Since starting pay per call campaigns last year, at the end of 2022, I've been intrigued by the idea of determining call quality. I'm just curious about data all the time.
I wanted to know how I could use the context of a conversation to help build better experiences for advertisers, which would then hopefully result in better relationships and payouts.
That's why I'm excited to share with you how I use AI to transcribe audio in real time and then get a sentiment score from it.
Why Get the Sentiment of Your Audio Calls?
Stock market traders use sentiment analysis all the time to get a sense of where the market is moving. They analyze articles, community forums, and other sources to understand the overall attitude of the market before making a decision.
We can do the same thing with affiliate marketing calls. Since most call-tracking platforms record calls, you can easily transcribe them with AI. This can be done manually, but there are also automation tools available that make the process even easier.
Once you have transcribed your calls, you can use sentiment analysis to identify the overall tone and attitude of the caller. This information can be used to improve your affiliate marketing campaigns in a number of ways:
- Choose better products and services to promote: By understanding what customers are saying about the products and services you are promoting, you can identify the ones that are most likely to generate sales.
- Create more effective marketing materials. Sentiment analysis can help you understand what kind of language and messaging resonate with your target audience. This information can create more effective marketing materials, such as landing pages, ad creatives, email campaigns, and social media posts.
- Improve your customer service: Even though you are not the one selling the product or service, you can still use sentiment analysis to improve your customer service. By understanding what customers say about their experience, you can identify areas where you can provide better support.
Overall, getting the sentiment of your affiliate marketing calls is a valuable way to improve your results. It can help you choose better products and services to promote, create more effective marketing materials, and improve your overall campaign.
Using Webhooks to Download Call Audio
If you happen to be using Ringba or Retreaver, and I've personally used both of these platforms, you have the capability to obtain the audio URL via a webhook integration. Here's how it's done with Retreaver:
Within the Retreaver platform, navigate to the campaign level and access the WebHooks settings.
What Make.com does in this context is capture the URL pointing to the audio file, and subsequently, it downloads the audio content. The downloaded audio file is then automatically stored in your Google Drive.
Here's an overview of my Make.com workflow:
Transcribe Audio with AI
Within the Make.com ecosystem, I utilize EdenAI's versatile suite of AI services to facilitate transcription. This unique feature allows me to select from various transcription AI services, and my choice for this task is Assembly AI, renowned for its accuracy and proficiency in transcribing calls.
The process begins by taking your recorded calls and converting them into a coherent and easily readable textual format. This transcription serves as a crucial foundation for further analysis and review.
After the transcription, I use another AI service to provide an in-depth assessment of the overall sentiment conveyed in the conversation. It's important to note that determining the overall sentiment alone does not always reveal the potential for the caller to continue or take specific actions. Therefore, I take an extra step by assigning a numerical score to the sentiment analysis results.
This scoring process adds a layer of granularity, allowing me to make more precise judgments based on the sentiment data. The score provides a quantitative measure of the caller's disposition, aiding in identifying not only their overall sentiment but also the likelihood of their future engagement.
Once these processes are completed, I instruct Make.com to compile the information into a Google Sheet for easy access and review. I can quickly peruse the transcriptions in this spreadsheet, review the overall sentiment ratings, and analyze the accompanying scores.
This comprehensive approach ensures that I have a clear understanding of the communication's content, sentiment, and potential impact, facilitating informed decision-making and action.
That's the entire process, and now you have a powerful tool to thoroughly review your calls, ensuring the legitimacy of the leads you send to your advertisers. This becomes especially crucial when addressing any potential quality-related disputes that may arise.
By retaining hard evidence, you can definitively demonstrate the variance in call quality, providing clear and indisputable support for your claims.
The advantage of implementing this system extends beyond call quality assurance. It can be harnessed to expand the depth and breadth of your audience pixels, allowing you to refine your targeting strategies and reach the precise audience segments that are most pertinent to your affiliate campaigns. This enhanced precision and granularity in audience targeting significantly elevates the effectiveness and performance of your marketing efforts.
I might do another post on that later on. If you know me, I like manually firing pixels to create a comprehensive audience pixel.