While operating with a profit-or-loss mentality can offer leadership opportunities and a drive for efficiency, there are several problems with a full-shared service model.
Several years ago, CIOs were admonished to build IT into a shared service, essentially creating a "business within the business" that would "sell" IT services to other business units and attempt to do so at the lowest possible cost. This strategy had some appeal. Presumably if IT was operating as a stand-alone business, it would be encouraged to streamline and operate in a cost-effective manner and would be available to anyone and everyone with a signed check. While operating with a profit-or-loss mentality can offer leadership opportunities and a drive for efficiency, there are several problems with a full-shared service model.
The ultimate commodity-shared service model is outsourcing
The most obvious problem with a shared service model is that it strives toward commoditization. If your "business within the business" had to maintain every potential competency, from maintaining desktops and network infrastructure, to having a raft of skilled implementation experts, there is no way it could be cost competitive. In this environment, your shared service tends to deliver primarily common commodity services. While there's nothing fundamentally wrong with this scenario, in even the largest organizations there is almost always an outside provider that can offer those same commodity services at a fraction of the cost, unless your company is in the IT services business.
While there's nothing wrong with outsourcing per se, from a perception perspective, the ultimate IT shared service is cheap and "out of sight, out of mind" until called into action. If you're a CIO striving to get a seat at the executive table and present IT as an entity capable of delivering on the organization's strategic and tactical imperatives, cheap and "out of sight, out of mind" are certainly not the fastest routes there.
Low cost usually equals lowest common denominator
Once you've built your shared service center and slowly optimized for commodity services, you're likely to be forced to exit the "solving complex problems" business. It's too expensive to keep high-end project managers and internal consultants on staff, and these needs will be fulfilled by outside vendors. There's obviously a balance since no IT organization can or should staff every conceivable skill internally, but a shared service mentality drives you too far in the opposite direction. In effect, the better you become at delivering a low-cost commodity, the more irrelevant an internal IT organization appears.
End the gymnastics
Perhaps the most wasteful aspect of an IT shared service is also one of its core concepts: chargebacks. Since a shared service is supposed to operate like a business within a business, other business units must generally "pay" internal IT for the services they consume. While conceptually straightforward, these exercises usually result in weeks (and in extreme cases, months) of accounting gymnastics, which at the end of the day do nothing more than move the company's money from the left pocket to the right. Most of these calculations do little to reflect market realities and occasionally have the unintended consequence of making internal IT appear grossly overpriced compared to the market, furthering the case against maintaining internal IT. In the worst cases, I've seen companies divide an employee's "billable hours" over the total cost of that employee (salary, benefits, paid time off, etc.), resulting in thousand-dollar internal bill rates for skills an average external vendor was charging $30 for.
Furthermore, it's tempting for an internal shared service to abuse what amounts to a monopoly position, again strengthening the case that there's no need whatsoever for internal IT. Put yourself in a business unit president's shoes and the case is pretty obvious. Pay exorbitant rates and sit through weeks of meetings with finance to engage in some accounting gimmickry or cut a check to an external vendor at a highly competitive price and be done with it.
There's a valid case for chargebacks in areas like a project environment, where providing a "blank check," sponsored by IT, could result in unending scope additions, but these scenarios should be the exception rather than the rule.
While a full-scale shared service model for IT is not a good idea, there are some elements to the model that can be beneficial. There will always be a commodity element to corporate IT, and this element should be delivered quietly, efficiently, and cost effectively. There's no shame in letting your network ops or desktop unit act as a "business within IT," striving to do things better, cheaper, and with less management oversight. Similarly, providing IT consumers with some idea of the cost of the services they are using is beneficial. I prefer a general menu-style approach, where everyone can gain an approximate understanding of what a particular service costs the company and which business units are the primary consumers of that service. Going off the deep end (I would put products that claim to calculate trivia like your "cost per single email" in this category) is not required.
The element of IT that should never strive for the shared service model is that which most demonstrates IT's value: the strategic and tactical component. While the majority of IT's time may be spent maintaining the boxes and wires, it's most visible activities should be around implementing the company's strategy through focused IT projects.