How to select the right matching workflow for your marketplace

I recently had the chance to help a young startup building a marketplace, and it reminded me of the long process we went through to find and refine our matching workflow (the one that help the buyers connect with the right suppliers). I decided to write this post hoping it will help other marketplaces analyze their challenges and come up with the model that fits their need.

If you’re leading a marketplace product, you know how important liquidity is.
Liquidity describes the chances of a buyer or a supplier to perform a transaction, and although it doesn’t have an accurate formula - it is driven by the number of options (number of buyers/suppliers) relevant to a certain service/product, in a given location and time.

Take speed dating as an example: if you invite 6 people to a speed dating event, the chances of some participants to find a match are pretty low. On the other hand, if you bring 100 people in - some of them will probably be able to find a match, and of course, the matching rate can be increased through optimization (gender, age, language…).

That’s the meaning of liquidity: the more availability each side of the marketplace has, the higher the chances are to accomplish their goal.

Early-stage marketplace startups often see liquidity as a matter of reaching certain numbers: you recruit enough suppliers on one side, then bring enough users to the other side, and you are set.

That’s a mistake: every marketplace has a different booking flow and different challenges that impact liquidity and as a result - the fill rate. 

There are many sensibilities involved in a marketplace "matching flow": price, quality, location, timing, experience, personal preferences, time management, and more... What works great in one industry can easily fail in another. 

Storytime: 

[Not interested? Scroll down to see types of booking flows and their pros and cons].

When we started Missbeez (a marketplace for lifestyle services), we estimated the number of suppliers we would need to cover a certain geographical area based on skills, treatment length, driving time, etc.

We created a bidding mechanism where a buyer creates an order, the system dispatches it to the relevant suppliers (service providers), they send their offers to the buyer, and the buyer can then select the preferred one (based on a business card, rating, reviews, photos...).

We assumed that once we reach a certain number of service providers - buyers will always have enough offers to complete the “handshake” and approve the booking.

Here’s what we found: 

1. Buyers' most popular hours (late night and weekends) were the least preferred hours for the suppliers. 
2. Suppliers were very sensitive to order sizes and locations (i.e. traffic, parking issues).
3. Buyers were so sensitive to the identity of the supplier - they often preferred to wait and see if additional offers will come in than to simply approve the booking. 

The numbers didn't work as expected.

In terms of flexibility - we were too “flexible” and the matching workflow was relying too heavily on human actions - it was inefficient, slow, and basically "encouraged" drop-offs.

This model is called: “double commit” and in the past few years marketplaces and VCs are walking away from it because it has a relatively low fill rate.

This problem led us to the realization that liquidity is not just about numbers, it involves additional factors density, trust, look at the different search and match workflows that are available for marketplaces and explore each of them:

1. Supplier pick: 

This is a good model if your marketplace provides homogeneous services such as rides or food deliveries.

Typical workflow: 

1. A buyer creates an order
2. Suppliers get the order as an open “request” 
3. The first supplier to accept the job creates the handshake and finalizes the booking. 
4. The buyer gets an approval that his order is booked

This workflow assumes that the buyers are not sensitive to the suppliers’ identities, as long as they know that the marketplace has certified and vetted them.

This approach fits perfectly with Marketplaces that deal with commodity services and are able to standardize their workflows, but they will have to make sure the product provides high and consistent standards and enough supply to ensure high availability.

Availability and shifts: 

To create an efficient workflow, the suppliers need to log-in and enter a “standby mode” so they can immediately accept any request that comes in.
This approach minimizes the buyer’s waiting time but doesn’t fit all types of services (i.e. Uber drivers are in their cars and can quickly respond to incoming orders, but massage therapists cannot wait next to their smartphones in a middle of a long treatment…).

Use the supplier pick model if you can standardize your supply, guarantee enough of it at peak hours, and your suppliers can respond quickly.  

I’ve heard of a certain beauty company in Europe that tried to solve this “standby” problem by providing wearables to their suppliers so they can work while in this “standby” mode.

Good for: 

1. Marketplaces dealing with commodity services 
2. Where suppliers are highly engaged and quickly available
3. Classic on-demand services (instant booking to “right here, right now” requests) 

Bad for: 

1. Sensitive or highly personalized services
2. Future booking 

[Recommended read: how to minimize leakage in your marketplace]

2. Buyer pick: 

This is a good model if your marketplace provides heterogeneous or highly professional services where buyers are expected to be sensitive to the supplier identity and quality of service.

Typical workflow: 

1. Suppliers enter their availability in advance
2. The buyer can browse through the available supply
3. The buyer selects a supplier and this creates the handshake
4. The booking is added to the supplier calendar

This workflow is the exact opposite of the supplier pick: it turns the buyer into the decision-maker and assumes that the suppliers are not sensitive to the buyer's identity (an assumption that must be validated in advance).

Buyer experience: 

When your marketplace offers a lot of options and high availability - the buyers' experience will skyrocket, but you also need to think about low availability scenarios: If a buyer can’t find the exact time/location he is looking for, it’s important to show other alternatives even if they feel like a long shot. From my experience, many buyers will be willing to compromise on the time or the day, just to get things done.

Availability and shifts: 

To make this model work, the marketplace must provide a friendly tool for the suppliers to manage their shifts, monitor their schedules, and make changes. If the marketplace focuses on future bookings then the suppliers are expected to enter their future availability in advance and commit to them.

If your suppliers are unable (or unwilling) to commit to certain availability hours - the buyer pick model will not work. 

If the suppliers are not “hooked” to their calendars, or simply not committed enough (i.e working privately or with completing marketplaces), they’ll end up canceling shifts or making last-minute changes, causing a bad buying experience.

There’s also an option to “fake” some future availability, ensuring 100% availability during the buying experience, leaving enough time to find the right suppliers after the booking is already secured. I had a chance to “test” one of our competitors that was using this technique, and out of 5 future booking I made, 3 of them ended up with phone call from their customer success team apologizing that “they had to change my booking due to an unexpected event“.

Supplier’s sensitivity to their bookings: 

The Buyer Pick model assumes that the suppliers are not sensitive to the orders they get, but as we’ve learned from our own experience - this is not always the case:
When we introduced shifts in our marketplace we started getting complaints from our service providers that wanted a certain level of control on their schedule:

• Some asked to be able to “reject” certain customers based on a bad previous experience they had
• Some didn’t want to travel more than X miles if the order size was below a certain threshold
• Some asked to avoid certain treatments in certain hours of the day
• Others said they wish to avoid certain blocks due to parking issues.

To make this work, we had to add some “settings” in each shift and let each service provider enter their personal preferences. Those were implemented on top of our scheduling engine that already had a set of rules based on skills, location, and time.

Good for: 

1. Marketplaces dealing with highly professional/expensive services
2. Marketplaces dealing with highly sensitive services (such as beauty treatments) 
3. Marketplaces focusing on future bookings

Bad for: 

1. “Picky” suppliers
2. “Right here, right now” on-demand
3. Marketplaces that cannot manage a reliable calendar

[Recommended read: 11 lessons learned while trying to become a data-driven company]

3. Double commit:

This model is the easiest to implement because it creates a free market where the users are making all the hard decisions by themselves, but it’s also the one with the lowest fill rate because the matching is done by humans.

Typical workflow: 

1. A buyer places an order request
2. Relevant suppliers get the request and submit their offers 
3. The buyer reviews the different offers and selects one. This triggers the handshake. 
4. The selected supplier receives an approval notification

Overall experience: 

This workflow provides the highest level of freedom for both sides of the marketplace: buyers throw out their preferred hour, regardless to what might be the availability, suppliers can decide when to “bid” and when to skip, and buyers get to make the final call about which supplier to select. There’s a lot of flexibility, and users are not committing to anything in advance.

Issues: 

1. The extra selection/approval step doubles the chances for a drop-off (supplier doesn’t feel like accepting, or buyer doesn’t select the offer)
2. All the steps in this workflow are done manually by humans, which makes it much slower than any workflow that includes automated steps. 

The more steps you add to a workflow, the higher the chances it will break.

Example: 

• A buyer creates a request
• X relevant suppliers get it, only a few pay attention, none of them accept it. 
• Problem #1: the buyer is waiting, nothing happens. 
• After a while, one supplier sees this request and accepts it. 
• The buyer gets the notification and evaluates this offer (he checks the supplier details, rating, reviews, photos) 
• No decision yet, the buyer is waiting for other offers to come in… 
• Problem #2: the process is stuck: the buyer doesn’t approve the match, the supplier is waiting for approval, the handshake doesn't happen. 
• After a while, another supplier sends his offer 
• The buyer is now busy driving, so he doesn’t see that notification 
• Problem #3: the process is still stuck, there are currently 2 suppliers waiting to get approved
• Momentum is lost... 

In marketplaces, too much freedom in the selection flow slows down the process, adds friction in both sides, and leads to a low fill rate. 

Double commit marketplaces suffer from low fill ratio even if liquidity is high. If you’re thinking of implementing this model, find ways to speed up the workflow (using nudges, encouraging immediate responses, timers, etc.) and automate some of the steps (allow suppliers to auto-accept certain jobs, etc.), or else your marketplace will create a lot of expectations but will end up frustrating both sides.

Good for: 

1. Early-stage marketplaces, still in MVP
2. Marketplaces dealing with highly sensitive services for both sides (buyers and suppliers) 
3. Marketplaces focusing on future bookings
4. Top-notch suppliers 

Bad for: 

1. Making money… 
2. Closing appointments quickly and easily 
3. “Right here, right now” on-demand

4. Hybrid workflows

Some marketplaces are using variations of the above workflows:

• Some fake the double commit model while in fact manage the suppliers automatically (to make the workflow more efficient) 
• Some remove the selection option from both sides (letting the suppliers commit on shifts and choosing the right supplier automatically per each order) 

Summary

Liquidity is driven by numbers, but it's also depending on the nature of your business and matching workflow. A good matching flow will streamline bookings and skyrocket your marketplace. A bad one will increase drop-offs and frustrate your users.

It's important to fully understand the business and the users before selecting the right model.
How sensitive are the buyers to the identity of the supplier? How sensitive are the suppliers to the booking details, what's the required level of standardization? How immediate is the need vs. future booking?
Answering those questions will help you find the right workflow for your marketplace.


Hey, before you drop off, check out the best of the mobile spoon or subscribe to my newsletter and become 23% more awesome than average.