Two years ago, Deloitte Consulting completed a research study predicting that by 2006, over 66% of companies would have deployed Voice over IP, a technology that allows phone calls to be placed over the Internet. They listed several of the common reasons for this migration ï¿½ shared network infrastructure, reduced complexity and costs, technology trends, and added functionality benefits. Myriad organizations have also joined in. Boutique technology consulting firms have published white papers sharing the success stories of companies converting to VoIP. Cisco Systems (News  Alert) provides detailed documentation explaining the protocols behind the technology, and even the business case for the technologies.
Yet, todayï¿½s VoIP technology is not as widespread as many have predicted. The executive offices and IT departments of many companies are eager. But they remain on the sidelines, likely because of several highprofile implementation failures. For all of the advertised benefits of VoIP, it has significant downsides. User expectations, technical details, potential business interruptions, and financial payoffs all come into play. Itï¿½s no wonder why many decision makers remain hesitant.
How to Analyze Risk?
Existing literature does an excellent job outlining and explaining the potential benefits of VoIP. Customers understand the advantages well, but itï¿½s the risks that they worry about.
Stable telephone communications is essential to many businesses. Companies could lose thousands, if not millions, in sales and revenues if the phone system went down for several hours. Executives taking on a VoIP project would also be concerned about potential telephone outages causing productivity loss, missed deadlines, and resentful employees and customers.
Even those who manage IP networks every day are hesitant to embrace VoIP systems. One network operations engineer who worked for a California energy company stated his concern this way: ï¿½When the computer network goes down, whatï¿½s the first thing we do? Pick up the phone and begin a troubleshooting conference call. With Voice over IP, the telephone wonï¿½t work!ï¿½
Upon hearing all of these worries, it became clear to me that those considering VoIP projects donï¿½t need stronger arguments for cost savings or a better ROI calculator. Instead, we need a quantitative analysis on understanding and managing the risks associated with VoIP.
One of the best quantitative tools for making sound decisions under uncertainty and risk is Decision Analysis, a mathematicsbased decision methodology taught in business schools worldwide. In this paper, we will explore the risks and possible outcomes of a VoIP project using a Decision Tree model, analyze the likelihood of each outcome, and draw conclusions on when and how to implement a successful VoIP project.
A Decision Tree Model: Mathematically Analyzing a VoIP Project
The decision tree depicted in Figure 1, despite its daunting appearance, is actually quite simple. It has one decision event, designated by the square, and three uncertain events, designated by the circles. The story reads like this:
#1 square: A company faces a decision on whether or not to implement VoIP
#2 circle: If the company chooses to pursue the project, they face uncertainty, because the implementation may succeed or it may fail.
#3 circle: If they are able to implement VoIP, they face more uncertainty. The new telephone system may be stable or it might be unstable and hurt business.
#4 circle: If the new telephone system is stable, the company may experience overall cost savings (or net benefits via new functionality), no cost savings, or increased cost (or net ï¿½lossï¿½ due to problems such as poor sound quality).
To build this model, we needed to make reasonable ï¿½best guessï¿½ assumptions:
ï¿½ The cost of a VoIP project is normalized to 100 monetary units. While this cost may be $100,000 or $5.6 million, we use a project baseline of $100 for simplicity.
ï¿½ The cost of ï¿½doing nothingï¿½ and staying with the traditional phone system is set at a baseline of $0.
ï¿½ A failed VoIP implementation will incur an additional 50% cost ï¿½ this cost includes reversing the telephone infrastructure changes, ï¿½managerial embarrassment,ï¿½ and costs of a damaged reputation.
ï¿½ An unstable VoIP system will cost 10 times the original project ($1,000 in additional costs). This includes emergency troubleshooting, loss of productivity, loss of sales, and of damaged reputation.
ï¿½ A stable VoIP system can leave the company ï¿½worse off,ï¿½ ï¿½about the same,ï¿½ or ï¿½better offï¿½
If the company is ï¿½worse offï¿½, it will incur additional costs equivalent to the original project costs ($100 additional cost, net present value).
If the company is ï¿½about the sameï¿½, the net cost and benefit will be about the same as traditional telephones ($0 more, net present value).
If the company is better off, it will save ten times the original cost of the project ($1,000 savings, net present value).
ï¿½ This company has a 70% chance of successfully implementing VoIP.
ï¿½ If the implementation is successful, the company has 70% chance of having VoIP run stably without affecting business.
ï¿½ If VoIP is implemented and stable, the company has 50% chance of being ï¿½better off,ï¿½ 25% chance of being ï¿½about the same,ï¿½ and 25% chance of being ï¿½worse off.ï¿½
The Results. What Do They Mean Anyway?
The decision tree recommends this company not to implement VoIP and to stick with the traditional telephone system. On ï¿½average,ï¿½ the VoIP project will result in a $92.25 loss and, therefore, the company should stick with the traditional PBX with its project cost baseline of $0.
The result, based upon our assumptions, makes sense. Conventional wisdom says if a company has a solid wellbuilt telephone infrastructure and it depends on telephones heavily for its business, making a big investment and taking on significant risk for a VoIP project is likely a bad idea. Mathematically, the results are consistent. As the CEO of this company, if you are worried about heavy losses and expect a 24.5% chance of success (properly implemented and stable and better off: 0.7 x 0.7 x 0.5 = 0.245), you certainly wouldnï¿½t want to commission a VoIP project.
Keep in mind, however, that this recommendation and ï¿½average lossï¿½ is dependent on the numbers fed into the model. While the result is pretty good for a stable company with similar costs and a similar perspective on the risk, those with differing assumptions may end up with a different result. If our model was used for an actual company, we would take painstaking measures to ensure accurate input. We would calculate expected cost for the VoIP project, create a financial model to analyze the value of a ï¿½better offï¿½ telephony scenario and the ï¿½worse offï¿½ telephony scenario, and determine the realistic cost an unstable telephone system. All of these financial estimates would used as inputs to create a more precise model.
The Problem with Probabilities
Take another look at the decision tree. Notice that probabilities are an important part of the model. Now the traditional notion of probabilities that we learned in school ï¿½ the probability of flipping heads is 50%, the probability of getting a black jack in a fair, shuffled deck is 4.8% ï¿½ doesnï¿½t apply here. We can flip a coin a 1,000 times in a controlled condition to calculate the true probability; we canï¿½t implement a VoIP project 1,000 times with the same company in the same way to calculate the true probability of success, nor would we want to.
Accurate probability inputs create a unique challenge. Financial estimates can be calculated fairly accurately, but probability estimates require a bit of guesswork. Consider the probability of a successful project. Even if a consultancy states that, historically, 90% of its VoIP projects were successfully implemented, an astute decision maker would make his own estimates. Many other factors, like current employee morale, the complexity of the current telephone infrastructure, the capability of the current IT support team, and relationships with the consultancy, could affect the likelihood of a successful project. In the end, these probability inputs will be subjective estimates.
We can examine how different subjective probabilities affect our recommendation by using sensitivity analysis ï¿½ that is varying the model inputs and seeing how our recommendation changes. Consider a company with multiple decision makers, each with his own probability estimates. The CEO, confident in his employees and the VoIP consultants, estimates the VoIP project will be successfully implemented with a 90% probability. The CIO, concerned about unforeseen technical obstacles in implementing VoIP, gives VoIP implementation a 70% chance of success. The VP of sales, who believes that his customer support staff will try to cancel the project at the slightest sign of any glitches, places a 50% chance of success on the project. How would these different probabilities change the results of the model? The sensitivity graph shown in Figure 2 contains the answer.
The graph explains when our base recommendation of ï¿½do not pursue a VoIP projectï¿½ changes to ï¿½pursue a VoIP project.ï¿½ The key is in the inflection points. When we believe that the probability of a stable VoIP (above in pink squares) system is above 68%, the VoIP project starts to become worthwhile. Likewise, when we believe that the probability of a successful VoIP implementation saving costs (above in yellow triangles) is above is above 79%, the VoIP project also begins to become worthwhile. The Yaxis shows the ï¿½averageï¿½ savings of the project. For example, if the believe that we have a 90% probability of the VoIP system being stable, the ï¿½averageï¿½ value of the VoIP project is $110; on average, the VoIP project will recoup its original cost plus 110%.
Interestingly, the sensitivity analysis graph above shows that the three executivesï¿½ probabilities all end with the same result: adopting the VoIP project is not recommended. Even if the company is 100% sure that it can implement VoIP successfully, the base assumptionï¿½s 30% risk of telephone instability and 50% chance of reducing cost makes VoIP a poor decision. These executives should, therefore, be more concerned about the stability and cost savings likelihood of a VoIP project than worry about the success of the actual VoIP implementation itself.
We can glean a few interesting points from the sensitivity analysis graph above.
ï¿½ The probability of stability is the steepest once we reach the inflection point. This means that increasing the odds of a stable VoIP system adds the greatest value. Anything the company can do without significant expense can create value for the VoIP project. A few examples may include:
Focus more on technical planning to ensure reliability.
Create an efficient outage response procedure, so VoIP issues donï¿½t affect the business.
Provide sufficient training to those maintaining the VoIP systems.
Establish strict security controls so that unqualified individuals do not accidentally bring down the system.
ï¿½ The probability of cost savings inflection point occurs the earliest, at 68%, but is not as steep as the probability of stability curve. This means that anyone implementing a VoIP project should be pretty confident (~70% sure) that it will reduce costs. Itï¿½s not as big of a risk as an unreliable telephone system, but a good VoIP project requires a sound business case.
ï¿½ The probability of a successful implementation never reaches its inflection point. This means that other concerns, such as stability and cost savings, should trump the worries of a failed implementation. A failed project is always embarrassing, but a project that disrupts overall business and hurts revenues would be much more damaging. Ensuring telephone reliability and building an effective, cost saving business case should take priority over creating a great project management team.
Twovariable Sensitivity Analysis: When Does VoIP Make Sense?
Simply stated, VoIP makes sense if a company can make it work. The telephone infrastructure must remain stable ï¿½ or at least stable enough not to impact daytoday business operations. Furthermore, the company needs to realize the proposed benefits. A fully stable new VoIP system would be pointless if it increased costs and didnï¿½t add any useful benefits. The combination of these two factors ï¿½ reliability and cost savings ï¿½ is key to successful VoIP projects.
Deciding to implement VoIP would be easy if we were 100% sure that the new VoIP system would be reliable and 100% sure that the company can realize the advertised benefits. No executive would be so naï¿½ve as to believe this to be the case. Figure 3 can help decision makers.
Letï¿½s use the previous example of a CEO and CIO as two decision makers. The CEO believes that the company has an 80% likelihood of realizing the promised cost savings and 90% chance of being able to establish a stable VoIP system given the right planning. Based upon these estimates, the table (created from the Decision Analysis tree) recommends pursuing the VoIP project. The CIO, however, estimates the probability of costs saving at 50% and the probability of a stable telephony at 70%. Given these estimates, the model suggests staying with the traditional PBX phone system.
Decision makers using this table as a tool should be careful at the border between PBX and VoIP. Research shows that most people are risk adverse, especially with the possibility of significant downsides. Therefore, in cultures that hesitate in risk, decision makers may be better off avoiding the risk of a VoIP project, even if the table makes a VoIP recommendation that lies on the border.
Decision Tree Variant: New Facility
Now suppose our company faces a different scenario. We are building a new facility, and the telephone network has not yet been built. In this new situation, does it make sense to install the new VoIP technology instead of the traditional PBX technology? A variant of our base decision tree can help us make this decision.
Our new decision tree is shown in Figure 4. This new model has three important changes.
ï¿½ The PBX system has not been installed. The installation of a new PBX telephone system faces the same uncertainties of VoIP system. For example, the implementation of the new PBX system could fail; the new PBX system might be unstable; and the PBX system may be more expensive than the VoIP system. These three uncertainties are represented with the new uncertainty event nodes (shown as circles).
ï¿½ A new VoIP system is less expensive to install than a new PBX system. The VoIP system will run on computer network infrastructure, thereby eliminating a second set of wires and cables.
ï¿½ An unstable telephone system, while still costly, wonï¿½t be as damaging as the base scenario. Employees arenï¿½t currently stationed in the new facility. Unreliable systems will mostly amount to more troubleshooting and delays, rather than interrupt crucial business processes and hurt sales.
This new decision tree recommends pursuing VoIP. This new VoIP system will have an ï¿½averageï¿½ price of 61.6 monetary units ï¿½ meaning that we expect the potential benefits and cost savings of this new VoIP system to pay off 38% of the upfront cost. The VoIP system is much less expensive than a new PBX system, which the model expects to have a TCO of $376.9. Notice that both telephone systems will have a net cost, so we choose the less costly of the two. The VoIP system faces more uncertainty around the implementation and the reliability, but it is less expensive immediately and in the long term.
Letï¿½s revisit one of our earlier questions: When does it make sense to implement VoIP? In our base scenario, we had to be mostly sure (>80%) that the company could make it work. But, in our new facility scenario, we donï¿½t have an existing telephone network and the stakes are lower. A decision maker would only need to be fairly sure (>40%) certain that the company can make it work. A twoway sensitivity analysis is shown in Figure 5.
With our base case of 70% VoIP stability and 50% cost savings, the model makes a solid recommendation to adopt VoIP. This sensitivity analysis table recommends, in general, those who are fairly comfortable with the new VoIP technology should implement it in their new facility.
Top 3 Variables: Whatï¿½s Important in a VoIP Project?
The greatest value in this decision tree lies not in the final recommendation, but the analysis. Astute executives and decision makers rely on both numbers and intuition to make sound business judgments. A helpful intuitive benchmark for understanding this VoIP decision comes from the tornado graph below.
The tornado graph (Figure 6) provides a visual explanation for the most important variables in the decision making process. The most important variables bubble up to the top of the diagram. The less important variables fall to the bottom.
To read the tornado diagram consider this question: What would happen to the total cost of VoIP if one of our decision variables changed while we were implementing the project? For some variables, the swing can be significant. For others, the difference would be barely noticed. Consider a successful VoIP project that saves the company money. If it saves just a little bit of money ï¿½ just enough to recoup the cost of the project ï¿½ the overall value of this project is almost negative $400. However, if we see a $3,000 net present value savings, the project can pay for itself almost six times over ($600 normalized).
The top three variables in a successful project, as shown in the tornado diagram, are
1. The value of savings, or benefits, in a successful, stable VoIP implementation that saves the company money.
2. The likelihood of a stable telephone system.
3. The cost (including lost revenue and reputation damage) that will be incurred if a telephone system fails.
The most important factor in approaching a VoIP project is the ongoing savings and net benefits of a new VoIP system. In terms of hard dollars, this means reduced maintenance costs (e.g., payroll, tech support, technology replacement, and long distance fees). Soft dollars should also be included in this estimate. For example, the opportunity cost for missing a technology trend or upcoming technologies and benefits that may be supported on VoIP but not on traditional PBX systems. To increase the chance of the overall success of a VoIP project, decision makers should pay careful attention to what vendors call Total Cost of Ownership (TCO), because ongoing maintenance cost is the main business driver and the most important factor in a VoIP project.
The risk of an unstable telephone system makes up other most important factors in a VoIP project. In our model, we break this risk up into two separate components: the probability of experiencing telephone instability that would impact business and the cost of this instability, should the telephone system actually become unstable. Companies interested in VoIP projects can take many steps to mitigate this risk; many possibilities ï¿½ such as better training for those maintaining the new VoIP system, thorough assessment of the technology, and strict security access ï¿½ were suggested above. However, reducing the impact of an outage, should one occur, is very important. Backup procedures and alternative technologies may significantly reduce the cost of telephony problems. For example, companies can leverage mobile phones or allow employees to work from home. Otherwise, the company may leave a few strategically placed traditional phone lines, in key areas (i.e., CEOs office or the network administratorï¿½s
office). This backup system can help important business continue or allow troubleshooters to effectively do their jobs during an outage. Another way to minimize the damage caused by VoIP outages is to roll out this technology slowly. By implementing one location at a time in a tiered system, the company can allow certain offices to cover for others, should VoIP become unstable.
Conclusion: Manage Your Risk
VoIP projects, because of the risk involved should, indeed, be approached carefully. A successful project requires a very good ROI and carefully managed risk of telephone instability. Converting to VoIP from a stable traditional phone system only makes sense if the decision maker is confident (>80%) that the project will result in significant savings and that telephone instability wonï¿½t hurt business, but those building new facilities only need to be somewhat confident (>40%). Through careful planning, such as detailed financial analysis, technology understanding, and well thought out measures to prevent telephone instability from affecting business, many companies should be confident enough to deploy VoIP ï¿½ at least in their new facilities.
While VoIP may not be the solution for everyone, those who implement it properly ï¿½ and in the right situations ï¿½ can garner significant benefits and cost savings from it. Successful use of VoIP requires making the right decision initially and carefully navigating the VoIP implementation process. IT
Ryan Tang is a Consultant with Deloitte Consulting LLP. For more information, please visit www.deloitte.com.
If you are interested in purchasing reprints of this article (in either print or PDF format), please visit Reprint Management Services online at www.reprintbuyer.com or contact a representative via email at [email protected] or by phone at 8002905460.
