Build the Model, Then Garden the Marketplace: Dan Hockenmaier on Growth Models and Marketplace Strategy

Hockenmaier began in consulting at BCG and private equity, then learned most of what he knows at Thumbtack, which he joined when it had about 30 people and left after a little over three years in which the business grew more than tenfold. He then built a strategy consulting firm that worked with growth-stage marketplaces, which was acquired by Faire, where he now leads strategy and analytics.

He defines a growth model as an analytical representation of how a business grows, usually built in a spreadsheet. The spreadsheet format is valuable because it is very hard to fake: when assumptions have to line up and link, you are forced to understand the business. He says about 50% of the value is simply building the model, and the result is useful for weighing opportunities, not for forecasting, which remains the job of the finance team.

For a basic SaaS business, he says, there are three building blocks: acquisition channels (paid, sales, viral) with traffic and conversion assumptions, retention with an activation rate and a survival curve, and monetization. A transactional business adds transactions per month, average order value and unit economics. A marketplace adds supply acquisition and retention, and how the two sides interact.

Things get interesting when the model becomes non-linear, as when virality means existing customers refer new ones, or paid marketing profits are reinvested. Linking those explicitly shows why payback period is a better measure of paid marketing than LTV to CAC, since how quickly you recoup money to buy another customer drives how fast you can grow.

Hockenmaier warns that growth models break down when assumptions stack, and marketplaces make that worse because there are two sides to model and some interactions are hard to understand. His example is an Amazon category manager who adds pet food and toy brands: how much of the resulting revenue is incremental, and does more supply make customers retain longer? Without care, he says, you get a junk-in, junk-out problem.

His preference has become one simple, high-level conceptual model describing how the system works, plus a mini model for each product pod or go-to-market team. Each team knows its north star and the inputs that drive it, and he does not try to stitch them into one master model. A growth team's model is usually a roughly linear funnel.

Other teams manage tensions. A marketplace quality team balances supply growth against customer experience, since letting on lower-quality suppliers lifts GMV at first but degrades retention. A fintech team balances extending more credit and higher spend against defaults. He and his consulting firm built perhaps 20 or 30 models, and he suggests first-principles building over templates, with Reforge as the place to go deep.

The first thing building these models shows, Hockenmaier says, is that growth is much more sensitive to retention than anyone can intuit, because a healthy retained base interacts with referrals, content and contribution margin. A small retention gain is often worth more than a bigger change elsewhere, which means resources may be misallocated. His tactical advice is to partner with a smart analyst or finance person, start building, and iterate over several quarters.

In planning, he says, the hardest part is a common currency for comparing teams that each claim to move a different metric, and the growth model is the function that provides it. Analysts and product managers run scenarios through it to produce a spreadsheet of what could be worked on and the likely impact on growth. Individual pods should also use their mini models in their strategy documents.

At Thumbtack, a model built with the finance team made it obvious how sensitive the business was to the repeat rate of new customers, who typically arrived through targeted SEO or SEM for one service among a thousand categories. The team had shipped hundreds of conversion experiments but done little on cross-selling, and the model led them to shift resources from top of funnel to deeper in the stack, which helped build a better customer journey.

Lenny recalls that Airbnb's analysts reached a similar conclusion about retention and found it hard to move. Hockenmaier says retention is the culmination of the whole product experience, so you rarely get there by sending more email or push notifications. In marketplaces it often comes down to depth of supply and how customers interact with it, which means working on core product levers, not growth-product ones.

The biggest wins, he says, come from improving the early experience, when you can teach customers why the product is valuable. He suggests looking for variability: Uber or Lyft drivers may have a bad first week through bad luck and conclude they only make $3 an hour. That is why both companies fight to guarantee first-week or first-month earnings, pulling below-average first experiences up to average.

He puts little weight on correlational analyses such as "our best users do X, so make others do X," because something unique about those customers usually drives the result. He likewise cautions against resurrection efforts, since the churned pool is made up of people who tried the product and decided against it. Focus on new users first, he says, and only spin up resurrection once earlier funnel work is exhausted.

Hockenmaier calls marketplaces a perfect fit for the venture model: hard and capital intensive to start, but very hard to stop once rolling. In most businesses, acquiring more customers means acquiring ever-marginal ones, so CAC rises and LTV falls. In a marketplace, improving liquidity and experience can invert this, so later cohorts show lower CAC and higher LTV.

For health metrics he lists four. The first two are GMV or transactions, which combine both sides, and unit economics, which are often poor or negative early on (he cites Instacart and Uber losing money on orders). The next two are specific to marketplaces and are covered below.

Liquidity, he says, is how reliable the marketplace is, ideally expressed in a dimension customers care about. For Uber and Lyft it is wait time, with a magic moment around four or five minutes, while for commerce it is search-to-conversion. Until a marketplace is liquid, he says nothing else matters, which is why advice is to narrow to one geography or category first.

The fourth is share of wallet: how much of a buyer's total spend, or a seller's business, goes through you. For a retailer on Faire it is how much of the store's shelf came from Faire. Higher share raises LTV and makes customers less likely to use another marketplace, and Hockenmaier says he would take 10% more of existing customers' wallets over 10% more customers because it measures depth, not breadth.

On supply versus demand, Hockenmaier says the answer is both, but advice tends to over-rotate on supply. Supply matters early because it is the product and has more product surface, but he says demand is ultimately the only thing that matters: whoever aggregates the demand in an industry wins, since suppliers will always take customers that make them money. Lenny notes that Bill Gurley made the same point and that the way to win demand may be to acquire hard-to-get supply.

Hockenmaier agrees and adds that you should acquire supply only as far as you understand how it affects demand. At some point extra Uber drivers no longer cut wait times, and extra pet brands no longer improve the pet buyer's experience. Supply has to be framed by the customer benefit it drives.

Rather than focus on buyer-to-seller balance metrics, he now favors dual-sided ROI equations. The cost of acquiring a customer includes the cost of acquiring the supply that customer needs, based on the ratio between the two. For Uber, that is rider CAC plus driver CAC multiplied by the share of a driver needed, compared with LTV to get a payback period, after which you push acquisition out to the payback you are comfortable with.

He notes one exception, which is externalities the equation cannot capture, such as disillusioned drivers leaving for Lyft when demand is too low. Lenny adds that for early marketplaces, most success comes from the same basics as any business, with marketplace theory as an extra layer.

Hockenmaier says every marketplace decision has second, third and fourth-order consequences. Pricing is the example: a SaaS business can draw a curve of price against conversion, but a marketplace's commission interacts with supplier sign-up and with the benefits it can fund for customers, like Amazon's returns and shipping. There is no simple curve, so he says a marketplace operator is like a gardener, with a very light touch, while a SaaS builder is more like a construction worker.

He has learned the hard way that a change in one place can cause an effect two months later that is hard to trace. His main advice is to tread lightly with core incentives and mechanisms, particularly when something is working. Lenny mentions Faire's Carla Pellicano and the large pricing recommendations team she led at Airbnb as an illustration of how complex pricing is.

On expansion, he says the markets marketplaces target, such as local services for Thumbtack and global wholesale for Faire, are so large that beyond a point TAM matters little. What matters more is adjacency, as a proxy for whether you can win it (Instacart into convenience stores before traditional retail), and whether the move strengthens your network effect, as Uber Eats reuses drivers and customers who want both rides and meals.

His third lesson is that product, not go-to-market spend, decides expansion. There is an arms race to reach liquidity first, but he has seen repeatedly that what matters is delivering an incredible end-to-end experience first, even for fewer customers. Good cohorts, such as a smiling curve with rising engagement, give the company and investors conviction, so go-to-market should not get ahead of product.

Asked about marketplaces adding SaaS and the reverse, Hockenmaier says it is easier for a marketplace to move into SaaS. A marketplace already generates demand, a higher-value activity with commissions of 10 to 20% against 2 to 3% for SaaS, and has relationships on both sides, whereas a SaaS company must acquire a whole new type of demand. OpenTable followed the SaaS-to-marketplace route, and he cites Solve in healthcare as a newer example, but he calls it very difficult.

His advice to a marketplace considering SaaS is to first be sure it has a liquid, high-performing marketplace, and then ask what customer pain justifies building a deep set of tools, such as integrations with back-office or accounting systems. Such tools make the product stickier and lift retention. More often than not, he says, it is better to focus on the core marketplace.

On vertical versus horizontal, he says the unbundling thesis is over-hyped. Verticals such as StockX in sneakers worked because trust in eBay inventory was lacking, but most unbundling arguments over-focus on user experience and underweight the scale advantages in customer LTV and network effects. Thumbtack tracked hundreds of vertical competitors, and few gained traction because it could upsell customers into a thousand categories and win on SEM keyword bids.

Exceptions are verticals that are high frequency or high dollar value, like Airbnb versus Craigslist, or that form a self-contained network, as with Workrise (formerly Rigup) for blue-collar work. Lenny adds that oil rig operators are unlikely to be on LinkedIn. Hockenmaier says what you are solving for is customer LTV, and low-frequency businesses face the challenge that customers forget you.

Asked why B2B marketplaces are rare, Hockenmaier says there are fewer founders who understand B2B problems, since most people are consumers, so discovery takes longer. He understood Faire at once because he had run an e-commerce business and dealt with a hundred suppliers, line sheets and PDFs. He adds that B2B often has lower fragmentation, which hurts marketplaces.

He defines fragmentation by asking what share of volume the top 5% of suppliers account for. If ten companies do 80% of the volume, they have their own sales teams and operations, need less from a marketplace and will pay less, and disintermediation becomes a problem. The dollar size of each transaction matters too: a $2 or $3 ride commission is not worth going around Uber, but on orders worth tens or hundreds of thousands of dollars a supplier would rather phone the buyer.

Looking ahead, Hockenmaier charted commission against year founded and found newer marketplaces charge more while doing more. Marketplace 1.0, like Zillow and Home Advisor, aggregates demand at about 5 to 10%. Airbnb and Etsy added trust, while DoorDash and Instacart took over logistics and Faire underwrites transactions, taking on default risk.

At the end of that continuum, he says, you charge 100% commission and are no longer a marketplace, as Opendoor, which is really an e-commerce site, shows. He thinks the deciding variable is how much creativity matters: where customers want a commodity, as with ride-sharing once autonomous vehicles arrive, businesses will consolidate away from the marketplace model. Where suppliers must bring new things, as on Etsy, Amazon, Faire or Steam, marketplaces will stay.