The problem is, customers usually hate the platform and stay away. I’ve built several 2SMs over the last dozen years and advised several others, including a couple right now. I’ve learned that for a two-sided marketplace to get any traction, the platform has to match vendor strengths with customer needs. But again, this isn’t as simple as it sounds.
Most 2SMs will tell you that they do the matching, but if anything, their simplified scoring is usually only beneficial to the vendor, or worse, only beneficial to the platform.
Here’s why that doesn’t work and how to fix it.
The directory model leaves it to the customer to find the right vendor to fit their needs. It’s the equivalent of making them walk into a giant shopping mall (which customers don’t do), choose a store at random (which customers don’t do), and then ask the cashier if the store has what they need (which customers hate to do).
When matching is promised, that match is usually based on metrics that provide the best chance of customer contact, not the best chance of customer success. This is good for the vendor and great for the platform, but a crap deal for the customer.
The solution is a matching algorithm.
Now, I’m not going to walk you through complicated math in a blog post. But I will give you the framework for basic algorithmic matching that results in happy customers, happy vendors, and a profitable platform.
The best match isn’t going to be a simple score. It’s going to be a matrix. Here’s a simple way to think about how your matching matrix needs to come together.
Think of your algorithm as a puzzle with a solution that puts your customer and vendor on the same graph. That graph has axes. Maybe the horizontal axis is cost and the vertical axis is time.
To some customers, price is everything. To other customers, time is more important. Spoiler alert: Most customers will land somewhere in the middle.
Some vendors can get the service done more quickly, and those vendors might be more expensive. So a customer who is willing to pay more for their time has a better match with the vendor who is more expensive but faster. At the very least, they should be in the same quadrant.
But note that I said that it would be a better match, not a perfect match. What about the customer who doesn’t want to drive as far to save money or get the service done quicker? That’s another graph.
The more of these graphs you can layer on top of one another, the better match you will make.
To be successful, your marketplace needs to deliver the service in a new and unconventional way. Otherwise, your marketplace is still just a directory of vendors, and so is Google, and Google will beat you.
Thus, your platform will need unique differentiators in the way the service is transacted and executed. The more of these differentiators you have, the more valuable your platform. Your matrix needs to conform to the value proposition that fits those unique differentiators.
A simple, universal example: Let’s say the service your marketplace offers is fitness training via video chat. I know this is not a very unique differentiator, but it’s easy to understand. Your matrix not only needs to score the vendor on how well they do fitness training for cyclists but how well they do fitness training for cyclists over a video chat.
What are the differences between cyclist fitness training in person and cyclist fitness training over video chat? Those are the kinds of questions you have to answer before you build your matching matrix.
Vetting vendors is your first shot at getting a baseline for their scoring. It also ensures that all your vendors meet a minimum standard to be on your platform. A lot of platforms don’t use a minimum standard — until the first time they get sued.
Vendors can’t be all things to all people, but let me assure you that they will tell you they can indeed be all things to all people. This is especially true if you put a Google form in front of them.
There should always be some human interaction when vetting, but that doesn’t scale. So at some point, you’ll need to automate enough of the vetting to at least ensure all your vendors meet your minimum standard.
Just like vendors can’t be all things to all people, customers can’t have everything they want, right now, and for one low price. You have to baseline customer needs too.
Now here’s the problem with setting a customer baseline: You can’t and shouldn’t “vet” your customers. Not only won’t it scale, but forcing a customer to interact with a human or complete a survey before they can transact is a huge amount of friction.
You need to remove the friction from the customer side of the process, but you still need to know something about the customer’s needs to make a good match. It’s almost like you have to find all this out via magic.
There’s only one way to do this:
You’ll be able to refine your customer scoring as they transact within your system. But you should have some kind of initial baseline to start with.
Referring back to Step 2 — Your marketplace needs to deliver the service in a new and unconventional way.
You’ll be selling all the advantages of your new, unconventional service to customers. But vendors will be used to a certain way of doing things — their way might not be what you’ve told your customers to expect.
You should have multiple sets of guidelines, best practices, scripts, steps, or instructions for every variant of your matching matrix. If you’ve promised a “fast” service to the customer based on their needs, your vendor and your customer need to be on the same page as to what “fast” means.
Now you’ve got the ability to make an initial match, but that’s only the beginning of the process.
To ensure the continued accuracy of your matching, track your matches and how often a match triggers contact. The more contact, the better the match.
Then of course, if you can get feedback from the customer on the success of the service itself, you can use this information to update the vendor scores. But you have to be careful.
You can completely derail a vendor by calling this feedback a rating and tagging it to the vendor. The truth is, if the service wasn’t a success, it’s probably on the platform for either putting a bad vendor in front of a customer or mismatching a customer and vendor.
To solve this, the feedback should include two parts: How accurate the customer felt the match was and how successful the service was.
You can get a sense of why this kind of algorithmic matching is rarely done. It’s difficult to do, it’s expensive, and if you want it to remain accurate, it needs constant review.
But when done right, proper matching is the difference between a truly successful two-sided marketplace and just another directory of service providers.
Joe Procopio is a multi-exit, multi-failure entrepreneur. He is currently the Chief Product Officer at Spiffy, on-demand vehicle care and maintenance startup. In 2015, he sold Automated Insights to Vista Equity Partners. In 2013, he sold ExitEvent to Capitol Broadcasting. Before that, he built Intrepid Media, the first social network for writers. You can read more and sign up for his newsletter at www.joeprocopio.com