Using standardized data to create the next giant services marketplace
To create a universal services catalog, build a template that works for a niche service, then expand using generative AI techniques - "Everything Starts Out Looking Like a Toy" #156
Hi, I’m Greg 👋! I write weekly product essays, including system “handshakes”, the expectations for workflow, and the jobs to be done for data. What is Data Operations? is a post that grew into Data & Ops, a team to help you with product, data, and operations.
This week’s toy: the sole Barbenheimer tie-in of this newsletter is to note that Oppenheimer is being shown on 70mm film with the help of software that emulates a Palm Pilot. It’s a bit meta to be sure, since the original hardware Palm Pilots are no longer in use. Edition 156 of this newsletter is here - it’s July 31, 2023.
If you're reading and haven't subscribed (yet), give it a try! And if you have any comments or are interested in sponsoring, hit reply.
The Big Idea
A short long-form essay about data things
⚙️ Using standardized data to create the next giant services marketplace
If you’ve hired a service provider recently to do work on your house or searched for a new car mechanic or a new hair stylist, you’ve encountered one of the areas of the economy that have been slower to standardize and digitize than other things you might use your phone to order.
You can order a ride, get food for delivery, and do many things from a mobile app while having a high degree of certainty for what you’re going to get. But there are some key areas where the service experience is harder to predict before the actual service is delivered.
When the service might be a little different at every house you visit, how do you know if it’s correct? If you’ve never changed a light fixture, how do you know whether it’s done well? If you’ve never fixed a leaky faucet, do you have a clue whether the plumber is doing it right? If you have never cut someone’s hair, can you guess whether it’s going to turn out well for curly or straight hair?
Experience is everything. How do you solve this disconnect? Word of mouth from friends, online reviews, and other information can help point you in the right direction to have a good service experience, but really understanding what you’re going to get and how it matches the problem you’re trying to solve can be super hard.
You’re probably not a plumber, an electrician, or a hairstylist
Dan Hock has written a great piece on the future of service marketplaces. These are places where you bring together demand (people who want the service) and supply (the providers who offer this service, brokered by the marketplace. Hock describes the service catalog: a key element needed for service marketplaces. In a marketplace for services, every service needs a SKU (stock keeping unit) that corresponds to something like the Amazon ASIN (Amazon Standard Identification Number). In other words, every unique service needs a unique description.
This works pretty well for highly understood services like those performed by plumbers and electricians. There is a standard and building code-approved process for installing a toilet, changing a ceiling fan, or replacing a light fixture. For the functional requirements of a job, you can absolutely define whether the service has been successful.
The key problem with service description? When the service requestor (you) has specific requirements about how the service should be completed, but has a hard time telling you (the provider) exactly how to meet those requirements to a high degree of success.
Identifying the elements of a great light fixture install is not too hard (well-mounted, looks good, works when it’s done), but try the same description for an amazing haircut. It’s hard to put your finger on the exact service description to make that happen.
Because you’re not a plumber, you don’t have a good way to describe the way your plumbing job differs from any other plumbing job that might have been done before.
A good service description is just the start
The argument for service marketplaces goes that if you are able to build a flexible service description, you’ll be able to solve any service need without mismatching the skill of the provider, the time needed to complete the job, and the complexity of that job.
This needle-threading is exactly what we tried to do when we started up the services marketplace between homeowners and contractors at Pro.com back in 2013. We built a catalog of plumbing services and other categories over a series of weeks, identifying the most common jobs and writing detailed descriptions.
The problem with matching the description of an ideal service with the actual experience? It’s hard to find a service provider who is good, will be able to complete the service, and know the expected price for that service. Did I mention that you’d also like the provider to arrive on time, leave the place cleaner than they found it, and make any needed recommendations for other plumbing work you might need? It sounds like you’re asking for a unicorn in addition to needing a leaky faucet fixed or a garbage disposal installed.
And it turned out that this was true. This marketplace (like many others) had a supply problem. Fundamentally the best resources for regulated or certified jobs like plumbers or electricians are always busy. To make those marketplaces go, you need to create more supply. Training more electricians or plumbers could take years.
Marketplace growth comes from long-tail services
But what if we could create structured service data from the requests that people are making? For many non-regulated businesses, you could create a standardized way to describe that service, making it easier for customers to know if they are receiving good service and for providers to deliver an excellent interpretation of that service.
But which service would you start with, and how would you know if your catalog item made any sense? In 2013 at Pro.com, we had no way to do this task beyond interviewing customers, identifying the most popular services, and authoring a service catalog.
In 2023, there are entirely new generative tools that make it possible to take the consumer description of a service and transform that simple description into a service catalog entry.
What data do you need to describe a service?
The template for a successful service is relatively straightforward to describe but hard to get right.
A service catalog needs:
categories that people understand - these are canonical items like “plumbing” or “electrical” that match the kind of provider who can deliver the service
synonyms between service offerings - these are other words people might use to describe the service colloquially but aren’t the standard, formal way to name the service
well-written service descriptions - this is a content management task to produce understandable summaries for the service that are short, medium, and lengthy
exception definitions - in the customer journey, there will be mistakes. Understanding what these look like ahead of time improves the possibility of success
a price or modifiers to the price - so that you can know if someone is charging an unusual amount for a canonical service
For a long-tail service, no person is going to write the catalog version of this service description unless it’s done with a generative process like ChatGPT and then edited and corrected by suppliers and buyers.
How would you start building the supply for this services marketplace? You could start finding providers from Etsy, Angie, Yelp, Fivvr, and Upwork, and asking them to describe what they do in a video, a voice mail, or a short description. Then, use a generative process to create a provisional catalog page. You’d need to filter these through a human verifier or other kinds of filters to confirm you get the kind of services you want to have in your catalog.
Now that you’ve got the bare bones of a service catalog, you can start facilitating appointments, taking payments, and validating whether buyers feel that the services offered are meeting the expectation set by the service.
It doesn’t matter (yet) whether you have achieved a brand that people recognize; it does matter that the data required to describe a single service and its outcomes can be easily compared to another service in the system.
Solve the long tail with great tools and you can compete in a crowded market
The trick to building a great services marketplace? I think there is no trick. You need to be able to deliver differentiated service for a single provider and validate that the service was both objectively good (matches the service catalog describing an archetypal service) and subjectively good (appreciated by the customer).
The giant services marketplace that Dan Hock describes is not going to look like the aggregation rollups achieved by Uber, Doordash, or similar. It’s also not going to look like a collection of all of the best Electricians, Plumbers, and Handypeople. Those people are already busy.
The services marketplace that Dan Hock describes is going to emerge from a better service catalog that emerges automatically from customer questions and is carefully matched to providers who can achieve high CSAT results from niche categories that don’t require a traditional certification or a license.
By creating experts in a category of one, the giant services marketplace of the future will identify services where people are willing to pay a premium for an unusual thing done well. What will that look like? I’m not sure, but it will be another service in an interrelated catalog that will eventually be a pretty interesting knowledge asset.
What’s the takeaway? Building a brand to achieve success in a marketplace starts with well-defined data that identifies the service to be completed, the exceptions to be handled, and the category expected by the buyer. But the next great marketplace will do this by emerging the categories automatically based on the questions posed by the buyer.
Links for Reading and Sharing
These are links that caught my 👀
1/ 1973’s view of computing - take a trip in the Wayback machine to 1973 and learn how people viewed computers. These computers read cards at the rate of 200 cards per minute, and terminated with an EOF (end of file) card.
2/ How to use AI to do stuff - Whether you consider yourself a Luddite or you think generative AI is the next best thing since sliced bread, Ethan Mollick has suggestions for you to make the most out of AI. Who knows? You might learn something.
3/ A novel way to pay less attention to your phone - Use two! Seriously, this is an interesting idea to create a “boring phone” and a “dopamine phone” (Trung calls these “Kale” and “Cocaine”, which is maybe extreme, but you get the idea). By alternating the exposure to “boring” and “exciting”, you might be able to break away from the phone a bit more.
What to do next
Hit reply if you’ve got links to share, data stories, or want to say hello.
Want to book a discovery call to talk about how we can work together?
The next big thing always starts out being dismissed as a “toy.” - Chris Dixon
This article is a fascinating take on creating a standardized services marketplace using generative AI techniques. I can't wait to see how this approach will revolutionize the way we experience services. Keep up the excellent work.