yumipop — the acquisition model

Who do we sell to?

Three ways in. Builders are the funnel; managed agents and platform whales are the revenue. Set how many of each you reach at full build.

The book

—

What's each worth — and what's it cost to serve?

ARPU is our net fee per month. Our edge cost is a slice of the usage inside that bill, computed to the cent. Margin falls out.

Gross margin

—

Assumptions — how infra is computed

Our cost per $1 of metered usage10%

edge + DB + storage; ~10% ⇒ ~90% block margin

Always-on baseline infra / mo$300

Builders — % of bill that's usage20%

Managed — % of bill that's usage45%

Platform — % of bill that's usage90%

Builders — fill rate35%

% of ad requests that become served impressions

Managed — fill rate55%

Platform — fill rate70%

What's it cost to acquire — and keep — them?

CAC is what we spend to land one. Churn logos out; expansion grows the survivors. Together they set net revenue retention — the metric that funds usage businesses.

LTV : CAC (blended)

—

Assumptions — retention, expansion & sales cycle

Segment	ARPU/mo	Margin	NRR	LTV (36-mo)	Payback	LTV:CAC

What's it cost to run us?

The team and overhead that don't scale with usage. Infra from step 2 sits on top and grows with the book.

All-in monthly cost

—

Headcount4

Avg loaded cost / person / mo$11k

salary + tax + overhead

Assumptions — the rest of overhead

Tools & software / mo$2k

Other / G&A / mo$7.5k

Can the round get us there?

Set the goal and the ramp. Platform deals carry a sales-cycle lag, so cash goes out before revenue lands — that's the dip. The verdict tells you if you clear the hill alive.

ARR at month 36

—

Cash in bank (the round)$1.0M

Goal — reach ARR of$1.5M

… within24 mo

Ramp — months to the full book18 mo

Downside test — lose whales mid-plan

Shock month (0 = off)0

% of platform accounts lost25%

Simulates a concentrated-book shock: a permanent step-down in platform revenue at the shock month.

—

Cost to build the book (one-time)

—

Break-even month

—

Lowest the cash gets

—

Cash in bankARR Break-evenGoal reached

—

What's real vs. what's a dial. Pricing is live product ($19/mo Pro; managed $199/mo + usage; Curator $99/mo + $0.12 CPM; blocks metered — Auction/Proof $0.05/1k, Trust $3.50/1M, Courier $0.15/GB, Settlement $0.02/1k; Discovery free). Infra is computed to the cent, so gross margin is an output. Counts, ARPU, usage, CAC, churn, expansion and sales-cycle lag are assumptions — grounded under "The evidence." ARPU is our net fee, not media pass-through. Burn is flat from month 1 (real teams hire as they grow) while the ramp assumes go-to-market capacity — read those two together. LTV is capped at 36 months of retained, margin-adjusted revenue (no 8-year fantasy). NRR = expansion less logo churn, annualised. Platform revenue lags acquisition by the sales-cycle months, so cash goes out before it comes in.

The cases are anchored to how comparable usage-based and adtech-infrastructure businesses actually raise, retain and price. Each card notes how the fact shows up in the model.

The milestone · why the goal defaults to $1.5M

The Series-A bar is a real number.

2025 median Series A went to companies at ~$1.5M ARR; competitive is $2–5M. Investors underwrite 2–3× YoY growth, NRR >110%, CAC payback <12 mo, burn multiple <1.5×.

In the model: the goal defaults to $1.5M ARR / 24 mo, and NRR is now a first-class output checked against the >110% bar.

CRV · SaaStr

The economics · usage + subscription

Hybrid metered models retain best — via expansion, not just low churn.

Subscription + usage hybrids deliver best-in-class NRR because usage grows as customers grow; $1–5M ARR startups run ~8-month median CAC payback (top quartile 5).

In the model: each segment has an expansion dial (usage growth on the base) that compounds into NRR — the reason ARR keeps rising after the book is built.

High Alpha 2025 · Dev Corporate

The wedge · why a platform pays us at all

Neutral, two-sided delivery proof — the line item incumbents don't own.

DoubleVerify and IAS verify the buy-side for advertisers and already sit in every MSA. The unserved job is neutral proof of delivery for the sell-side and between agents — exactly-once, agent-native settlement proof. Matching an incumbent's price isn't a wedge; being the thing they structurally can't be is. Verification is billed around $0.05–0.10 CPM on multi-year contracts (order-of-magnitude, from public vendor filings/coverage), which sizes a volume CTV platform at roughly $8k–$40k/mo of net fee.

In the model: platform ARPU defaults to ~$8k/mo of net fee; the story is displacement-proof because it's a new, two-sided line item, not a third verification tax.

DoubleVerify IR · IAS IR (CPM range = estimate)

The closest comp · agentic advertising infra

Scope3 shows the category is fundable — on a few big logos.

Scope3 raised $25M led by GV, grew usage revenue 12× YoY, and landed Coca-Cola, GM, Mastercard. It scaled on a small number of large, usage-based accounts — not a signup flood.

In the model: validates the shape — a handful of high-value usage accounts is the pattern VCs already fund in this exact category.

BusinessWire — Scope3 / GV

The pipeline · where the whales come from

Network-led, not cold. The 8 are mostly warm.

Enterprise CTV/SSP deals are 9–18 month, procurement-heavy sells — uninvestable as cold outreach from a 4-person team. They become investable as warm intros: the Series-A book assumes roughly 5 warm (an ex-SSP CRO's network + a GCC/ex-Kochava lead + MENA supply relationships) + 3 stretch, sourced partly through SSAI vendors and AdCP co-sell, not direct cold BD.

In the model: the platform sales-cycle lag (default ~6 mo, warm-led) reflects cash-out-before-revenue — and the ramp is a believable 1 → 3–4 → 8 cadence. Cold enterprise deals run longer and cost more — see the Downside case.

Internal — team network (CTV/SSP + GCC/MENA)

The floor · enterprise vs self-serve ARPU

Enterprise ARPU dwarfs self-serve.

Median B2B SaaS ARPU is ~$250/mo; enterprise routinely reaches $5,000+/mo. Adtech infra monetises on take-rate + per-transaction on real media volume.

In the model: builders sit at ~$30/mo as funnel + credibility; the revenue lives in managed and platform.

Culta (directional)

Caveat, stated plainly. The agentic/AdCP market is early — mostly announced, little live at scale. These comps establish the shape (few large usage accounts, high margin, network-led sales), not guaranteed counts. The model finds the mix that clears the hill; then you go get the evidence for it.