Don't Spend It All at Once

Article Featured Image

Last year, Touchstar donated a call center solution to the president of Guatemala for use by the country’s agricultural and exporters associations. The solution provides AgExport’s Assistance Center for International Commerce with a tool to quickly and accurately give callers information about products, tariffs, customs requirements and procedures, international trade agreements, general market conditions, and more.

Touchstar, a call center and unified communications solutions provider based in Aurora, Colo., has pledged to make similar donations where it can to nonprofit groups, governments, and enterprises that are creating economic opportunities around the world. But, because the Guatemalan system is estimated to be worth more than $60,000, the company is not prepared to give hundreds of them away.

Thus, most organizations and enterprises can expect to pay for any system they want to deploy, and that will require a great deal of budgeting and planning far in advance.  

In the fast-changing world of technology, though, budgeting for a speech system is probably the most difficult step in the entire process. Financial projections for any longer than three or four years at most are hard to make, and trying to rigorously enforce a set targeted dollar amount is futile because there are so many variables that can affect the price of even the most basic applications and technologies, both in the short and long terms.

In the case of an interactive voice response (IVR) system, for example, initial system costs will depend on the number of locations, ports, and agents; anticipated call volumes; applications that will be supported; hardware, software, and telephony platforms already in place; the number of prompts and menu trees that have to be written; the amount of usability testing that needs to be done and the number of rewrites needed as a result of testing; and so much more.

"Budgeting for speech is so difficult because most speech self-service solutions include so many things," acknowledges Steve Lawrence, a customer management technology research associate at Aberdeen Group. "Sure, you have a basic IVR, but then you add in call routing, analytics, call recording, etc., and the numbers keep going up."

A Shot in the Dark
Additionally, there are many different types of solutions and pricing models to consider. Costs vary wildly between hosted versus on-premises solutions and between packaged application suites versus customized solutions.

Establishing the types of applications to be used from within those categories has to be the first step in any budgeting process. Without it, "you’re really shooting in the dark," says Aaron Fisher, director of speech services at West Interactive.

With any application there will always be up-front fees, something that most vendors have become pretty good at predicting; however, there are much more far-reaching hidden costs to consider. "One thing most people don’t consider is the cost to all the other parts of the customer-interaction aspects of their company," says Ian Jacobs, a contact center strategic analyst at Frost & Sullivan. "Things will either work or they won’t, and in either case they will affect other things. If it works, you can take things out that have been clogging up the contact center. If the customer is dissatisfied and zeroes out of the application, that leads to higher customer dissatisfaction, which leads to higher agent turnover, which leads to more hiring and training, and so on."

Industry analysts, consultants, and vendors agree that it’s important to be flexible in any budgeting process because of the many variables that exist. In pricing a speech system, so much has to be built around the specific needs of each individual client that it’s very difficult for systems vendors and services providers to quote any kind of a standard rate. No vendor can ever truly say what a system will cost without first looking at each prospective client’s operations from the ground up.

"It’s almost impossible to compare vendors because their cost structures are all so radically different. It’s such an apples-to-hand grenades comparison," Jacobs says. "It’s not like buying a car, where there are set costs that do not change much from car to car."

Touchstar, for example, offers both hosted and on-premises speech solutions, and though it does its best to determine costs with clients ahead of time, it’s not something that comes easy. "We have calculators to determine line costs and usage. We have a lot of experience sizing companies—even without a history—so we can estimate costs based on current numbers and expected numbers," says Brian Smits, the company’s global director of marketing. But even then, "we have so many solutions, and the cost depends on whether you’re purchasing just one solution or how many applications you’re pulling together, and the hardware in place," he adds.

The same can be said for SoundBite Communications, a Bedford, Mass.-based provider of hosted speech-enabled outbound campaign solutions. "We do some things to make it clear to customers what they will need to spend. We try to make things as predictable as possible," says John Keyes, the company’s director of product management. "We offer a fixed fee at the beginning of the process of writing scripts and doing recordings, but it varies based on complexity."

A simple script could cost as little as $1,500; more complicated scripts could cost thousands of dollars, he notes.

West Interactive, one of the largest hosting services providers in the industry, says it can give potential clients a fairly accurate picture of what a speech system will cost ahead of time, but even that has a degree of variability. "With us, you know pretty much what it will cost," Fisher says. But "we approach every client differently because the scope of what they require is different, rather than throwing a blanket cost out there for everyone."

But few, if any, chief financial officers will sign off on a project where the final cost is marked as a big TBD. For that reason, consultants recommend first looking at the stated goals for any speech deployment and spelling those out right away. "Assuming the projects get funded, you’ll need to have stated goals in mind," Jacobs says. "You can’t budget anything until you know what you want to do. And right after you’ve gotten the type of end result you want to see, you need to put it into language the CEO and CFO can understand."

Jim Larson, an independent consultant and VoiceXML trainer, agrees. "A big part of the budgeting process needs to detail exactly what you want, what is needed, and where the money will come from. You don’t need to have the money in hand, but it’s important to have an idea of where it’s coming from and what it’s going toward," he states.

Any work at this stage should include a business discovery that takes into account the reasons that end users call, call volumes, call lengths, hold times, completion rates, and customer satisfaction rates. It should also include a technical discovery that considers not only what hardware—such as computer systems, servers, and networks—is in place already but what changes or additions will be needed to roll out the desired solutions.

In looking at the stated goals, the easiest starting point is a basic comparison of costs associated with having a live agent versus an automated solution. "You know approximately what it costs to have a live agent do a particular function, and you can have a speech application come in cheaper than that," Jacobs says. "Automation rates are a great stat to use. They get CFOs excited because they can see the value right away."

And those values can be applied regardless of what type of solution is planned. "You can look at self-service rates, reducing the number of repeat callers, reducing the number of redirects within the call center, and reducing the number of live agent minutes," Fisher says. "That’s the same whether it’s on-premises or outsourced.

"Show a solid ROI—what you are going to save—and then build stakeholder support around that," he continues. "If you can show real ROI, you can pretty much get what you want. There are very minimal objections the CFO can bring up."

Conversely, failure to provide that analysis can prevent a speech project from getting out of the starting gate. "If people do not understand the goal of the project, the technologies involved, or the ROI numbers, it will be difficult to get projects through the CFO," Fisher says.

"You’ve got to have the ROI numbers beforehand or the CFO will boot you right out of his office," adds Bruce Pollock, vice president of professional services at West Interactive.

When pitching the ROI, most analysts suggest against including total cost of ownership projections, though, because costs are so fluid in this industry. "Any initial stab at [the time of] deployment will not be right," Jacobs says.

And before long, systems will have to be upgraded or replaced, and that will have to be budgeted for as well. "There is a life cycle for all speech applications. You probably have about three years out of any application before you start pulling your hair out and saying that the application is not doing what you wanted it to any more," Jacobs adds. "Even if you keep your legacy system, you still have to throw money at it to keep it current and up and running right."

Larson is a little more optimistic. "Though the industry average is about five years, any application’s time line really depends on its nature, intended users, and the platform on which it’s built," he says.

Speech systems, like any other technology, are fast moving, and a string of new releases with greater capabilities are never more than a few months away. To bring them online often requires system compatibility tests. "One thing people seem to miss is that every two or three years there will be a new software release that will require work to support," Fisher explains. "They may nail their year one and year two projections but not look out to year three."
As an example, Fisher cites the growing interest right now in natural language applications, particularly among early adopters of speech systems.

That’s why there is wide consensus across the industry that budgeting is better done in stages rather than all at once. Larson, for example, maintains that any budget needs to include separate allocations at five stages of the deployment cycle:
1. Investigation, during which the enterprise decides exactly what it wants the speech solution to do and builds its requirements around that;
2. Design, during which the enterprise analyzes whether the hardware it has will be compatible with the software it is going to buy, if the system can support the number of calls anticipated, etc.;
3. Development, during which the actual prompts, scripts, menu trees, and codes are written;
4. Usability Testing, involving both automated means and actual end users; and
5. Sustaining, during which the system is tweaked, upgraded, and fine-tuned over time.

While it is possible to estimate costs for each of these stages, Larson and other industry insiders adamantly insist that  anyone deploying a speech solution invariably will have to refine costs as the project moves forward. "You’ll likely change your mind along the way, cut some things, and add some things, so at the completion of one phase, you will need to juggle funds for each of the next phases," Larson states.

Jacobs agrees. "You really need to budget at each phase," he says, "and you can’t really budget for one [phase] until you finish the previous one."

A Political Move
Along that same vein, "there’s definitely some politicking that needs to go with budgeting for any app," Jacobs continues. "Create one part of the project, see how effective that is, and then use that as ammunition to get funding for the next part."

Larson also says that budgeting in stages will help project managers avoid the all-too-common mistake of overestimating one stage and underestimating the next. When they do that, they often wind up having to cut corners to stay within budget, he says. And, unfortunately, that usually winds up affecting the testing and tuning of an application, something that is widely agreed to be one of the most important parts of any speech deployment. "A lot of people make the mistake of thinking that coding is the most expensive and exhaustive part," Larson says. "The coding is actually only a small part; design and testing is where the most work goes."

Testing, he says, should really take up about one-third of the total initial budget, with another third devoted to development and the other third earmarked for the investigation and design phases. "One-third [of the budget] has to go to testing. Don’t ever short change testing," he stresses. "Everyone always underestimates it, and says it’s the last thing to worry about. Don’t do that!"

Every application will go through testing at one time or another, whether companies plan for it or not, says Lizanne Kaiser, senior principal consultant for voice services at Genesys Telecommunications Laboratories. "You can do it early on or scrap it and go live and then find out that customers are not happy with it," she states.

Kaiser recalls one client who followed that pattern, subjecting its application to a round of usability testing only after it went live and customer satisfaction rates began to plummet. "That’s when it redid the app, and it cost a lot more," she says.

Bruce Sherman, a product manager at Syntellect, says the soft costs in cutting usability testing prior to fielding the application can be even stiffer than the dollar amounts. He recalls a company that scrapped testing to save money.  "The deployment cycle took longer than everyone thought, it cost more than expected, and the organization’s momentum for speech got lost as a result," he says.

As any project manager knows, nothing irks C-level executives more than a project that goes over budget. And more often than not, projects go over budget because the manager underestimated the amount of time it would take to deploy an application. With speech applications taking anywhere from as little as two or three months to as much as 18 months to complete—depending on how complex things get—one can definitely say that time and costs are intertwined.

"Before you begin any speech application project, you need to have some schedules in mind and hold solutions providers to those schedules," Larson states.

Part of the budgeting process needs to outline a complete business schedule—a start-on date, an end-on date, a go-live date, and task completion dates along the way, he continues.

"Any self-respecting speech company will be able to estimate costs, work with you to develop sustaining cost numbers, and work with you on a payment schedule," Larson adds. "You usually pay a fraction to [the vendor] up front, and then there are a series of events when [the remaining] payments will be made."

Touchstar’s Smits goes so far as to recommend that prospective clients specifically demand those kinds of schedules in the requests for proposals that they submit to vendors, and then hold the vendor to those schedules. "For us, a project is a failure if it goes grossly over budget," he says.

Financial Considerations with Speech Automation
  • Application acquisition costs
  • Application maintenance costs
  • Professional services
  • Increase in automation rates
  • Time value of money
  • Iterative changes to business processes
  • Training costs
  • Training costs due to change in call types passed to agents
  • Hardware acquisition costs
  • Hardware maintenance costs
  • Desired result for application deployment
  • Customer service levels
  • Customer satisfaction levels
Source: Frost & Sullivan

Writing the Right RFP
When soliciting proposals from speech technology vendors, including—and leaving out—the right information in a request for proposal (RFP) can make all the difference in establishing a price later on. Brian Smits, global director of marketing at call center solutions provider Touchstar, has seen and written many RFPs, and offers the following tips:
  • Make sure the vendor knows your business requirements. A lot of RFPs talk only about features and prices, but do not address your unique needs.
  • Make sure the vendor can clearly demonstrate how its products will make your business better. Don’t tell vendors what you want. Let them tell you what they think you will need, just to see if they understand your business.
  • Vendors need to know your business and its challenges before they can recommend something, so give enough information for the vendors’ salespeople to make that assessment.
  • Clearly state what the vendor will be expected to do and what you will do so that everyone knows ahead of time and can budget time and resources accordingly.
  • Require performance guarantees that limit service costs while outlining performance minimums.
Once the vendor has been selected, Smits says it’s just as important to properly conduct negotiations. He suggests the following:
  • Ask yourself if the vendor understands your business.
  • Have the sales rep build your initial ROI for you.
  • Obtain a performance commitment.
  • Always ask for a better price or more services. Especially with hosted services, most providers are able to secure volume discounts for their clients from the data and voice carriers they use.
  • Tell the sales rep "no" at least once.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues