HP Goes All In With An IT Transformation
Randy Mott's telling an incredibly persuasive story. Hewlett-Packard's CIO is laying out the results of a three-year journey by his team to transform the way it builds, delivers, and manages the technology that runs HP, and the results look impressive: triple the bandwidth at half the cost; 70% of employee time spent on new development with just 30% on IT support; 85 data centers consolidated to six; 700 data marts to fewer than 55; 6,000 applications to about 1,500.
In rough numbers, HP spent about $1.7 billion on the IT transformation, including construction of six new data centers. At the same time, it cut ongoing IT spending from 4% of revenue three years ago to just under 2% today, generating about $1 billion in annual savings. What CIO wouldn't love to have those numbers to tout?
Yet all I can think, as I sit at an HP event at a Miami resort, surrounded by HP customers who've come to learn what Mott and his team are up to, is this: Will even one CIO in the audience follow his advice?
That's because duplicating HP's business technology transformation comes with medicine many CIOs will find hard to swallow. On the budget front, it's capital-intensive, focusing on reducing IT staff size in part by spending on more automated and efficient software and hardware--not an easy sell in this economic environment. On the organizational front, it takes full support at the executive committee level, giving the IT team the clout to take on business units that want to circumvent centralized tech prioritization and measurement. When it's a stare-down between the CIO and the general manager of some high-flying business unit, will the CEO back the IT team? Mott's vision is a high-risk, high-reward proposition.
And here's the toughest part: CIOs must do it all at once, all in. They identify the big areas that need transforming and go after them all as one effort. For HP, there were five big areas--portfolio management, IT workforce effectiveness, world-class IT organization, global data center consolidation, and a single enterprise data warehouse. This isn't a pick-and-choose transformation, where companies do the data center consolidation now and push the portfolio management piece somewhere deep into a five-year plan. Mott puts it this way: "Choosing is losing."
The former Wal-Mart and Dell CIO says he's getting more adamant on this point now that the heavy lifting is behind him and he's starting to see the payoff. Do only part of the transformation, Mott warns, and "the parts you don't do will undermine the parts you do."
An aside: We all get that Mott works for Hewlett-Packard, and that HP sells the stuff needed to make this kind of IT transformation happen, from OpenView and Mercury software to manage and automate data centers to servers and storage to the data warehouse, a market it's trying to break into against entrenched incumbents. One of the four main goals CEO Mark Hurd set down for Mott after recruiting him from Dell was to showcase HP technology.
But Hurd's other three goals are to provide better information, lower risks related to tech failures, and lower costs. This is a CEO who says he doesn't like "binary goals"--as in, be an HP showcase or cut costs. Yes, Mott has some serious advantages over other CIOs, including a CEO who's got his back. Mott is the first to acknowledge that you're crazy to try a transformation without executive support. But this isn't a fairy tale that could happen only at a technology vendor. CIOs might find Mott's transformation pitch a tough sell, but, given his track record as a CIO and the financial results HP has been delivering under Hurd, it's one worth a serious look.
THE REVENUE OF IT
A big problem for many CIOs is that all they get to talk about are costs. They have no consistent, company-wide way to value what technology investments deliver. So a cornerstone of Mott's philosophy is to put a measurable value on the work his group does--what he calls the "revenue of IT." That's the sum of all the benefits, both hard dollar and intangibles, that a project delivers in the 12 months following full implementation.
"Every business has revenue, but IT typically doesn't ... because we don't have the discipline to capture the benefit of projects in a way that we can show the CEO or executive committee and have numbers that are real," Mott says. That's why portfolio management and prioritization were one of Mott's five big priorities. The "revenue of IT" figures have credibility with execs because they're based on a cost-benefit analysis agreed to by business unit leaders and their controllers, so they're finance numbers, not IT department numbers. "These groups are very comfortable reacting to data," Mott says. A side benefit is it potentially boosts IT morale, giving teams a credible gauge of whether they delivered what they promised and its value to the company.
But count on this: Getting to that revenue figure, and changing the company's approach to focus on that revenue, will be a battle. HP's technology organization now requires business units to fill out a cost-benefit analysis--a CBA--before any project even gets started in the company's annual planning process or any midyear project reviews.
"I had business areas that said, 'These are my 10 things, but I don't want to do your CBA,'" Mott says. "That's a real-life example: 'I know these are the right things, I've been doing this business for X years.'" That's when there's no substitute for CEO support.
And that support can't just be for the CIO. "All of us got tested," says Matthew Minetola, VP of global IT and financial services IT. Some partners in business units, he says, wanted "to go back to the way they had it before." Minetola's response is an oft-repeated expression among HP's technology team, prefaced always with a credit to Hurd: "Don't blink." No IT projects happened without a CBA--period.
The next point of tension comes in forcing business unit leaders to rank the technology projects they want. "We had a lot of people who did a stacked ranking that was 1, 1, 1, 1," Mott says.
Most companies manage only the top 10 or 20 tech projects closely, he says, and that probably represents only half of discretionary spending. Plus, those are the ones least likely to change. Hiram Davis, a manager in HP's global portfolio management office, says most of the debate at budget time concerns the "top of the below the line"--projects that just missed the cutoff for financing. The competition got tougher, too, because HP focused on doing fewer projects at a time, from 1,250 active projects to about 500. The goal is to put more resources--10- to 30-person teams, instead of two to 10 people--on fewer projects and give them shorter time frames. "It's not about doing less work," says Minetola. "It's about doing less work at a time."
THE NEW 80-20 RULE
Mott's goal is to have only 20% of IT staff time spent on support and maintenance and 80% on new project development, a reversal of the old 80-20 rule. He's not there yet--it's about 30-70 now, a reversal from the 70% spent on support three years ago.
In the process, Mott slashed the IT payroll. Three years ago, HP was paying about 19,000 IT pros, half of them employees and half contractors. Now it's less than 10,000, of which 90% are on staff. That ties to the 80% development goal: If people are doing mostly support and maintenance, it's not as important that they be employees who understand the complexity of HP's business. "I put a much higher premium on people being HP employees because of the development we're going to do," Mott says.
To know that reversing 80-20 wasn't just a theoretical exercise, HP started documenting what people work on, not what they're assigned to. Once a week, tech employees document their time by project--including if they're pulled off an assigned project to put out a support fire. It's a big effort to set up, Mott says, but "if you don't have good information on what people are doing, I don't know how you make decisions to take an organization in a new direction."
Definitions matter, too. There's only support or new development--no middle ground of "enhancements." Either it's a new project, with a priority and revenue target and business unit backer, or it's support of an existing system or application.
That discipline has helped in the often-unpopular effort to slice HP's application portfolio. Business unit leaders and employees aren't keen to hear that software that's working for them is getting eliminated for the greater good of efficiency and consolidation. Tracking the real support costs is IT's best chance. "It becomes easy to make an unpopular management decision when you have facts behind it," Mott says. "Most management decisions are unpopular to start. People don't want to change."
DATA WAREHOUSE: THE ROAD TO ONE
Data warehouse technology is where Mott's role as vendor salesman will be the most prominent, because HP's likely to be the largest customer for some time of its 2-year-old Neoview system. In a market led by Teradata and Oracle, and where HP is a nonfactor, Mott promises HP will get much more aggressive. "It's a technology whose time has come," he says.
HP has loaded 300 TB of usable data (a petabyte of raw data) into its Neoview warehouse, has registered 32,000 employees as users, and will have 50,000 by early next year, Mott says. Next up: Giving access to suppliers and channel partners, which could bring the number of users to 100,000.
HP used to have more than 700 data marts, each doing some specialized task, and that's now down to fewer than 50, mostly tied to closing the year-end books, after which they'll be eliminated. "All have death certificates that are already written out," says Mott.
His argument for a single enterprise data warehouse starts with the notion that "all data's important, and all data is affordable with today's technology." From there, he contends it's less work to put everything into the data warehouse than to pick and choose.
Mott insists he wasn't forced to use Neoview, but let's assume the corporate pressure was at least implicit. He says the technology team was told to evaluate whether HP had a marketable, scalable product, and it worked with the development team to fix any bugs. The result is a product built around the Tandem NonStop OS and database. Mott says HP's will be one of the largest data warehouses in terms of size and number of users, and one that will "not go down." Since HP's moving data for everything from supply chain and financials to customer and partner interactions onto it, there will be no hiding from any failures.
Of all the projects Mott's team has undertaken over the past three years, the most visible are the six new data centers, built at three U.S. locations, totaling 342,000 square feet of computing space. That massive project consolidated 85 large data centers and 370 smaller sites with servers. It also included what Mott calls an "interruption" of underspending on IT capital, a one-time burst of spending to get the company off old, inefficient technology.
This one seems the hardest of all Mott's initiatives for companies to replicate given current economic conditions--to spend 2% of one year's revenue (that's what HP plunked down) to refresh to the latest processing, cooling, automation, and other technologies to improve IT performance. Most CIOs already are furiously virtualizing servers in hopes of postponing spending on more computing capacity and real estate. More are buying software as a service rather than incur the capital cost for software and the servers to run them, and some are just delaying infrastructure purchases and squeezing more life out of servers. A recent InformationWeek survey of 451 business technology professionals finds that about 20% plan to cut spending on data center hardware or software, and half plan no change. Just 4% plan "significant" increases.
Mott says that's the problem--that most companies have long underspent on IT capital, even before the downturn, so they're spending too much on support, energy, and even telecom to keep older systems running. "The fundamental problem with most IT organizations is they don't have a modern infrastructure," he says. "... It's like driving a car that gets 6 miles to the gallon, like we all drove when we were kids."
With more efficient technology, HP says it cut energy costs 60%, reduced its number of servers by 40% using more powerful machines and virtualization, and cut networking costs in half.
Mott points as an example to HP's networking and telecom costs, which were driven by having to connect those 85 data centers around the world and hundreds of data marts to share information. "A miracle had to occur every day," he says of the complexity. "The bad news was that a miracle never occurred." By concentrating computing power in three locations, the server-to-server connections either run inside the data center, where there's no telecom cost for data transfer, or through the major broadband pipes between the data centers, which can be bought for lower costs per megabyte. HP's a global company, with two-thirds of its revenue outside the United States, so "we have to connect to employees around the world," Mott says, "but we don't have to connect server to server outside the U.S." The data-intensive computation work is done in the data centers, and only the results are delivered to end users.
Can other companies get the kind of savings HP did? Few IT organizations are as decentralized as HP's was, with 10 different groups operating in 100 sites around the world, so other companies might not have the same upside. Yet Mott contends most have many of the same ingredients, such as a far-flung network of legacy applications that proliferated through organic growth, acquisitions, and decentralized IT decisions.
To get a handle on the opportunity, Mott suggests CIOs take two steps. First, get an accurate inventory of "all-in costs" for IT, from equipment to people to energy to real estate. Second, get an accurate account of the active projects IT's doing, whether they're on time, and what their cost-benefit or ROI promise is. "You have to dissect the business in more detail than IT typically has," he says.
A NEW SHAPE FOR THE IT BUDGET
So does Mott's vision for transformation make sense in your company? And where are CIOs likely to meet the most resistance?
Mott's most convincing when he talks about "changing the shape of the IT budget, not just the size of it." Business technology organizations went through radical change in the last economic downturn, as large chunks of work that were done in-house were sent to offshore outsourcing companies. Yet for all the talk of moving routine IT work offshore to focus people on more innovative new project work, companies still struggle to devote enough firepower to innovation. Mott lays out a compelling strategy, heavy on IT process automation and standardization and reduction of applications.
One of the critical pieces many companies are missing is Mott's concept of the revenue of IT. The challenge will be to get business units to spend the time and make the commitment to quantify those benefits, the way HP does, in a way that has the standing of revenue.
The hardest sell of all is Mott's approach that "choosing is losing." Rather than present something as a three-year transformation, most CIOs will knock off these projects incrementally, even if they have a vision for how they'll work together. That approach will deliver benefits, Mott acknowledges, but for an IT group to change the company's edge over competitors--to "be the benchmark," as Hurd tells Mott--means taking on all IT's weak spots at once.
Understand that Mott isn't advocating a Big Bang approach that doesn't yield results until everything's done in three years. HP's effort started lowering costs throughout those three years, with benefits delivered incrementally even as IT spending was lowered from 4% of revenue to 2%. But even with that approach, CEOs will be wary of something billed as a three-year transformation that comes with a big capital cost.
Even those CIOs willing to take on a high-risk, high-reward, capital-intensive IT transformation in this economy are likely to run into an executive committee that would rather hedge its bets. No one said it would be easy.