The emergence of mobile messaging apps, AI virtual assistants, and voice-based platforms have bred a new generation of apps with “conversational” interfaces. Such apps may be the key to unlocking a new, 3rd generation of the classic marketplace business model.
While there are many conversational commerce applications, or “chatbots”, it still feels like we’re still searching for their killer app. One area to explore is the “higher order” service verticals that have otherwise been notoriously difficult to execute.
Each major technological shift has catalyzed a new generation of online marketplace; let’s explore what’s happened previously, and where the next opportunity may lie.
1st Generation Marketplaces: Directories
The rise of the internet provided an unprecedented platform to aggregate long-tail supply & demand, spawning the first online marketplaces. Marketplaces of this generation looked more like online classifieds, or directories, such as EBay and Craigslist (as horizontal platforms), or Match.com and Yelp (as vertical offerings).
These businesses were largely funded by “listing fees”, and later “pay to communicate” models for enabling buyers & sellers to meet. However, these platforms offered little else, leaving it up to the matched parties to process their own payments, and manage other logistics.
As such, these directories were fairly “static”, creating profile pages for sellers, and bringing new buyers in via search. However, without real-time inventory and availability management, these profiles tended to “age” over time.
2nd Generation Marketplaces: On-Demand
The convergence of smartphones, mobile apps, and payment technologies, gave rise to the second generation of vertical, on-demand marketplaces, such as Uber, Airbnb, and Handy.
The technological advancements enabled these models to be more “dynamic” than their predecessors, by supplementing seller profiles with location, real-time availability, and messaging. As such, these platforms can offer end-to-end experiences that not just match buyers & sellers, but also facilitates their bookings, payments, and reputation management. There is often no charge to either party, and instead a transaction “tax” is applied as the primary means of revenue. Similarly, discovery of these platforms lend themselves well to peer-to-peer referrals, as well as App Stores.
Dominant players here succeeded in the lower order marketplace types, disrupting many incumbent directories.
3rd Generation Marketplaces: Concierge
While the immediacy of on-demand marketplaces lend themselves well to commodity goods & services, higher-order verticals don’t fit this model well, due to the level of consideration and customization of the product / service transacted.
In these “higher order” verticals, the trust barrier is substantially higher, which make conversational approaches, that deliver personalized advice, a perfect fit. However, to execute on this model, the current “chatbots” must evolve to this new “concierge” model.
What makes for good trusted advice
Back in post-bubble apocalypse of 2001, at my first startup, WiseUncle, we built early automated advisors based on a dialog, or conversation, model. From those models we developed, and the learnings we gathered, here are some principles that an effective automated advisor should follow:
- Offer a multi-step, branching conversation, that adapts based on the sum total of answers from the customer, rather than a strict decision-tree. This is an evolution over those seen in poor IVR menu systems, or context free models like Siri and Alexa.
- The conversation should be optimized to elicit the customer’s needs in the way that they think about them, rather than finding an optimal path through a solution space.
- The high-level needs a user articulates should be intelligently mapped to potential solutions, based on experts, community data, and ongoing training
- The system must then map those potential solutions to feasible ones, based on real-time inventory, and configuration rules.
- When the system cannot find a feasible solution, it should provide intelligent fallbacks, or provide “background searches”.
- To help users understand how their needs are met by the recommended solution, the system must be able to “introspect” and provide explicit justification for its decisions.
What’s needed for this to take off?
I think we’re still in the early innings of this paradigm, where these models must evolve from catering to developer audiences to ones accessible by marketers, subject matter experts, or other “knowledge engineers”. So while we’re still in the “DOS-era” of AI, understanding and adhering to the aforementioned principles can help to unlock key marketplace verticals, and find a “killer app” for conversational technologies.