1000x an idea

MEMO To: Marcus Webb and the exec team From: Staff PM, Product Re: The 1000x version of the tip-out calculator, and what to build next quarter

---

1. The mechanism: our workers are a sales channel

Kenji's calculator was built to save managers 25 minutes a night. The beta shows that it also turned workers into a distribution channel.

• Weekly worker-app use at beta restaurants rose from 34% to 81%. Workers open the app at close to see that night's tips.
• 41% of those workers also work at another restaurant.
• 57 restaurants signed up in one quarter because a worker asked them to. That is 0.19 per beta restaurant per quarter. Those deals closed in 9 days instead of 41, at about a third of the sales cost.

Today we grow by selling to owners: 380 restaurants a quarter, about 6% of our base. In the beta, each restaurant generated referrals equal to 19% of itself per quarter, through workers who carry Tally into their other jobs. That changes the shape of our growth curve. A curve that compounds on its own is what gets us to a much bigger company. A longer feature list does not.

The one dimension I would push is the worker, not the owner. We should make Tally the worker's own account across every restaurant they work at. Today we treat workers as users of each restaurant's account.

The data shows workers already try to spread Tally. Across the ~9,000 beta workers, 29% said they asked a manager at another job to use it. That is roughly 2,600 asks, which produced 57 signups. Demand is not the bottleneck. Conversion of worker asks is. Nobody currently sees those asks, routes them, or follows up on them.

2. The most ambitious version

Tally is where every hourly restaurant worker sees and gets their pay, across every job. A worker has one linked account and sees tonight's tips from both jobs. They can cash those tips out the same night. With one tap, they can ask their other manager to join. Every restaurant that joins brings in more workers, and those workers bring in more restaurants.

Later, once we hold the licences, that account becomes the base for earned-wage access.

Sizing from our data (ARR per restaurant ≈ $31M ÷ 6,400 ≈ $4,840):

Caveats. The high case compounds a rate from 300 restaurants that asked to join the beta. It also assumes no churn (we have no churn data here) and no local saturation. The payout range rests on stated intent, and "would pay" usually overstates real use. So treat $70M–$185M as the range to test, not a forecast.

Even the low case is roughly double our current growth rate, reached at a third of the acquisition cost. That low case is not 1000x. The high case gets us to a $1B path within a few years of compounding.

Why this matters against Crewbook. We lost 11% of deals last year on price, mostly to Crewbook. Crewbook sells to owners, while this channel reaches workers. Referred deals cost a third as much, which gives us room to compete on price where we need to. The calculator also enforces state tip-pool law, so an owner can't easily switch away without losing that protection.

3. The sequence

Next quarter: two squads

Squad A: roll out the calculator and build the referral loop. - Roll the calculator out to about 1,500 restaurants that did not ask for it. This tests whether the beta rate survives outside self-selection. - Build an in-app "ask my other manager" flow with tracking of where each ask comes from. - Route every worker ask to sales within 24 hours. - In the second half of the quarter, if the first gate below is met, pilot instant payout of earned tips at beta restaurants. This needs no licence, costs $0.99 per payout, and Tally keeps $0.60.

Squad B: build the linked worker account. - Linking a worker's accounts with their consent is about one squad-quarter. - It is the foundation for everything after: one view of tips across jobs, payouts across jobs, and a worker identity that belongs to the worker rather than to one restaurant.

Gates before each next step

1. Before rolling out to all 6,400 restaurants: - The rollout cohort produces at least 0.08 referred signups per restaurant per quarter (roughly $2.5M+ new ARR per quarter at full base). - Referred deals still close in under 15 days. - Worker weekly use exceeds 60%. 2. Before scaling instant payout: - At least 10% of pilot workers pay for a payout in a month. - At least half of those pay again the following month. 3. Before committing to earned-wage-access licences (9–12 months in 3 of our 5 core states): - Linked accounts cover at least 30% of multi-job workers. - Payout usage meets gate 2. - Compliance can prepare the applications in parallel, since that uses no squad time. We only file once the gates are met. 4. After that: owner products built on network data. For example, labour benchmarks across restaurants. This is where Marcus's vision comes back in (see section 4).

The result that would make us stop

If the non-self-selected cohort produces under 0.03 referred signups per restaurant per quarter, and worker weekly use stays under 50%, the loop was a beta artifact. In that case we would: - stop the network push; - keep the calculator as a retention and compliance feature, since the 25 minutes a night is still real; - move both squads back to the core roadmap.

We will know within one quarter, and it costs us one quarter to find out.

4. What to do with Marcus's idea

My recommendation is not to fund the "AI operating system" next quarter, but to keep it as the destination. My reasons:

• Nothing in our data supports it yet. We have no evidence owners want Tally for inventory, pricing, marketing or reservations. Each of those is a mature category with established vendors.
• We don't have the capacity. Two squads cannot build four products well. Spreading across them would starve the one mechanism we have actually observed.
• It aims at the wrong side of the market. The beta shows our leverage is with workers. An owner-facing OS goes head-to-head with Crewbook and everyone else on features and price. The worker network is something they cannot copy quickly.

Marcus is right about the ambition and the $1B scale. The worker network is how we get there cheaply. It also makes the owner OS far easier to build later. At 20,000–40,000 restaurants, with worker and pay data linked across them, we would have distribution and data that no single-restaurant tool has. Owner products built at that point sell themselves through a base we already own.

Proposal: we treat step 4 as the explicit home for the owner OS. Marcus names the first owner decision he would want Tally to make smarter. We scope it for the quarter after gate 1 is met.

Decision needed: approve the two-squad plan and the gates above for next quarter's planning.