cleansing and filtering built into their Digital Twin’s inputs so it yields the guidance the manager in question’s gut insists is right. Unless you also program digital twins of these managers so you can control their behavior, Digital Twin Gaslighting is just about inevitable.

Not that simulations, models, and/or Digital Twins are bad things. Quite the opposite. As Scott Lee and I point out in The Cognitive Enterprise, “If you can’t model you can’t manage.” Our point: managers can only make rational decisions to the extent they can predict the results of a change to a given business input or parameter. Models and simulations are how to do this. And, I guess, Digital Twins.

But then there’s another, complementary point we made. We called it the “Stay the Same / Change Ratio.” It’s the gap between the time and effort needed to implement a business change to the time the business change will remain relevant.

Digital Twinning is vulnerable to this ratio. If the time needed to program, test (never ignore testing!) and deploy a Digital Twin is longer than the period of time through which its results remain accurate, Digital Twinning will be a net liability.

Building a “Digital Twin,” simulation, or model of any kind is far from instantaneous. The business changes Digital Twinning aspires to help businesses cope with will arrive in a steady stream, starting on the day twin development begins. And the time needed to develop these twins isn’t trivial. As a result, the twin in question will always be a moving target.

How fast it moves, compared to how fast the Digital Twin programming team can dynamically adjust the twin’s specifications, determines whether investing in the Digital Twin is a good idea.

So simulating a wind tunnel makes sense. The physics of wind doesn’t change.

But the behavior of mortgage loan applicants, is, to choose a contrasting example, less stable, not to mention the mortgage product development team’s ongoing goal of creating new types of mortgage, each of which will have to be twinned as well.

Bob’s last word: You might think the strong connection to business data intrinsic to Digital Twinning would protect a twin from becoming obsolete.

But that’s an incomplete view. As Digital Twins are, essentially, software models of physical something-or-others, their data coupling can keep the parameters that drive them accurate.

That’s good so far as it goes. But if what needs updating in the Digital Twin is its logic, all the tight data coupling will give you is a red flag that someone needs to update it.

Which means the budget for building Digital Twins had better include the funds needed to maintain them, not just the funds needed to build them.

Bob’s sales pitch: All good things must come to an end. Whether you think KJR is a good thing or not, it’s coming to an end, too – the final episode will appear December 18th of this year. That’s should give you plenty of time to peruse the Archives to download copies of whatever material you like and might find useful.

On CIO.com’s CIO Survival Guide:6 ways CIOs sabotage their IT consultant’s success.” The point? It’s up to IT’s leaders to make it possible for the consultants they engage to succeed. If they weren’t serious about the project, why did they sign the contract?