July 14, 2010
Who is to blame for IT project failures? My colleague, Michael Krigsman, argues that when IT projects wander into the “IT Devils Triangle,” all three participants–the vendor, the integrator, and the customer–are to blame. Michael is very insistent about this; in a recent post on Marin County v. Deloitte, he says, “In my view, it is highly unusual for a project to fail without some culpability on the part of the customer.”
Michael is the guy who almost singlehandedly took IT project failure out of the closet and into the open, and much of his professional life is spent analyzing these failures. For that reason, one should not criticize what he says without good reason. But I’ve always been uncomfortable with his line on things and felt that we would all be better off if we had better analytic tools for this kind of situation. I want to know when the “culpability” is minor and merely contributory and when it is really important or even primary.
One’s normal intuition about such things is that there are broad spheres of responsibility. If I get an artificial hip implanted, it’s up to the manufacturer to provide my doctor with a hip that works, up to the doctor to put it in correctly, and up to me not to abuse it. This is the intuition that underlies a lot of contract law and liability law, and it even applies to very large, complex engineering projects. When the architect found a flaw with the Citicorp building in New York City, which flaw could have caused it to collapse, the architect and the contractor took responsibility for it and had it fixed. They didn’t blame the customer, and they didn’t ask the customer to adjust his use of the building.
If you apply this intuition to technology, it works pretty well. In the case of IT projects, the software company is responsible for making sure that the design and building of the product is adequate to the demands that will be placed on it, the integrator is responsible for deploying the product in a way that meets what one might call the normal expectations of the customer, and the customer is responsible for using it in ways that are consistent with the design. If any of them fail to meet their responsibilities in their sphere, then they are culpable. And the others are not culpable (except perhaps in a minor way) if they don’t wish to compensate for that failure.
Now, I think Michael would agree with this, but he would say that, as a practical matter, what happens in these large project failures. is that all three people fail to deliver, so all are culpable. (Michael, please correct me if I’m wrong.)
But it seems to me that doesn’t go far enough. I think you need to be able to figure out when one participant is clearly morst culpable? Perhaps, for instance, one ought to apply a notion of priority, so that if the vendor fails to deliver, the vendor is primarily culpable by definition, and the most that the next two parties can do is provide some minor contribution. (This tends to be the approach in products liability, with some major caveats.) Or perhaps there are some other tools that need to be applied, tools that would allow one to extend (or perhaps even change) the “pox on all their houses,” line that Michael takes.
Whatever you do, though, it isn’t easy.
Let me try to illustrate what I mean with an example that has made a lot of news recently: the iPhone 4. Here you had three participants: Apple, AT&T, and the customer. You had a product that “did not work properly.” And you have a dispute about levels of culpability.
The problem, in case you’ve been in Tibet for the last month, is that the iPhone’s reception isn’t very good for reasons that still aren’t clear. (My wife has one. It’s true; it’s not not good.) The customers have basically been told by Apple, “The problem is in the way you hold the phone. Either hold the phone differently, or fork over $25 for a case.” The customers have said, if I may paraphrase, “It’s a phone. You should be able to hold the phone any way you want. It’s up to Apple to give me a phone that works no matter how I hold it.”
Now, the IT Devil’s Triangle analysis says, basically, that the customers are wrong. According to this argument, Apple made a good phone that made some design tradeoffs. (Basically, the antenna was put on the rim of the phone.) If one of those tradeoffs means they have to hold the phone in a special way, so what? They should either buy themselves a case or hold the phone correctly. If they don’t want to do that or can’t, they are contributing to the problem and are at least as culpable as Apple.
So is AT&T off the hook, here? Under the Devil’s Triangle argument, not at all. The phone service that they provide is weak and erratic, and that makes it very difficult to troubleshoot the phone. (I have experienced this in spades.) Their customer “service” is designed to handle normal problems expeditiously, but is not designed to track and resolve complex problems, and as a result, it makes what could be an irksome experience into something verging on horrible. (I have also experienced this, believe me.) And that means that customers are not as tolerant of the iPhone 4′s design choices as they should be.
It’s something of a conundrum, isn’t it? One’s normal intuition is that Apple should design a phone that people can hold any way they want. But one can certainly make a cogent argument that it’s really the customer, as much as anyone else, who is to blame, because they get mad at Apple, rather than being willing to hold the phone in the correct way.
In the end, the solution to this riddle is a matter of values. If you think (as Consumer Reports does) that it’s up to the vendor to get things right, when it comes to simple things like how one holds the phone, then the Devil’s Triangle argument is simply wrong. If you think (as Steve Jobs and many of my friends who work for vendors think) that the customer is to blame if they can’t deal with the “flaws” that they find in a complex and highly-engineered technology, then you’re going to agree with Michael and assume that there are very few situations where the customer doesn’t have some culpability.
Let me just tell you where I come down on this and hope for some illuminating comments. I think the benefit of the doubt should be given to the user. If the vendor is clear about the design tradeoffs and limitations and the deliverer (integrator) provides service that takes those limitations adequately into account and is clear about that, then I think, “Yes, it’s up to the consumer to deal with the limitations.” If the vendor and deliverer fail to do this–if the vendor isn’t clear about what the design tradeoffs are or if the vendor and deliverer allow you to get the idea that the product will do things (like make phone calls in normal situations when being held normally) that it won’t in fact do–then the onus is on them.
So, for instance, if the building architect makes a mistake in the way the building was designed or the general contractor deviates from the design in a way that turns out to be dangerous, it’s up to them to fix it; the building owners are not culpable because they don’t want to limit the number of people allowed on each floor to a number far below what they’d been led to expect. And if the hip designer uses a brittle material and the doctor chips it when he or she installs it, it’s not up to the patient to walk less.
Just my view.
So why is this important? It’s important because this is not how the software industry works. It’s normal for vendors to be very close about the design of their technology, for vendor and integrator salespeople to make unrealistic claims about what the technology does or about the benefits to be received, for the people delivering the product to view limitations in delivery as upsell opportunities, etc., etc. So if my view were to prevail, and it were up to the vendor and deliverer to make sure that the product works as you would normally expect and the installation is what it should be, the industry would have to change.
Is that likely? Maybe, maybe not. But it gets more and more likely every time a user asked to adapt to some software’s vagaries asks himself or herself, “Isn’t this like being asked to hold an iPhone with tweezers?”
April 9, 2010
In yesterday’s post, I argued that ROI would not be an adequate measure of the benefits conferred by new-gen (or pseudo-new-gen) applications like Workday, Business By Design, or Fusion Application Suite. The previous-gen applications were all about automation. The new-gen suites confer real benefits (I think), but not necessarily benefits that fall through to the bottom line.
What benefits are they? Well, they have to do with working more effectively: making fewer errors, putting more time into work and less into busy work, making more accurate decisions, faster. Is there benefit from this kind of thing? Sure. But how do you measure it.
In the post, I suggested a hazy term, “operational effectiveness,” for the benefits one should expect. What is “operational effectiveness?” Let me admit freely that I don’t know for sure. In this post, let me propose an analogy, which should help you to understand what I’m getting at.
The analogy comes out of a historical situation that always posed a problem for ROI analysis, the transition in business from the typewriters that sat on secretary’s desks to the PC that sat on executive’s desks. This transition occurred in two different phases. First, the typewriters on the secretary’s desk were replaced with big, clunky word-processors that sat next to the desk. These word-processors automated the secretary’s document production work. Then, the secretarial position itself was eliminated, and the typing function became something that executives did themselves on that PC.
The transition to word processors could easily be justified in ROI terms. We could get more work out of the secretary or else hire fewer secretaries. Whether the justification was real is an open question. But it’s certain that that’s how people thought of it.
The next transition was much more problematic for ROI analysis. Expensive executive time was now being put into jobs that had been performed more efficiently by much cheaper labor.
At the time, people didn’t put a lot of thought into figuring out why they were funding this transition. Executives saw the PCs, knew that everyone else was using them, needed them for some functions (e-mail, spreadsheets), and just decided. “We’re doing it this way.” At least in my recollection, that’s what happened.
So were they just loony or lazy or wasting shareholder money on executive perks? I don’t think so. I think what they were plumping for was the same “operational efficiency” that I’m talking about 25 years later.
True, they spent more time typing. But they also had more control over the final product; they could change the product more easily; and they could distribute it without much overhead. And, at the same time, they were changing the form of what they were doing. They weren’t just producing typed memos; they were documents with fancy fonts and illustrations; and they were creating Power Points. True, many an executive was spending ridiculous amounts of time fiddling with type sizes so that they could get things on one page, but even acknowledging that, they thought the new way was better.
Indeed, by the time the transition was finished, justification wasn’t even a question, because the new tools changed the nature of work, and now you couldn’t get along without the tools. When executives were doing the typing, they stopped creating long reports. More and more of the time, a corporation’s decision-making was even wrapped around a full document (minutes, memos, or formal reports), it was wrapped around Power Point decks.
So by the end of the transition, ROI analysis had become entirely moot. How could you get a tangible measure of benefits when you were comparing apples and oranges?
Could we be seeing a similar transition now? It’s certainly possible. The analog to the word processors is that first generation of enterprise applications, which were funded by the automation benefits they confer and by ROI analysis. The analog to the PC is the second generation of enterprise applications.
(One caveat. As I’ve said before, I don’t think that Fusion Applications or the versions of Business by Design that I’ve seen are in fact second-generation applications. They’re more like Version 1.3. But they’re close enough to next-gen to raise the problem I’m talking about.)
If the analogy holds and if second-gen apps work as the developers hope, the benefits that businesses are going to experience will be equally hard to get your arms around, partly because the benefits are so subtle and disparate and partly because you’ll see a shift in the way work is done.
Does that mean that we won’t be able to talk about the benefits and we’ll just bull ahead with them? Well, that’s why I’m introducing the notion of operational effectiveness. It does seem to me that we can get clearer about what the benefits are.
So come on guys. Make comments. What is operational effectiveness? And how can we tell whether we are getting it?
April 8, 2010
This blog post is as more an open question than a pronouncement. So please feel free to comment on this or take the idea in a different direction.
I’ve been thinking about the next generation of Enterprise Applications, the value that they (might) bring, and about how people might justify replacing the enterprise applications they have with the new generation.
Generally speaking, you justify an investment in infrastructure using ROI. You invest this much, get this much return. For the first generation of enterprise applications (most everything designed between 1990 and 2003 or later), this made sense, because they were basically automation apps. They automated work done by people. The ROI showed up because you didn’t have to pay people to do it any more.
Now these new applications simply don’t do that, that is, they don’t automate appreciably better than the old applications do. And this means that ROI is a pretty crummy tool for evaluating whether an investment is worthwhile. Yes, there will be ROI if the enterprise application works the way it’s supposed to. But the return will be highly indirect. You won’t be able to fire people and pay for the application.
Brian Sommer has been talking about this problem for years–essentially, he points out that automating something that’s already automated doesn’t justify an investment on the same scale. But he has never really explored whether there are other forms of justification.
Nenshad argues in his book and his blog that good performance management enabled by modern tools will get you to a place you want to be, and he tells you a lot about how to do it. And while it is true that these new applications help you manage performance better and that’s one of the reasons you want them, he doesn’t offer a way of thinking about justifying the move to what he recommends. (Nenshad, if I missed this in your book, I’m sorry.)
So what does one use? Well, let me offer a concept and sketch it out in a paragraph or two, and you my small but apparently very loyal readership can then take me to task.
I’ll argue that what these new applications really do is improve “operational effectiveness.” What’s that? Well, to start out with, let’s just say that they let each employee put more effort into moving the company forward and less effort on overcoming friction, that is.
Well, that’s suitably hazy. So here are some things that you could measure that would, I think, be indicators that employee effort is more coherent and focused. You could, for instance, take a page from the black belts and measure operational errors or even just exceptions as part of operational effectiveness. Or, you could look at corporate processes that are nominally automated and see whether they are managed by exception. (Truly best, automatable practices should require almost no routine manual actions.)
People sometimes try to look at operational effectiveness by measuring what percentage of revenue is spent on things that feel like pure expense, like IT. So, a company that spends 3% of its revenues is less effective than one that spends 1% of its revenues. People also try to get at it by trying to look at which activities are “value-added” and trying to get people to do more of them. Both ideas are silly, of course, in themselves. (The 3% company may be spending on stuff that makes them effective, while the 1% isn’t.) But I think there might be indicators of operational effectiveness that are better. Wouldn’t operationally effective companies spend less time in meetings, send fewer junk e-mails, work fewer hours/employee (!), resolve more customer complaints and deflect fewer, etc., etc., etc.?
You get the idea, I think. So why is this a good measure for the new generation of applications? Because, bottom line, I think that’s what they’ll do. They’ll help organizations and people avoid wheel-spinning, error correction, and pointless processes or rules by getting to what matters, faster.
Of course, people have always accused me of being a ridiculous, blue-sky, naive optimist. But that’s how it seems to me.
What do you think?