Over the past decade, a number of generational consumer-facing marketplaces have been built on the internet: Uber, Airbnb, Doordash. The wave of technologies that enabled these businesses (mobile, payments processing), and the tailwinds that drove exponential growth (adoption of e-commerce during COVID-19 lockdowns) have accelerated the growth of a number of B2B marketplace platforms as well. Amazon Business had $10B+ in sales in 2019, and Bank of America projects that it could generate $125–245B of EV for Amazon as a whole, even assuming a small fraction of share of the $1.4T market for B2B ecommerce.

In this post we explore how new entrants can build valuable businesses to capture this opportunity.

Framework for B2B business models

We start with a core framework to analyze different B2B business models, and the archetypes that result from the underlying supply-side characteristics (e.g., variability) and buyer preferences (e.g., level of trust / validation required for the transaction, willingness to adopt software).

Frequency / retention: the first piece of the framework is whether or not both sides of the platform transact at a high frequency, with a strong cohort of “power users.” The best B2C marketplaces have this characteristic, e.g. Uber, Turo, Airbnb. Similarly, best-in-class B2B marketplaces need to have high repeat rates in order to get operating leverage at scale.

• Lenny Rachitsky and Casey Winters point out in this excellent piece that NRR is more useful than gross or logo retention, especially in cases where CAC is low and there is long tail upside from the “power user” cohort (e.g., in the case of Shopify).

Variability of supply: an important axis when evaluating a B2B business model is whether there is a high degree of heterogeneity on the supply side, and therefore the platform can provide the functions of providing curation, discovery, and trust to the transaction. In cases where this is true, the business model is a “marketplace.” In cases where this is not true, and the supply-side is commoditized, the platform’s main function is to digitize a procurement process and move the same transaction that would have happened in an analog manner to a digital process.

• For the most part, more heterogeneity on the supply side leads to more network effects and platform stickiness. In some models there is diminishing returns to new supply beyond a point, e.g. the utility of ridesharing platforms plateaus at around a 4m waiting time, and there is limited value in reducing wait times beyond that. On the flipside, the marginal unit of supply on Airbnb can have infinite utility for the demand side. In the B2B setting this analog for infinite supply-side heterogeneity is most strongly reflected by Upwork and Fiverr, where each freelancer can have sufficiently different skills and style to satisfy an incremental unit of demand.

Operational complexity: whether or not the supplier is willing to adopt software determines the operational complexity of selling to them — for industries where software adoption is slow, a seller often needs to vertically integrate into providing the end service. The degree of operating complexity determines the margin structure of the business, and influences how platforms should think about pricing and expansion strategy. For example, a managed marketplace should have large AOVs and high take rates, given the additional operational lift of supply qualification required as compared to more fragmented B2C/C2B marketplaces.

We dive into the resulting 4 archetypes in more detail below:

Fragmented marketplace with B2C/C2B qualities

In this archetype the supply-side is highly fragmented and heterogenous. Buyers value the platform’s ability to offer a breadth of supply, and to provide both curation and qualification of supply quality. Transactions tend to be frequent and relatively small. For example:

• Frubana has independent restaurants on one side and growers on the other, and restaurants value that the platform can fulfil their procurement needs across a breadth of categories
• Wonolo has gig workers (independent blue-collar workers) on the supply side. Instawork has gig workers on the supply side and SMBs on the demand side. In both cases employers want local high quality workers, often with specialized skills
• Faire has small boutique stores on the demand side, who want to curate a variety of merchandise
• Upwork and Fiverr have individual white-collar workers on the supply side with a broad range of skills represented (e.g., design / creative, admin support)

In many cases, these platforms are initially replacing an activity that is already happening with a faster, better, and cheaper service. The best version of this archetype expands TAM because it enables the buyer to grow their business. For example, Wonolo enables some customers to expand in geographies they might not have been able to profitably serve before, because they can now spin up local labor supply to complete those projects.

Metrics that matter: given the fragmentation of the supply-side, CAC payback (target: <6 months for supply side) is an important metrics to track to ensure it can be acquired efficiently.

Managed marketplace

In this archetype, the supply side is more concentrated or limited than in the B2C/C2B archetype, and the demand side has a stronger need for qualification of that supply. This results in higher operational intensity to facilitate matches than in the B2C/C2B archetype, and lower frequency transactions with higher AOV.

Usually there is a barrier to entry on the supply side (e.g., required licenses), and the platform is providing the function of validating the credentials of the supply side. In many cases platforms forward integrate into providing that credentialing to increase available supply. For example, Seso helps workers with the H-2A visa process, which helps them increase their available supply on the platform. Wonderschool helps people launch licensed childcare centers, and assists them with all aspects of scaling their business.

There is some risk of disintermediation in this model given the relatively constrained supply side. To build a protected platform, the demand-side needs to value variety (and, therefore, discovery) on the supply-side, and the supply-side needs to add value to the transaction (e.g., in the form of credentialing, financing, or digitizing).

Metrics that matter: share of GMV from retained cohorts (target: >70%) matters a lot in this model, given the operational intensity of the model. Operating leverage can come over time from scale in customer acquisition costs as a larger share of GMV comes from retained customers over time.

Vertical SaaS with transactional component

When the supply-side is relatively homogenous and the buyer is willing to buy software, the primary value of the platform is in digitization of the core B2B procurement process. There is limited value to discovery or curation, so to create stickiness companies offer a vertical SaaS procurement tool to digitize workflows, get into the flow of purchases, and build marketplaces on top of that. Given the relatively commoditized supply side, take rates on these transactions tend to be relatively low (~2%) or even flat. Additional monetization is possible, e.g. through payment processing services.

For example, Cvent provides an event management platform for a SaaS fee, and charges users a per-attendee transaction fee. Given that Cvent does not add any value in finding or qualifying attendees this fee is a volume-based flat per-registrant fee. The resulting business is ~80% SaaS, 13% transactional, and 8% professional services revenues.

Metrics that matter: traditional SaaS benchmarks and metrics apply here, e.g. rule of 40, magic number.

Tech-enabled service provider (retailer / broker / distributor)

In markets that are slower to adopt new software (e.g., shipping, agriculture), platforms often need to vertically integrate to fulfill the procurement service as a retailer / broker / distributor, and become a 1:1 replacement for the incumbent. The value add of these platforms over incumbents in these cases is to

• [Retailer or distributor] Bring the benefits of e-commerce to the transaction (e.g., digital merchandising, visibility, analytics) and provide the functions of logistics and fulfilment to the buyer
• [Broker] Increase speed and lower costs of matching demand and supply, and provide services to enable the transaction (visibility, payments processing)

In this model, the platform is matching and then fulfilling high AOV and low take-rate (or fixed fee) transactions for relatively commoditized or pre-vetted goods. For example, Shelf Engine is building a next-gen grocery distributor. Their value add is not in the curation or validation of SKU assortments, but in using AI/ML to optimize purchase volumes and merchandising and reduce shrinkage at the store.

In cases where goods are pre-vetted, the platform does not need to provide a curation or validation function, but is instead adding value with digital merchandising and fulfilment. For example, Material Bank charges a flat “sampling fee” rather than a percent of the transaction to discourage platform disintermediation. In their case there is a relatively fixed universe of buyers and sellers, but a massive set of potential SKUs. The value of the platform comes from the convenience of visualizing options in a central place and outsourcing fulfilment, and Material Bank has built a highly sophisticated fit-for-purpose automated fulfillment system for their particular single-SKU supply base.

Given the operational intensity and low take rates these models tend to require a significant amount of capital to scale. Cost of capital and access to competitive debt / working capital financing is therefore important in this model.

Operating leverage at scale typically comes from increasing volume through the cost base and reducing unit costs (e.g., amortizing fulfilment costs over a larger volume, automating fulfilment operations).

Metrics that matter: share of wallet for the customer matters a lot in this model, as a “winning” outcome is one where the platform becomes the retailer / broker / distributor for the market and has high throughput.

Are B2B marketplaces good businesses?

We can look to the public market performance of B2B marketplaces to test the archetypes above, and compare them to B2C models. A few highlights when comparing them side-by-side:

• B2B with NRR > 100% is more attractive than its B2C counterpart: typically, B2B businesses have an advantage in customer acquisition costs, and B2C in customer retention. A B2B business that also has high retention is therefore doubly attractive. Particularly in high ticket value / low frequency categories (e.g., car buying), B2B can be more attractive than B2C because of the ability have NRR > 100% and therefore gain operating leverage
• Many-to-many is valuable: Across B2B and B2C, highly fragmented C2C / B2C / C2B platforms tend to outperform tech-enabled retailers. Upwork and Fiverr are the most fragmented of the B2B comps, and the heterogeneity of their supply gives them the “Airbnb effect” of infinite marginal utility for each incremental unit of supply
• “Winning” B2C platforms get big — very big. DoorDash and Upwork trade at a similar EV/sales ratio (12–14x), even though DD is 10x the size of Upwork from a topline perspective. I infer from this that the market sees a lot of upside potential for DD even at this massive scale

In summary, there is a huge opportunity for businesses to capture the $1.4T of B2B e-commerce opportunity for both labor and materials across multiple verticals. The business model that will best serve a given market is dependent on the nature of supply (particularly, the heterogeneity), buyer behavior (willingness to adopt software), and the degree of trust required for the transaction. For each of these models, there is a slightly different metric that companies should be laser focused on — but net retention is the one metric to rule them all.

Highly non-exhaustive market map of B2B marketplaces