7 Frameworks For Predicting Online Marketplace Success

Hands in a swirling pattern

What makes some online marketplaces billion dollar successes and others costly failures?

Uber and Airbnb are forging ahead relentlessly. Yet Homejoy, having raised $40 million in funding, recently shut down.


How much is down to execution and luck, and how much to fundamental differences between the industries they chose to play in or their precise business models?

What if there was a framework that could help us predict whether an online marketplace would be successful or not in a given industry?

line drawing of man pondering

It turns out that over the last few years a number of smart people have thought about this and shared their frameworks for assessing the potential for online marketplaces.

Let’s take a look at what they had to say…

Framework 1: Ben Holmes of Index Ventures

Ben HolmesIn a post from July 2014, entitled “Autobutler and the Six Characteristics of a Successful Online Marketplace“, Ben described how, when looking at online marketplaces, he and his colleagues at Index Ventures look for a proposition that fulfils 6 key criteria:

For the consumer:

  • Convenience : Is it significantly easier to use the marketplace than not to?
  • Choice: Does the marketplace have a good selection of sellers?
  • Clarity: Are price and service level clear?

For suppliers:

  • Promotion: Does the marketplace provide a way for sellers to efficiently promote their offering to highly qualified buyers?
  • Productivity: Does the marketplace help sellers increase their utilisation or productivity?
  • Practicality: Does using the marketplace fit conveniently into the sellers’ workflow? (A characteristic that can be seen strongly in so-called market networks.)

(They look for strong teams, too, just as they would with non-marketplace businesses.)

Framework 2: Imran Ghory

Imran GhoryIn his blog post Markets for Marketplaces in September 2014, Imran Ghory (previously an analyst at Index Ventures) presented a great list of factors that affect whether an industry is attractive for an online marketplace.

Demand side:

  • Multiple-transaction behaviour (do customers make at least a few transactions per year?)
  • Low vendor loyalty (do customers regularly switch between vendors?)
  • Large market (customer spending of $10bn+?)
  • Expandable demand (is there ‘untapped demand’ that can be unleashed by a marketplace?)
  • Potential to improve consumer experience (e.g. reviews of suppliers)

Supply side:

  • Fragmented (do small suppliers control a large part of the market?)
  • Low conversion/high conversion (do customers typically shop around a lot before making a transaction? If so, a marketplace can have an advantage in customer acquisition costs vs any individual vendor.)
  • Supply-side servicing (can marketplace provide value-add services to suppliers because of their position?)
  • Expandable supply (would a marketplace reduce the barriers for new suppliers to enter the market?)


  • On-platform transactions (can the entire transaction be captured on-platform?)
  • Market open to a marketplace (are the participants willing to move to an online marketplace?)
  • Willingness to pay (are participants willing to pay for a marketplace?)

Framework 3: Bill Gurley of Benchmark Capital

Bill GurleyIn Bill Gurley’s 2012 post “All Markets Are Not Created Equal: 10 Factors To Consider When Evaluating Digital Marketplaces,” he described 10 factors to consider:

  1. New Experience vs. the Status Quo. (Can marketplace offer a better experience for the customer?)
  2. Economic Advantages vs. the Status Quo.
  3. Opportunity for Technology to Add Value. (e.g. providing data about market or facilitating workflow)
  4. High Fragmentation (ideally buyers and suppliers)
  5. Friction of Supplier Sign-Up. (high friction is costly but can be a barrier to entry)
  6. Size of the Market Opportunity
  7. Potential to Expand the Market
  8. Frequency (higher frequency better)
  9. Payment Flow (being part of the payment flow is best)
  10. Network Effects (is experience for customer N+1000 better than for customer N directly because of 1000 extra participants?)

Framework 4: Josh Breinlinger of Sigma West

Josh BreinlingerIn his post “The Ingredients for a Successful Marketplace” from 2014, Josh listed 5 top factors to consider as a framework for evaluating whether a marketplace should exist in a given vertical:

  1. Recurring Usage
  2. Irregular Usage
  3. Standardized Work
  4. Little Trust Required
  5. Non-Monogamous Relationships

Framework 5: Fabrice Grinda

Fabrice GrindaFabrice has written some great articles about marketplaces as well as being an active investor and operator.

In “Thoughts on Building Successful Vertical Marketplaces” from March 2013 he considered marketplaces that specialise in a particular category in contrast to broader multi-category marketplaces such as Craigslist and eBay.

Fabrice identified three factors he believes are important:

  1. Frequency: Users purchase the product or service reasonably regularly.
  2. Value in Specialisation: The vertical marketplace must offer something that is not easy for the horizontal marketplace to offer and truly improves transactions in the category.
  3. Market size

Framework 6: Andrei Hagiu of Harvard Business School and Julian Wright of the National University of Singapore

Julian WrightAndrei HagiuIt’s interesting to think about the similarities and differences between being a reseller (where you purchase inventory) and being a marketplace (where you don’t purchase inventory). In both cases you are an intermediary enabling goods or services to pass from sellers to buyers.

In this HBR article from March 2013, Andrei Hagiu of Harvard Business School and Julian Wright of the National University of Singapore looked at which of the two business models is most appropriate to a given situation.

They argued that the following 4 factors are important in this decision:

  1. Scale effects – high-demand products favour reselling since a large buyer can capitalise on economies of scale.
  2. Aggregation effects – if products and services have much higher value to buyers when bought together than when purchased separately, then this can favour reselling.
  3. The buyer and seller experiences – if customer experience is often poor, reselling tends to work better as it allows more control over the experience.
  4. Market failures – in some cases, marketplaces can collapse, e.g. due to uncertainty about suppliers or large mismatches in bargaining power. Reselling can add stability.

Andrei and Julian also concluded that different models might be preferable at different times. You might, for example, want to start as a reseller and then move towards being a marketplace. By their nature, marketplaces tend to need less capital to operate at a given scale so may be better suited to rapid growth.

Framework 7: Boris Wertz of Version One

Boris WertzIn this May 2015 article, Boris considers what characteristics make an industry attractive for an Uber-style marketplace:

  1. Underlying commoditized services
  2. High purchase frequency
  3. True on-demand use case (In most cases you don’t need a cleaning service or house painter to show up within minutes, or even the same day. But taxis are a different story.)

Further Reading


All these frameworks were designed with slightly different viewpoints in mind and to serve slightly different purposes, so their authors naturally arrived at different results. For venture investors looking for huge exits, for example, the potential size of the market is extremely important. For others, it’s less important.

There are, however, two common factors that appeared in several frameworks as important factors for the success of an online marketplace:

  • Frequency of purchase
  • Potential to improve the buying/selling experience

Imran and Josh both listed a factor that can be a result of Boris’ ‘underlying commoditised services’:

  • Low vendor loyalty

Lastly, one factor that a couple of the frameworks included and that I personally suspect is an important one:

  • Fragmentation of existing players


It’s an interesting time to watch what’s happening with online marketplaces, especially given the excitement around the likes of Uber and Deliveroo and the rise of ‘market networks‘ such as AngelList and Houzz that layer social network and workflow components on top of marketplaces.

Which industries do you think are best suited for disruption by an online marketplace next?

Read More

How Thumbtack Built a Huge Local Services Marketplace


Thumbtack, an online marketplace for local services, yesterday announced a $100M funding round led by Google Ventures.

Building a local services marketplace is a notoriously difficult problem that has beaten many entrepreneurs. So how did Thumbtack get where they are? What tactics are behind their apparent success?

The founder of Thumbtack, Marco Zappacosta, was interviewed by Andrew Warner of Mixergy.com earlier in the year and covered some of the tactics that Thumbtack used to build its marketplace. You can read the transcript of their interview here.

Here’s a summary of the tactics and approaches Marco described:

1. Built Supply Side Using Scraping and Customised Mass Emails

  • Recognising the chicken-and-egg problem of all marketplaces, Thumbtack saw the supply side as easier to get in place and focused on that first.
  • They used web scraping to compile data on local service providers and developed software to process and organise it.
  • They used the extracted information to bulk send customised emails to service providers inviting them to create profiles on the site.
  • They were concerned about being blocked by email service providers so were careful about monitoring the relevance of their emails.*
  • Their technology wasn’t limited to a specific category or part of the country, so they covered a broad range of categories and the whole US from the start.**
  • With their first $6M of funding, they had listed hundreds of thousands of service pros.

Here‘s an example of an email Thumbtack reportedly sent in 2012. What do you think: was this spam?

** It’s interesting to see that Thumbtack succeeded despite going against the common wisdom of focusing on a single city and/or category.

2. Built Demand Side Using SEO (Based on Unique Content of Service Provider Profiles)

  • Service providers creating their own profiles resulted in lots of unique content.
  • As the service provider profiles were crawled by search engines, this resulted in lots of organic search traffic.
  • More and more customers now coming from word-of-mouth and direct.
  • Have started running search and display ads, including retargeting.

3. Tried Different Revenue Models (Commission -> Subscription -> Per-Quote)

  • Initially, tried commission-based model. But had collection issues (hard to tell if an introduction resulted in paid work, lots of providers didn’t pay).
  • Next, tried subscription model. Worked okay, but wasn’t well aligned with marginal value of each introduction.
  • Currently charging providers to respond to quotes. (They can pay per response or buy credits in bulk.)

4. Off-shored Labour-Intensive Tasks

  • 75 of 200 employees are in the Philippines.
  • Main tasks for these employees are: responding to support inquiries (from customers and providers) and helping to onboard service providers.

In summary, then, in terms of building up their marketplace, Thumbtack’s use of mass emails to contacts scraped from the web seems to have been key. Another important decision was to keep their supplier profiles accessible to search engines. This allowed them to grow their demand side through SEO.

At Thumbtack’s current scale and with a large amount of money to spend, it will be interesting to see whether other paid acquisition channels will now start to pay off for them.

What other tactics have you seen used to successfully establish marketplaces? Do leave a comment below.

If you’re interested in tactics for building marketplaces, you make like to read my posts on Getting to Critical Mass: How to Start a Marketplace Business and Skillshare: How to Build a Marketplace for Online Education.

Photo by kellee_g

Read More

What is the Average Churn Rate for a Subscription Box Business?

Foodzie subscription box

I wrote in a previous post about typical subscription e-commerce retention rates and how, when considering launching a subscription box businesses, your assumptions about the retention rate you’ll achieve have a huge bearing on whether or not the business looks like it’ll be profitable.

Lately I’ve come across published details about the retention rates and churn rates of a few other subscription box businesses:

  • According to this interview with Andrew Warner of Mixergy, one of the founders of Foodzie (selling monthly tasting boxes of artisan food) said they had average subscription lengths of 6-8 months. That implies a churn rate of around 14% (and a retention rate of 86%).
  • BarkBox (selling monthly boxes of toys, gifts and treats for dogs) reportedly boasted a 93-95% retention rate in their first year. I assume that’s a monthly retention number. Taking the midpoint of 94% retention rate, that equates to a 6% monthly churn rate or an average subscription length of around 17 months.
  • UmbaBox (selling monthly boxes of curated handmade goods) reportedly claimed a retention rate of 90% (churn rate of 10%) a few months after their launch, with a not-insignificant $18,000 in monthly revenue. That’s equivalent to an average subscription length of 10 months.

It would be dangerous to read too much into these numbers, especially the ones for businesses in their early stages, but they at least give some ballpark numbers that you might find useful.

Note: do bear in mind that there may be some survivorship bias here – businesses with low retention rate would have been less likely to raise money and get written about. So these numbers may be on the optimistic side.

Have you come across any other published numbers for subscription box churn rates or retention rates?

For more information about subscription commerce, see Subscription Commerce Insider.

Photo by ceonyc

Read More

How to Find a Good WordPress Developer on oDesk


You have a WordPress blog or website. You want some changes made to it, so you need to find someone who can help. But who?

One effective way to find a WordPress developer is through a freelancing website like oDesk or Elance. But finding the right person can be very time-consuming and error-prone. How do you make sure a developer you’re considering is the right fit for your task?

Here’s the 9-step process I’ve used:

1. Craft your job posting to help good developers self-select

Write a detailed description of the task you want completed. The more details the better. A serious developer (the kind you want) will read what you’ve written carefully to see whether it’s something they feel confident about doing. They will only apply if they believe they can do it, so give them as much information as possible to help them self-select.

2. Ask the developer to describe similar projects they’ve worked on

You want to find developers who have already successfully completed the kind of task you’re looking to get done, so you want to establish to as high a degree of confidence as possible whether they’ve done that. Ask them to describe similar projects they’ve completed and provide URLs if possible. Ask the developer for a brief outline of how they’d tackle your task.

After posting your job, you’ll likely receive a large number of bids that you’ll need to filter through. Here’s to do that…

3. Remove bids that don’t refer to anything about your task

People who haven’t bothered to read your task description are probably mass-replying to a large number of job postings. Given this scatter-gun approach, they’re more likely than average to not be a good fit for your task. Eliminate them now.

4. Remove any developers who have a less than excellent rating

A quick and easy filter – just get rid of anyone who has a rating of less than 4.5. If they haven’t left a good impression with their previous tasks, there’s a strong chance they won’t leave a good impression on you.

5. Remove any developers who have fewer than 100 hours of work logged on oDesk

Someone who’s completed very little work via oDesk may be inexperienced as a developer or may simply be inexperienced with the oDesk system. Either way, their limited oDesk history means they’re more of a risk. You don’t know whether, if they had built up more history, they would have received good ratings or not. If you still have enough bids to be picky, eliminate all bids from people with fewer than 100 hours of logged work.

6. Review bid details and work histories

Review their oDesk work history and the past projects they’ve mentioned in their bid. In each case, look at what exactly their task was. Was it something very simple, such as installing a WordPress extension, completed for a very small amount of money? Or was it a much more complex task, involving custom development work over multiple weeks?

If you’re looking for a developer to take on a relatively complex project, make sure they’ve completed projects of at least a similar level of complexity in the past. This is very important! Reject any developers who have only done much simpler things than you’re looking for (even if they have great feedback).

Note: this is an easy place to go wrong. It’s not enough to select someone with a lot of hours doing WordPress work if what they were doing was much simpler than what you need them to do.

7. Consider other factors

You may have preferences about such things as the developer’s location, standard of English, or hourly rate. If you still have plenty of candidates, now’s the time to reject any bids that are a clear mis-match against these factors.

8. Hold Skype interviews

You should hopefully be down to around 3 to 6 developers by this stage. Schedule 25-minute Skype interviews with each one (I like using youcanbook.me to give a block of time when I’m available and let the developers pick time slots).

A lot of developers will want to communicate via Skype text rather than audio. Decide whether voice communication is a must-have for you. If it is, make it clear when you arrange the interviews that you’ll want an audio call and that they’ll need a working microphone and speakers or headset at their end.

During the interviews, discuss the task you’re looking to get done and ask the developer how they’d tackle it. Ask them about similar work they’ve done in the past and how they tackled that. As well as getting a feel for their technical fit, you’ll also quickly find out whether they’ll be able to communicate well with you.

9. Make your choice

For a relatively small task, simply hire the developer who seems the best fit and politely decline the other bids. If your first-choice developer doesn’t work out for some reason, you may want to go back to one of the other developers.

For more complex tasks, consider hiring two or three of the developers for a limited paid trial. Agree a trial amount of time or small milestone. Make sure everyone is clear how the process will work. After the trial is complete, hire the developer who has performed best.

That’s It!

If you’ve followed the steps above, you should now have hired a WordPress developer who’s a great fit for your task.

Going through this process can be pretty time-consuming, so you’ll probably only want to do it if you want to hire someone for ongoing tasks or a fairly significant one-off piece of work.

Update: in case you want to take a short-cut, I’m preparing a pre-vetted list of good WordPress developers for $15. Let me know if you’re interested. You can contact me about it here.

Photo by thenovys

Read More

Skillshare: How to Build a Marketplace for Online Education

I always like hearing how entrepreneurs have managed to launch marketplace businesses. TechCrunch recently posted this interview with Michael Karnjanaprakorn of Skillshare, a platform for online education, about how they seeded and grew their marketplace.

He described two main phases in a marketplace’s journey:

Phase 1: Seeding

Where you “Roll up your sleeves up and get it done.”

Michael explains how, at Skillshare, they focused on getting teachers onto the platform as they recognised that teachers already had their own student communities who they’d be able to bring along with them. They reached out to their friends and personal contacts and tried to minimise the friction for teachers to get started, including doing non-scalable things like finding them real-world venues for the classes.

The key points from Michael are:

  • Don’t try to do everything. If you can focus on one side of the marketplace and get them to bring in the other side, then do that.
  • Reduce friction and get to liquidity as soon as you possibly can because once you do, powerful network effects then come into play.
  • At this stage, don’t worry if what you’re doing doesn’t scale.

Phase 2: Scaling

The labour-intensive approaches used in the seeding phase tend not to scale, so you then need to transition to building things into the platform to power further growth. Skillshare are currently improving their product to deliver more value to their users and help users create content and resources that, in turn, bring in more users.

Further Reading

Photo by: Emilian Robert Vicol

Read More

What Makes a Good Ecommerce 2.0 Startup?

PandoDaily have been running a series of articles lately on ‘ecommerce 2.0’, featuring video clips from a dinner with the CEOs of some of the top ecommerce startups: Birchbox, Warby Parker, One Kings Lane and Jack Threads.

I’ve rounded up the clips from the series here as they’re worth a watch if you’re interested in some of the high-level trends in ecommerce at the moment.

[ Edit: the videos are currently unavailable :-( ]

Some key points put forward:

  • Adding commerce to an existing content business can be extremely tough to pull off successfully.
  • Ecommerce businesses may increasingly look to generate demand rather than just fulfil demand.

What do you think will be the key success factors for ecommerce companies starting now? Have the CEOs in the videos got things right?

I’ll try and keep this post updated as more videos are added to the series.

In the meantime, if you’re interested in more thoughts from Ben Lerer, have a look at this post.

Read More

Our Share of Visits from iPads is up 23% Since Christmas

iPad Share of Visits Up 23%

In the month before Christmas, 13.2% of our unique visits came from iPads. In the seven days since Christmas, that has leapt to 16.2%. That’s a whopping 23% increase in a matter of weeks.

I guess plenty of our customers received iPads for Christmas!

Is It Just Technophiles?

Who are these gadget-hungry consumers, you ask?

Well… it turns out they’re not Fab’s hip design-lovers. And they’re not a bunch of geeky dads who finally have a much-anticipated tech gizmo to play with.


Our site sells knitting wool.

Our customers are, by and large, knitters. Someone looking for a dozen balls of Rowan Pure Wool DK or a new pair of knitting needles isn’t the most likely to be using the latest technology. And that’s why this is particularly noteworthy. These are everyday people — a fair proportion are middle-aged women — and, according to Google Analytics, 26.0% of them are now using mobile devices to visit our site (up from 22.7% before Christmas).

The One Mobile Device You Should Worry About Above All Others

So what should we take away from this?

First, in case you were trying to ignore it, ‘mobile’ is clearly a very real factor in today’s e-commerce. Significant numbers of ordinary people are now using mobile devices to buy stuff online.

But let’s be more specific. ‘Mobile’ might conjure up images of smartphones. And that wouldn’t be accurate when nearly two-thirds of our mobile visits are from just one familiar (non-phone) device. At least for now, if you run an e-commerce site, there’s one mobile device I suggest you worry about before all the others: the iPad.

How about you? Have you also seen a big increase in iPad traffic on your site? If so, how are you optimising your site for it?

Read More

Magento E-Commerce Checkout Conversion Rates

Screen Pages, a UK ecommerce agency, publish some great data on their blog. Their latest post is about Magento e-commerce checkout conversion rates.

It’s interesting for a couple of reasons:

  1. It gives some useful benchmark conversion rate numbers.
  2. It shows how conversion rates can and do vary significantly between sites with checkouts that are functionally almost identical. (Clearly other factors are at play, e.g. brand, initial purchase intent, removal of any impediments, etc.)

Here’s a quick summary of the findings.

Data collected:

  • Analysed 2.5m visitors to over 27 websites.
  • All used standard Magento ‘accordion-style’ checkout page (customised for each brand).
  • All niche brands selling lifestyle goods.


  • Average conversion rate (visit to sale): 2.47% (ranging from below 1% to over 10%)
  • Average % of visits leading to a basket page view: 5.76% (mostly between 3% and 7%)
  • Basket page exits: 9.29% (mostly between 7% and 11%) [people who leave the site at the basket page]
  • Overall checkout completion rate (basket page view to sale): 37% (mostly between 22% and 48%)



Read More

Perceived Value vs. True Value: Is There a Difference?

This is a great TED talk: Rory Sutherland makes an entertaining case for thinking more about psychology in the way we design and enhance things (and less about technology).

One example: one of the best improvements to the London Underground in ROI terms did nothing to change the trains or how they ran. Instead, it was to install displays on platforms to show passengers when the next trains were expected.

If you have a few minutes spare, it’s worth a watch.


Read More

Google +1 Recommendations Rolling Out to All Users Today

According to an email I’ve just received from Google, they are today rolling out Google +1 Recommendations to all users.

When you hover over a +1 button, you may now see a popup showing other pages on the site that your friends have recommended via +1.

My account has been in the beta test for the feature for the last few weeks, but I must admit I hadn’t noticed any recommendations until today. I guess I just hadn’t been using those +1 buttons enough!

Anyway, it’ll be interesting to see if this extra functionality is useful to anyone. If users do find it a useful way to discover new content, perhaps it could encourage a few more people to actually click the +1 button as well. For now, I haven’t found the recommendations particularly interesting, so I remain a bit skeptical about the new functionality. What I think would be useful to users is if the +1 button recommended content on other sites. But that would be a tough sell to the website owners who need to choose to include the +1 button in the first place and who wouldn’t want to lose visitors to their competition.

How about you? Are you seeing any interesting recommendations?


Read More