With web computing, smartphones and voice becoming the new human-machine interface, the phone is now rediscovered. The objective of this conference is to present the call tracker tool created by Allo Media to collect and exploit data from telephone conversations.
Why do brands avoid call volumes?
The phone remains a powerful channel. As consumers, we all call the customer service of the brands we interact with. This is a natural mode of expression, unlike e-mail, which calls for more effort.
It's not the brand or the marketing department that hears customers when they call, but a customer service advisor who takes the calls. The relationship between call center employees and marketing employees is often almost non-existent. The client, meanwhile, expresses himself as if he spoke to the brand and his interlocutor deals with the call to provide an answer or a service, not to capture the entirety of his speech.
For 20 years, brands have considered call centers to be "cost centers" because the value they capture is problematic. That's why signs are trying to limit call volumes, while customer interaction is the basis of the business. In addition, 60% of customers believe that their request is not generally well treated.
What is the technology developed by Allo Media?
Allo Media has developed a call tracking technology by working on a language model and acoustic model adapted for each of its customers. The objective of this call tracker is to be able to tackle the Grail of automatic recognition, namely the telephone conversation. Indeed, the telephone conversation combines several major technical difficulties, such as a sign of very poor quality and unstructured language between two human beings.
Using an acoustic model and a custom dictionary, the tool analyzes an entire telephone conversation to convert it completely into a complete transcription. Added to this is machine learning to understand the meaning of the conversation. This is an automated semantic analysis to determine for example what bid, for which keyword Adwords, was able to generate lead on phone. This technology is called "voice cookie".
What are the features of this tool?
A live demonstration was performed during the presentation, using a volunteer from the audience. On a travel site presenting six destinations and different types of stay (week-end two, family week ...), the volunteer chooses his formula and passes a phone call picked up by the tool. Several technologies are at work at the same time, namely:
- An acoustic model, made specifically for the phone.
- A language model adapted to build language elements specific to the business sector of the company, namely travel.
- A semantic analysis model to understand the content of the call.
Thanks to this phone call, and the use of the call tracker, it was possible to collect the digital route (the pages viewed on the site and the page that generated the call), as well as the link made with the conversation. even though nothing links the computer used during the presentation and the volunteer's phone.
Generally, companies use the technique of call tracking. This method looks at the amount of calls generated by pages on which a specific number is displayed. Allo Media goes further with a clean and unique dynamic phone number for each browsing session. This is what connects the call to the online visit. Other features must also be mentioned:
- The data is captured, structured and sent to a CRM. Allo Media's call tracker work is to build the data and send it in this type of tool in order to target customer profiles and propose personalized offers (for example, a Valentine's Day travel offer for a couple).
- The voice cookie is also used to evaluate the advisors. These assessments are a time-consuming and expensive process for brands, as an advisor processes around 100 calls a day. The call tracker is used to check certain predefined waypoints during a conversation to assess the quality of the call.
- Connectors make it possible to exploit the data. This makes it possible to create a new event, calls, in Google Analytics, and to correlate the pages and keywords that generated the different types of phone calls.
- Today, ads that go back to the engines are generally linked to the click or navigation course. During the live demonstration of the presentation, nobody clicked on anything. The content was built on the basis of the telephone conversation. It is therefore possible to either hyper-personalize or to exclude from campaigns those customers who have already bought a trip so as not to annoy them.
In conclusion, the voice cookie allows to analyze the performance of the marketing campaign and to optimize it with a reduction of the acquisition costs and an increase of the conversions. All this by retargeting afterwards and by generating commitment.
Speaker: Frédéric DANIEL
Discover the Useradgents workshop on the impact and technological contribution of voice assistants
Discover the impact of voice on brands and content at the plenary of the One to One Biarritz 2018