Service with a Smile–and a Wink and a Nod
February 20, 2006 Timothy Prickett Morgan
The AS/400, the iSeries, and the System i5 have a problem, and it is a big one. The problem is not any particular hard or soft ware. It is not that it supports legacy applications as well as modern ones. The problem is perception, both in the IT market at large and within IBM itself. At IT Jungle, we spend a lot of time dissecting and discussing the former, but because we are not inside the IBM Company, we can’t really do much about the latter.
Here’s the problem, in a nutshell: In the last decade, IBM has transformed itself from a hardware company that makes a little money on services and software into a services company that makes a little money on hardware and software. While IBM has been innovative in how it has shifted around hardware and software costs inside the AS/400, iSeries, and System i5 platform–breaking the line into systems with 5250 capacity and servers without it, then moving to interactive processing features, and now having Standard Editions and Enterprise Editions of software–the product is still largely a hard sell with a big price tag, and it has virtually no services component aside from Software Maintenance, the fees IBM charges for tech support and software updates on the i5 platform.
A platform with no services revenue potential is an orphan at IBM. If we want IBMers to champion the System i5, we have to convince IBM to change the packaging of the i5 to include lots of services, and thereby change the perception of the box to being good for all of IBM–its hardware, software, and services units–and a weapon with which to fight against IBM’s many competitors in the market.
IBM’s shift to services has been dramatic, and you really need to see the numbers to get a grasp of the magnitude of the problem. While IBM had an arm’s-length subsidiary called ISSC to do services since it signed a consent decree in 1956 to settle an antitrust lawsuit, in the early 1990s IBM was able to wiggle out of parts of that consent decree and re-absorb its services unit, thus establishing what would become the Global Services behemoth. Global Services united all of the services units IBM had running around the world in different geographies, including those that did break-fix maintenance as well as systems integration, outsourcing, and other forms of consulting.
Back in 1992, which is as far back as I can go in the online IBM annual reports, Big Blue had $5.5 billion in services revenues, plus another $1.8 billion in its Federal Systems Company unit (which did systems integration for the U.S. government) and $7.6 billion in hardware and software maintenance revenue, yielding just under $15 billion in total services sales. IBM had $33.8 billion in hardware sales, $11.3 billion in software sales, and a total of $64.5 billion in sales. The company booked a net loss of $5 billion, and sales shrank even further in 1993, and IBM booked a staggering $8.1 billion loss. IBM had a loss of nearly $2 billion in 1991, so losses for those three years came to over $14 billion. From 1990 through 1992, IBM slashed about 75,000 employees from its payroll, which was 375,000 strong at the end of 1990, and when that didn’t help the company turn itself around, John Akers lost his job as chairman and CEO in 1993 and Louis Gerstner left his job at Nabisco to take over the company. He proceeded to slash another 75,000 jobs.
One of the first things Gerstner did, ironically, was sell off the Federal Systems unit to Loral, thereby boosting services gross profits to 20 percent in 1994, an increase of five points even as revenue stayed flat in services (excluding maintenance). Software gross profits were nearly 65 percent and hardware gross profits were about 42 percent in 1992. These were better businesses to be in, to be sure, except for one thing: There was no revenue growth potential. And so, Gerstner focused IBM on becoming not just a computer company, but a services company.
Fast forward to year-end 2005. IBM is now a $91 billion company, making respectable if not insane profits, and it has 315,000 employees after shedding about 15,000 employees in a layoff last summer. About half of IBM’s employees and a little more than half of its sales come from the Global Services unit. Specifically, Global Services, while under pressure as companies transition from long-term, blockbuster contracts to smaller, short-term services deals, raked in $47.4 billion in sales and had gross margins of 26 percent. IBM’s hardware sales are way down since it exited the disk drive and PC businesses, to $24.3 billion, but doing so raised gross margins by five points to 35 percent. Software has grown to $15.7 billion and has a whopping 87.5 percent gross margin. Here’s the important thing to realize: Even with all of those people-costs in services, the Global Services unit delivered $12.3 billion in gross profits, which was on par with the $13.8 billion in gross profits of Software Group. IBM’s overall hardware sales generated a meager $8.5 billion in gross profits. Back in 1992, IBM sold close to 40 percent more hardware and had about the same amount more in gross profits. The aggressive price cutting in the PC and disk businesses, which IBM founded, made them useless as far as IBM’s bean counters were concerned.
Which brings us to the OS/400 platform. In many ways, the OS/400 platform has a negative services potential, and it represents a snapshot in time of the IBM Company in the late 1980s.Not technically, but financially. IBM’s big push in the services business since 1993 has been to help companies adopt distributed Unix, then Windows, and now Linux systems and to charge for the integration and operation of these systems and the applications that run on them. Because these systems are complex, and so are the applications that run on them, this costs lots of money.
An AS/400, when properly used, is a network in a box. It can house every infrastructure tier–database, application, and presentation layers–and integrates well with other architectures when necessary. It is designed to be assembled relatively quickly and easily. To make matters worse, the 200,000-plus OS/400 shops in the world overwhelmingly prefer to either write their own applications or to do their own integration. They do not, as a rule, go for outsourcing, since they prefer to run their own data center operations. There just isn’t a lot of services here, and even though the software content is now pretty high (in terms of percent of revenues in a total system sale), the volumes of the AS/400, iSeries, and i5 machines are so low that the software stream from the OS/400 platform is utterly dwarfed by the software IBM sells to Windows and Unix shops.
The OS/400 server business at IBM has shrunk in absolute terms by a factor of three since it peaked at around $6 billion in 1991–including servers, peripherals, systems software, and storage. Unit volumes are down by a factor of two in a broader server market that has seen volumes explode by a factor of three or so. The OS/400 has, by comparison, shrunk by a factor of nine or so compared to the broader market.
As I have said many times before, I believe that IBM is not only charging too much for OS/400 servers, it is punishing customers who love the 5250 protocol, which they understand, no matter what pretty GUI face they want to put on the screens. IBM is charging at the front end of a system deal for the familiarity of an OS/400-RPG stack, and it is charging a lot more than the broader market would tolerate. If that were not the case, IBM would be selling close to 200,000 OS/400 servers a year, would have an installed base of about a million customers, and generating something more like $7 billion a year in sales. It would still have 15 percent of the server market’s revenue, as it did nearly two decades ago.
The OS/400 platform has generally been on the front wave of many innovations in the computer business, but it has been a serious laggard in the financial packaging of the box. If OS/400 looks strange to customers who know Windows, the pricing and salient characteristics of the i5 line are mind-numbingly complex. This has to stop, and I think I know how to change it and move to a sales model that is more amenable to Software Group and Global Services. The System i5 needs to shift to a combination platform sale-services sale.
Imagine, if you will, a cable company that operates the way IBM operates the i5 platform. First, if you wanted to install cable, IBM tell you that you would have to pay, up front, for the full installation costs of the cable wire strung from the central office to your home. Then it would tack on the full cost of the set top box and the back-end systems required to send you your signals. This would be an integrated system, you see, from end to end, very reliable and very familiar to you. So you wouldn’t mind spending thousands of dollars to have the line installed. Then, if you wanted basic service–the kind of channels you have always watched–you would have to pay an exorbitant fee. But if you wanted premium channels like HBO and Showtime, these would be very inexpensive, and if you wanted a box that only played premium channels without the basic service where you spend most of your time, then it would be even cheaper. So instead of paying $80 a month for cable, in this model, you would be socked with a bill for $4,800. If you wanted Internet cable service, the same rules would apply. Instead of paying $50 a month, cough up another $3,000. And you want VoIP? That will be another $2,000, please. And you also have to pay $40 a month to help us keep the system software that runs the cable system up to date, and we make you buy it in a 12-month block, so add another $480. That comes to $10,280. But the installation and integration would be free, so that’s a pretty good deal, right? You ready to sign up for cable yet?
This is not, of course, how the cable business operates. The basic premise of the cable business is ubiquity. Since everyone is presumably going to want basic cable service (especially after you stop using transmission signals over antennas because you have a strong lobby in Washington), you can amortize the cost of the cable infrastructure, the set top box, and the back end systems that serve up the signals. By law, the basic service has to be provided–for free. Premium channels cost an incremental amount, as does the addition of Internet and VoIP service. When you are done, you end up paying the same $211 a month for the whole suite of services and ongoing support, but the way you feel about how you spent it–and the way it goes on your bookkeeping–is different.
This analogy isn’t perfect. But imagine if the System i5 platform was priced based on the assumption of ubiquity, not scarcity. Get Windows ported to it, get MVS ported to it, and you already have OS/400, AIX, and Linux ported to it, and just let people buy what they want and run what they want. No governors. No overcharging for basic service–yes, that was 5250 green-screen processing in my analogy, and Java was a premium service.
Now, imagine further that IBM made the box as inexpensive as possible–or even sold it at cost to customers. And sold a wide range of performance and price points. And let any workload run as fast as it could on these machines. IBM has to have some hardware money to appease the Systems and Technology Group, and some software money to appease the Software Group, and some services money to appease the Global Services group. To make it simple, match IBM’s own overall revenue distribution: Take the total charge of the i5 systems that we have today, add up the number for acquiring and supporting a box with all of the stuff IBM makes–and I mean all of it–and just cut it 25 percent for hardware, 15 percent for software, and 50 percent for services. Don’t worry about what those services will be yet. Just trust me.
Say an i5 solution costs $500,000 over five years. That makes the hardware piece of the deal–the upfront cost–$125,000. The systems software stack–i5/OS, DB2/400, WebSphere, the compile set, the SQL tools, every little and big thing IBM sells in a standard system, including Notes/Domino and other middleware–is only $75,000 of the stack. The services component, which you are going to spread out over five years, is $300,000, or $60,000 a year.
The tough question is, what are those services, and why would anyone spend money on them? Well, there are all of the capacity on demand technologies, so say IBM rolls in enough capacity on demand to deal with peak processing needs at the end of the week, month, and year. Call that $10,000 a year. That leaves another $50,000 a year. While green-screen capability is free, maybe WebFacing and Host Access Transformation Services–for the love of God, it has services in the name!–are no longer products that are “free” with WebSphere Development Studio, but are charged for on metered usage. If you want pretty screens, you pay a service to use WebFacing and HATS. This seems reasonable. It is like paying for a premium channel (or, if the analogy held, it would be like paying a little extra for HDTV over basic analog or standard digital service). Call that another $10,000 a year. Now we have to scrape up another $40,000 in services. Well, there is break-fix service, but that is only after a three-year warranty runs out, so that is really only in the last two years of the deal, and maybe only worth a total of $5,000 a year because of the reliability of the components. So that only shaves $2,000 per year out of the total $500,000 for the solution. But it does get us down to $38,000 in services revenues we have to come up with. Now, let’s get crazy. Remote backup services and high availability services. From now on, IBM Global Services and the iSeries HA partners promise to work together to deliver remote backup of data and high availability clustering–not as a product, but as a service–for a total of $35,000 a year. It has proactive management, and IBM hosts the hot standby box in one of the zillion data centers. That leaves $3,000 a year for proactive capacity planning and other tech support, which could be performed by the vast network of iSeries resellers. Give them the break-fix money, too, and let them manage that problem. Give them a commission on sales, and a piece of the services revenue stream over time. Be innovative and generous: If the resellers find new customers, they get a bigger slice for the first couple of years.
The trick is to create services that we all know people want, and to price them on a System i5 based on the assumption that everyone will want them, not just a few thousand shops. If you price HA services so the entire base can afford them, let me assure you: the entire base will buy them. And I think the HA vendors would be better served if IBM sold a lot more machines with HA software on it. Imagine 200,000 System i5 boxes a year, people. If the HA service component of the iSeries system ranged from $3,500 a year on a $50,000 i5 solution to $350,000 a year on a $5 million solution, and the normal distribution of i5 systems prevailed, you would probably be looking at somewhere around $10,000 per year per server. Even if volumes only tripled to about 60,000 units (back to the old normal from the 1990s), that is $600 million. I think the six major iSeries HA vendors and IBM can build a very sophisticated HA recovery operation with this kind of money. If they can’t, I know some people who can.
Now, with this scenario I am painting, customers get to use as much or as little green-screen capacity as they want, Java and WebSphere or not, and have a whole line of services that they obviously need, they have always wanted, but they could never afford. Most importantly, the entry price for the hardware and software has come down by 60 percent, Global Services has a vested interest in promoting the System i5, and IBM gets the exact same amount of money in the end. But the way the customer feels about this deal is very different.
For one thing, the System i5 platform I have described would be priced exactly like an X64 server running Linux and an open source software stack for low-end boxes and like Unix machines for high-end boxes. And packaged the same. If Linux licenses cost $300 for a two-socket, four-core server, then i5/OS does, too. If virtual machine partitioning costs $5,000 on such a machine (as it does from VMware), then the Virtualization Engine does, too. And so on for the database, for the middleware stack, etc. Moreover, this way of doing it shifts $300,000 out of the capital equipment budget that has to be depreciated over time and into the operating expenses budget.
The AS/400 used to be a leader in technology, and in many ways, it still is. It is now a time to be a leader in pricing and packaging, and to throw out the way the product has been diced and sliced. We tried it IBM Somers’ way, and it didn’t work. The AS/400 has been cut up so much, a traveler from 1988 would hardly recognize it.
Keep it simple. There are too many features, too many different ways to build an i5, too many conditions. No one can remember this stuff, no sales person can explain it, and no one has time to listen to it when they are buying.
Be generous. Take the governors off, once and for all. RPG and 5250 are basic services, and maybe Java has to be as well.
Think services. Sell people services they want, not high hardware and software bills that make them move to other platforms.
Stop worrying about what to call the box and start worrying about redesigning the way it is priced and packaged and sold. Turn the iSeries into a pull box, where people want to acquire it, from a push box, where you have to convince them to buy.
Compete with Lintel and Wintel platforms on the low end, and beat them. Compete with Unix machines in the midrange and high end, and beat them. GHz for GHz, GB for GB, feature for feature, dollar for dollar. No holds barred. Stop explaining why the current strategy is right and start discovering how a new strategy–one that engages all of IBM–is better for Big Blue, the vast OS/400 installed base, and the as-yet unsold tens of thousands of OS/400 shops we want to have in the fold in the coming years.