The Real Costs of Deploying ASR Solutions
In my 15+ years as an Engineering Pre- and Post-Sales Support Executive, I rarely encountered a company that didn’t have some existing analytics platforms, which often operated in separate silos. When evaluating a new technology solution, it’s crucial to perform due diligence on integration with other products and process flows.
Evaluating ASR solutions is no different. While you usually can customize your own integrations, it may come with an additional cost. This may work for your business, especially if you are new to ASR or to call recording and analysis in general. But most established speech analytics vendors have processes built around the use of particular components and tools, which can’t be removed easily from their platforms. Before you commit to a specific solution, it’s important to understand if you’ll be charged additional fees to continue using your existing stack.
Having flexibility of deployment — whether you need ASR solutions on-premise, in the cloud, or for both — is another factor to consider during evaluation. ASR investments that deliver the greatest return can be deployed flexibly and scale across your organization to meet your business needs of today and tomorrow. However, as Jay Pascarella pointed out in his blog, this flexibility should also be aligned with your data security needs. If you don’t protect your data adequately, your business and reputation could suffer a serious blow, and your customers could face even worse consequences.
While you can expect to pay an initial fee for software implementation, training, and support, the scope and pricing for these services differs from vendor to vendor. Unfortunately, not all vendors disclose the full implementation costs upfront, and sometimes you discover these mandatory extra charges and hidden professional services fees when it’s too late and you’re already planning out the project. Understanding the scope of contracted services during the evaluation period will help mitigate any future surprises.
No on-premise ASR solution can operate without dedicated hardware. Some vendors specialize in customizing hardware configurations for your required infrastructure, while others use an “ASR in a box” approach in which customers choose their hardware from limited number of options. Both methods are valid depending on your needs. Some ASR vendors may also require you to use their hardware instead of purchasing your own, or require deployment on their proprietary hardware. While you might not need to buy, install, or maintain hardware, you may incur processing charges on top of the transcription cost. Added hardware costs, deployment flexibility, and possible charges for any deviation from the standard offerings should all be considered as part of the solution evaluation process.
Data accessibility also is important. Some vendors will prevent you from accessing your recordings and transcriptions unless you pay them an additional fee. Mike Smocer addressed this issue in his blog, but to keep things short, it’s crucial to understand who owns your calls’ data.
The final parameter to consider is metadata. While metadata isn’t directly a part of your audio files, this information significantly improves analytic capability. Is your metadata handled in the same API as the transcriptions, enabling an easier and more flexible integration of your ASR solution? Most vendors have a separate integration for metadata, warranting additional cost.
When it comes to choosing an ASR solution, performing due diligence on its deployment will likely uncover flexibility and scalability benefits, data security vulnerabilities, and hidden costs.