Larry Learns Philosophy
The Oracle vision swerves once again.
Introduction
We've just come back from the sparsely attended
Oracle Applications Conference. (I measure attendance
by the number of people at the CEO's keynote--3500 for Larry, as compared with the 5000-plus at other recent ERP conferences.)
Our perception going in was that the Oracle application suite is slowly, but inexorably losing ground to the other major vendors. As far as fundamentals go, there is no reason for this. The suite is a good product with many excellent features; it should be the system of choice for the many people who like the design ideas that it embodies. But the loss of the public relations battle with PeopleSoft, the long history (still not ended) of quality problems, and the usual self-inflicted wounds have damaged Oracle enough that people lose sight of how good its product now is.
Our trip to the conference did not change this view. Things have improved (we tend to credit Chuck Phillips for this), but not enough (so far) to stem the tide. What the trip did do is
provide some insight into where Oracle's problems lie and what they might do
to change things.
As always with Oracle, one has to start with the vision. The
vision at Oracle matters; it drives behavior, explains resource allocations,
informs sales pitches. Oracle works hard on its messaging, and usually the
message has one other thing going for it: it's interesting. Larry is one of
the brightest people in the business (Chuck Philips is no slouch), so
it's always worth thinking about what he says.
This year, moreover, the vision has veered off in a new direction, a heading
that, with a little tacking, could move the company back toward reality.
I'd like to devote this
Short Take entirely to that vision; in the next Short Take,
we'll talk about
what customers and partners were saying and what is happening with the
product.
This piece was written just before Oracle announced that it was
increasing its bid for PeopleSoft. I will cover that in a separate article;
I assume, still, that the bid will not be successful.
The Single Source of Truth
So what is the vision this year, and how has it swerved from previous years?
For the last four years or so, the Oracle e-business suite has been the right
solution for customers because it allows you to run your business with a single, integrated
set of software. The appeal of this, at bottom, was that it would save you money.
With a single set of programs, you would save money on management, hardware, and
integration, and, if you did as Larry said and didn't customize, on customization.
Oracle is still selling the e-business suite, and they still claim
a TCO edge. But now, the primary attraction of what they are selling is
that the suite (or components of the suite)
provides a "single
source of truth" (SST).
This formulation should touch some nerves. As Chuck Phillips
said in his keynote, too many business decisions are made on gut
feel. The more decisions are based on facts, the more likely they are
to be good decisions.
Senior executives know this (or at least pay
lip service to this), but complain that their corporate systems can't
provide that information. Whenever there is a meeting where important
decisions are made, all the players come with their own spreadsheets,
their own interpretations of the facts, and their own hard positions.
Republicans and Democrats do it when talking about the budget, and
so do the political constituencies inside corporations.
A single source of truth can reduce the number of arguments,
weaken the defenses of the indefensible, and help everyone to
understand the facts. Not a bad thing to have.
But perhaps more important from Oracle's point of view,
this formulation allows it to expand its potential markets. It
turns out that you can have a single source of truth from Oracle
even if you don't swallow the entire Oracle e-business suite. Oracle is doing two
things, now, to make the benefits of the SST available to more companies::
- The e-business suite is being opened up, so that
integration is much easier.
- You can soon use the Oracle data model to establish an SST,
even if you're not using the suite. A new product, the Customer
Data Hub, will allow you to store or distribute master data
about customers to all your corporate applications.
There are thus three ways of getting to the truth with Oracle: buy
the whole suite, integrate to some of it, or buy the data hubs.
This may sound like a big change, but if
you listen to Larry, it's not so. A single source of truth
was always the idea behind the e-business suite, he says. The
trouble is that there
will always companies that can't or won't run everything on Oracle. So now we're doing what we can to help them have an SST.
At insurance companies, for instance, claims processing is the heart of the business,
and Oracle doesn't have a claims processing application. But with
these new capabilities, insurance companies can connect their
claims processing applications to the e-business suite modules that they
do run. Or, if they're hopelessly mired in their legacy apps,
Oracle will provide the customer data hub, so they can at least have
an SST about their customers.
Autocrats have always felt a certain latitude when it comes
to revising history, and we should not be surprised that
Larry is no exception. Obviously, as an objective account of
Oracle's application vision, this
is hooey. But otherwise, it has a certain patina of plausibility.
SSTs don't appear to be a bad thing. Increased integration capabilities
are a good thing. Software that allows you to centralize information about customers is a good thing.
It's also a relief to hear Larry
acknowledge that there are some customers can't run their
entire business on the e-business suite. This was always true
and is always going to be true. So in previous years, when
Larry maintained what appeared to be the exact opposite,
you had to wonder. Was it just that he thought the truth was inconvenient, or
was he actually deluded?
When I was listening to this, I was also pleased at the subtext.
Larry seems to have gotten interested in applications again. A year or
two ago, he thought applications were boring, a commodity. Now,
he seems to recognize that there are some interesting design problems
associated with building this single source of truth.
The Skeptics Weigh In
For analysts who have been following the e-business suite, skepticism has to be the default position. The new vision and its presentation at this conference certainly would do nothing to convert a skeptic.
According to the skeptics' view, the SST is just Oracle playing a weak game of catch-up.
- Yes, Oracle is recognizing that you can't run your entire business on a suite from a single vendor. But SAP figured that out five years ago.
- Yes, Oracle is providing integration capabilities built partially on web services. But every other ERP vendor (including JDE) has been offering integration products for years. Even Baan was introducing one with its version 6 release.
- Yes, Oracle is selling a Customer Data Hub. But isn't that just like SAP's MDM
and Siebel's UAN, which have also been released for quite a while now?
In the view of the harshest skeptics , Oracle is now a half a generation behind its major competitors. (Figure that the median replacement rate for ERP apps is about ten years.)
Indeed, if you want to push the skeptical view to the limit, you would say that you can figure out just what Oracle is going to do simply by noting what SAP did five years ago.
SAP,
after all, was the first company to sell the idea of a single
application suite running on a single instance. And it was about five years ago that SAP figured out that their product was not the
be-all and end-all and that you had to a) open it up to
other applications and b) provide integration capabilities.
On this view, the fact that Larry Ellison feels free to paint the past in today's colors is actually rather ominous. Such a person, the skeptic argues, becomes one
of Santayana's victims, people
who cannot remember the past and so are condemned to repeat it.
I have a lot of sympathy for this skeptical view, but I think it's overly simple. Yes, Larry might have done better to read some Spanish philosophy and listen to Ray Lane when he told Larry not to chase SAP.
But the product is now roughly what Larry envisioned. And Chuck Phillips
and Safra Catz have come in to replace Ray and make it work. And what Larry wants to do now is not catch-up.
Chuck Phillips Weighs In
An equally realistic, but more practical view of this new vision says basically, "Whatever happened in the past, the problem now is to make it work.". I haven't talked to
Chuck about this, but I'm going to ascribe this view to him. If apologies
are in order, I will tender them.
On this view, it really doesn't matter whether Oracle is
playing catch-up. As long as the product is in reasonable
shape, there are only two things that need to be done:
- Reconnect with business users by giving
them a message that has some business meaning.
- Improve execution throughout Oracle, especially
on the sales side.
I derive this view partly from Phillips' reasonable,
though hardly passionate, keynote speech and also from
the remarks I caught on the broadcast from Analyst
Day.
On this view, the SST message should resonate enough with
clients that it will spark interest and establish some confidence
that Oracle is still an innovator in the space. Even if it does
nothing to arrest Oracle's gradual decline in the space (and Chuck
would never, ever admit that there is such a decline), it is a good
holding action, while the fundamentals are being repaired.
Certainly, there is a class of Oracle customer who will like the vision. The Ciscos and GEs of the world, companies
that like the underlying design ideas in Oracle anyway and like
working with Oracle (more or less) are going to feel that this is a step forward. Beyond those customers with a special relationship, there will also be executives who
think a Single Source of Truth is A Good Thing. They, too, will respond.
If you take this view, the real crux is the blocking and
tackling. And in that area, there is plenty for Phillips and Catz to do.
To give you an idea of how much is needed, consider one simple fact. Right now, according to Phillips, Oracle has as many
developers working on applications as SAP does. (Let's say, about 5000.) But it has nowhere
near the license revenue that SAP has.
To beat SAP at its own game, Oracle will need to do applications better than SAP. They will have to be better at marketing, better at sales, better
at total cost of ownership, better at building applications.
Chuck and Safra will have to be the John Fox of the ERP game. (John Fox took Carolina from 1-15 to the Super Bowl in two years.)
I'm not saying it's impossible. But if you adopt this view, you have to say it won't be easy. And you may end up saying that the two new presidents of Oracle will have to be almost superhuman.
As worrisome as anything else, of course, is the fact that Oracle starts out this
new game with a major liability. They've lost the public relations
battle. In Oracle vs. PeopleSoft, Oracle is seen as the heavy, the bully, and the loser, an ugly combination. And in the apps world, they are seen as having lost importance, simply because PeopleSoft now sells more apps than they do. Blocking and tackling doesn't really do too much about this. And spending $10 billion on PeopleSoft will only make the problem worse.
Larry Learns Philosophy
It would help a lot, of course, if the vision were really good. If having a single source of truth made a huge difference to companies and Oracle applications were uniquely able to provide a single source of truth, Oracle would have a significant edge, and improvements in blocking and tackling would just help them exploit their advantage more quickly.
From what I can tell, however, I don't think this vision provides them with the edge that they need. The notion of a Single Source of Truth is conceptually flawed. It isn't just that SST is the glossy, marketing version of something that's much more complicated. It's that what Oracle (or any other ERP company) could actually deliver under the guise of a Single Source of Truth is so far from what they claim it is that its value today is pretty limited.
The basic conceptual mistake being made is clear to anybody who has taken a course or two in philosophy. There, you learn to make a distinction between "the truth" and "a representation of the truth." It is the truth that Abraham Lincoln was president of the United States, because there was a man named Lincoln and he was elected president. But the phrase "that Abraham Lincoln was president, etc." is a representation of the truth, one of many possible representations.
The information gathered in an Oracle database with the help of some version of the Oracle e-business suite is not a source of truth; it is a source of representations of the truth, representations that happen to be self-consistent.
In today's ERP systems, not just Oracle's, what these representations
give you is very problematic. You may be looking for The Truth from them,
but what you're getting is not inspired words issuing from the mouth
of a cave in Delphi. It's probably more like the ramblings of some guy
with disheveled hair who lives in a cave in the back woods.
Why is this? Well, the value of any representation depends on two things: the richness or relevance of the representation itself and its accuracy. In any ERP system, not just Oracle's, both are fairly weak. Much of the data in any ERP system is really bad, so the representation isn't accurate. And the representation itself is not really adequate, relative to what people generally expect.
Let me give you a simple example of this, one that I'll take up again in the next Short Take. In most ERP systems, there is something called a contract, which sets the
terms of a purchase order (price, credit terms, amount, etc.). ERP vendors routinely claim that they
manage contracts and that their systems can evaluate suppliers based on whether they satisfy the terms
of the contract.
As a practical matter, this claim simply isn't true, for two reaons.
First, the representation of a
contract that sits inside an ERP system is far too weak to allow
you to do contract management or to evaluate most suppliers.
The ERP system picks out only
a few things from an actual contract (the aforementioned price, etc.) and
ignores any other terms and conditions, whether or not they're relevant
to your relationship with the supplier or your evaluation of the supplier.
Second, even if your only concern for all supplier
contracts was price and credit terms, most systems would still not be
terribly useful because their data quality is horrendous. This is not
at all the fault of the ERP vendors; nevertheless, it
vitiates the value of the information that their systems provide.
Mind you,
ERP vendors are not exactly a force for good in this area: they do
a really terrible job of providing tools that help you keep data
clean and technologies for making it harder to junk data up.
Properly understood, the notion of an SST sets a fairly high bar for Oracle, which now has a good, but not great ERP system. To become an SST, they'd have to make their representations much richer and do some significant things at the same time about data quality.
To Oracle's credit, Oracle does seem get some of this. They clearly have some appreciation of the problems of data quality and representational richness. And they are doing something about them. The Customer Data Hub, it turns out, is based on what they
call a Trading Community Architecture, a representation of the customer relationship that is designed to be much richer than what is normally available in an ERP system. (Iit is this architecture
that you're really buying when you buy the Customer Data Hub.)
There is also something being done about data quality. In
a sense, the Customer Data Hub is precisely a technology for
keeping your data from getting junked up. But more than that, several speakers
made it clear that the hub has tools that help "data librarians" clean up the data or keep it clean.
Equally encouraging in such a top-down company is the fact that Larry seems to have learned something about these problems and enjoys the challenge they present. In his talk, he even told an amusing story about the changes they had to make in the e-business suite in order to improve what turned out to be inadequate "semantics."
If this vision keeps on operating within Oracle and keeps on pushing Oracle to improve its representations and data quality, it will be a good thing for customers.
But it's unlikely. At this moment in history, most corporations
don't appear to be too worried about the fact that their
systems provide a rambling, incoherent, and often incomprehensible
version of the truth.
You see this most directly when you talk
to somebody who is responsible for data cleanup or data synchronization.
You also see it if you talk to people who sell data cleanup services
or data cleanup tools. They've got hold of a legitimate problem. But
their customers would rather invest money in new systems than invest money in cleaning up what they already have.
In most corporations, moreover, there are always large constituencies that
are well served when access to the truth is not easily available.
There is a reason why data is locked in silos.
You even saw the low value put on truth at the conference. Oracle trotted out two customers who were early adopters of the Customer Data Hub. So far as I can tell, if you want to use the hub to get the truth, you must use data librarians. (Somebody has to operate the tools Larry is providing.) But neither user had them, and the processes they had put in place instead appeared to be inadequate.
It may not be simply that people don't want the truth. It may be just that getting there is pretty challenging. If you are a buyer of Oracle and you want to get at the truth using their systems, you are going to have to do a lot of head-banging and a lot of cleanup before you have a chance at getting what you need. It's a big project, and in most companies, such a big project is not a priority.
And remember, even if you do want it, the Oracle system still needs a rich enough system of representation. For many companies, that's still not the case. I talked to one company that bought Oracle's "e-business suite does all you need" message three years ago. Unfortunately, the fit for their business is poor. So far, the IT guys said, we've spent $10 million figuring out a way to make it work. We should have the design approved next month.
Does Oracle really have the commitment to build up its systems and
offer the improved ways to manage data quality that its vision probably requires?
If customers prove to be tepid about it or don't have the resources, what's the point?
Besides, going to all the work of making your system genuinely robust would be a pretty high-risk thing to do. Most customers and most analysts would be unable to value what you've done. To return to my example, what would happen if the e-business suite really did an excellent job of contract management. Oracle would be up against any number of vendors who claim (falsely) that they've been doing contract management for years. It's easier just to say that you're doing contract management and put together a few Power Points.
So, even i f Oracle sorts out the conceptual issues, puts in an appropriate set of resources, and gets serious about the Single Source of
Truth, they'll have to make a compelling case to customers that it's actually better
to have the truth.
Oracle as a Test Case
Oracle does have an opportunity to do this. There is in fact one company that is deeply committed to using Oracle applications to provide a single source of truth. Oracle.
If Oracle can actually use its access to the truth about customers and about its own operations to become materially more effective than all those other ERP companies and can sell people on the notion that what Oracle did can be applied to them, they would have a case.
Cisco did this to great effect during the go-go years. By being an effective Internet company, they sold a lot of people who wanted to become like them. Of course, much of this ended with the dot-bomb.
So again, maybe it comes down to blocking and tackling. If, for instance, Oracle can use the information they have in their customer database to define who are the most likely prospects and what products those prospects are most likely to buy, it will be very impressive. Oracle will be able to lower its cost of sales by quite a bit. And Oracle won't just have a better margin than anybody else. They'll have acceleration in apps license sales that far exceeds everyone else's.
So far, this hasn't happened, but given the quality of the systems Oracle has been using, I'm not surprised. As I said, the e-business suite doesn't start out being the best system out there when it comes to being a representer of truth.
Again, let me try to give a simple example. One of the things you want to do when you're building a system that is a source of truth is to recognize that data is time-bound and that the value of data decays over time. In systems like PeopleSoft's, a recognition of this is built into the representational architecture, because PeopleSoft has an as-of date recorded for every important piece of master data. That means you can date information accurately; you can change it when the information changes without losing the record; and you can even track information history. Not so in most other systems.
All this can be changed, of course. Oracle does have 5000 developers and a vision, anyway, that could take them toward significant improvements in an already competitive product.
But I can see why the skeptics don't even want to be bothered with such a charitable view. Yes, this could be something terrific. But so often, you have to take what Oracle says with just a little salt.
Let me close with an amusing example of this. Oracle always looks for "Oh, wow." statements, and at this conference, one of the "Oh, wows." was the advantages of moving to Linux on low-cost servers. Several times, Larry told us to do what he did and move to $6,000 dual-processor Dells for his middle tier servers. "$6,000 apiece for two of these machines," as opposed to, let's say, a $75,000 dual processor Sun. "What's not to like about that?"
Well, I'll tell you. As it happens, I talked to somebody who had done just that: dumped
an $80,000 Sun dual-processor and bought two $6,000 dual-processor Dells to run Oracle on.
And here's what's not to like.
It turns out that you also
need to buy RAC in order to use the two extra processors from Dell. And RAC costs
$20,000 per processor. (I'm using list in all cases.) The savings was not $68,000. It was roughly $28,000, much of which was used up by the conversion.
The story gets worse. Unfortunately, his boss
didn't have a sense of humor. So over one weekend, the team had to disable one
processor in each Dell machine. In doing this, somebody made a mistake. So
they lost all database service for 19 hours, in a very high throughput environment.
Somehow, the words, "Oh, wow." never seemed to occur to him when he was describing this.
To see other recent Short Takes, go to our summary page.
| 
