What is Conversational Conversion Optimization?
Conversational Conversion Optimization is the practice of optimizing automated conversations towards one or many desired outcomes. These outcomes could be: a higher rate of engaged conversations, a higher rate of emails captured or other types of data collected, a higher add-to-cart or purchase rate in ecommerce scenarios or other objectives matched to a specific use case for a chatbot, voice assistant or other forms of conversational AI.
Conversion Optimization has been practiced by marketers in one way or another for a long time. For digital marketers, it’s part of the air they breath – every decision can and must be data-driven, and they always try to establish clear controls that allow them to optimize an experience towards outcomes. Conversational AI is a powerful new way to engage consumers, but without better performing experiences, marketers and brands will never realize it’s full potential. That’s why we built conversation conversion optimization – read more to learn how it works and what it can do for you!
How does Conversational Conversion Optimization work?
In traditional conversion optimization, marketers use A/B testing to define and measure outcomes based on variables they include in any online or interactive experience.
With Conversational Conversion Optimization, marketers and conversation designers can optimize the standard conversation flows they build throughout the course of a conversation design. By following the history of interactions consumers have and outcomes, this process can be optimized towards desired goals like the ones mentioned above.
However, given the large amounts of information processed by each conversation, marketers need a sufficiently automated way to optimize themselves against the variety of outcomes they want to achieve. With Conversational Conversion Optimization, they can establish a variety of variables, weigh their value against one another (e.g. a purchase is worth more than a captured email), and then allow this feature to produce the best aggregate outcomes across each and every conversation.
Marketers might want to test pictures, emojis, different phrasing or language, or different multimedia approaches to communicating a point or delivering content. They might want to try out different lead generation or product recommendation processes, or simply test out a variety of natural responses that AI can give someone looking for assistance. Conversational Conversion Optimization allows both these variations and conversations as a whole to be optimized to generate the best result with consumers.
Using a variety of features like weighting, personalization and degrees of confidence, these outcomes are able to be rapidly achieved based with relatively limited data. Most brands and enterprises assume that they need large data sets to achieve highly efficient conversational experiences, but with the right design, optimization tools and domain-specific approaches, material benefits can be realized in a relatively short amount of time.
What do brands and consumers get out of conversations?
Consumers and brands get a variety of things out of talking to one another. The three primary benefits for brands are improved engagement, sales and satisfaction, but consumers also benefit from the experience that is provided to them. Consumers tend to find conversational AI a quicker, easier and more informed way to shop, and understanding the mutual value that is already generated by conversations is essential to understanding how conversations can be optimized.
Whether it’s to increase engagement with a particular question or concept, or to drive greater sales by altering the way or order in which products are presented, these are the kinds of objectives that can be considered by a designer in optimizing a conversation to convert.
In optimizing a conversation, almost any goal can be considered as a desirable “end state” that a conversation can be optimized towards. This “end state” of course doesn’t have to occur at the end of the conversation, but can include things like:
- Selection of a particular choice or initiation of a particular flow
- Achievement of a certain number of turns
- Diagnostic completion
- Email collection
- Product purchase
- Achieving a particular number of turns in a conversation
There is relative variety in the kinds of goals you can establish in conversational conversion optimization. This reflects the flexibility of conversational experiences overall, particularly when powered by AI. When we have conversations with other people (or brand representatives), we open more possibilities than we can often anticipate, and unexpected turns in conversations are for the most part what makes them enjoyable. Where AI is able to provide a dynamic conversational experience and then also optimize it, both brands and consumers win.
Understanding Conversational Conversion Optimization’s AI-Driven Capabilities
Conversational Conversion Optimization leverages different capabilities to achieve these best aggregate results for brands and consumers. These include weighting between goals, personalization based on known information about users and degrees of confidence to temper or exaggerate the way in which conversations are optimized depending on historical data.
Weighting: If you have multiple goals, you can apply weighting between them. This means that if different goals are not of equal value, the conversation can be optimized to the expected outcome. So, while a high value goal (like completing a purchase) doesn’t seem like it can be achieved, less valuable but still relevant goals (like capturing an email) can be focused on to maximize predictable returns.
Personalization: This feature also uses personalization, which takes the raw data available to conversation designers and creates filters that can optimize outcomes on the basis of user age, gender or features that are defined in engaging with the chatbot. This means that optimization can be personalized, while also being categorized to make it easier for a human analyst to make judgements about optimizations based on buckets.
For instance, users could be bucketed between ages 18-24, 25-37 and 38-75. A designer or analyst could then evaluate the performance of a given conversation flow based on the age group and identify qualitative differences in the way the chatbot communicates and use those insights to further obtain higher conversion.
Degrees of Confidence: Conversational Conversion Optimization uses degrees of confidence to adapt to results. Let’s say one type of variation is performing better than expected and another is performing worse than expected: within the bounds of existing weightings and personalization, if there is sufficient data to infer confidence in the positive result, more traffic will automatically be routed to that result. This accelerates the rate at which conversations are optimized, limiting the cost of testing for brands and improving the experience provided to more users.
The Future of Conversational AI
Aside from its essential benefits and having the right conversation design platforms and NLU capabilities around it, the most important thing to consider in using Conversational Conversion Optimization is who is using it, and whether their thinking and approach to conversation design is creative and unique enough to get the most out of it.
For instance, conversation flows can vary between users, but the best overall performance often relies on nuance, rather than speed. This means that it’s not always the quickest or most direct conversation that converts best – an evaluation that is hard to make in traditional conversion optimization, but is second nature with the Conversational Conversion Optimization built into our own platform.
A conversation is often more likely to convert if it takes more turns rather than less. A higher frequency of interactions actually improves conversions – an optimization that is counterintuitive compared to the usual digital marketing approach. Sometimes, a brand may also simply want to generate data and use it for marketing via other channels. All this requires creative and expert thinking, guidance and experience, something the team here at Automat is uniquely accustomed to providing in their work with many of the world’s largest brands.
Interactive experiences are still unusual to see deployed at large scale by many brands, but Conversational AI is rapidly changing that. The kind of nested optimization offered by our Conversational Conversion Optimization is essential for any sophisticated customer acquisition process – particularly when applied to conversations that produce anywhere from 4-6 unique insights on individual consumers. Real-time interactions can be optimized in a totally different way. Using this feature and the reinforcement learning that underlies it, you can truly maximize your marketing performance – all with the same machine intelligence that guides rockets back to Earth!
Arman Kizilkale is a Research Scientist at Automat and led the development of Conversational Conversion Optimization. Contact Automat to learn more about how this new feature can benefit your brand.