The AppLovin playbook, for native & DTC  ·  built by operators

Don't become an ad network.
Own the layer above every one.

A neutral engine that measures and optimizes to realized retained profit across Taboola, Outbrain, Meta and the rest — the one thing none of them can build, because each only sees its own traffic and only the front-end conversion.

"

That is the actual AppLovin analogy. AppLovin never won by buying publishers. It won by owning a proprietary, self-improving signal no competitor had — and renting an engine on top of it. Our equivalent signal is realized downstream profit, measured the same way across every channel.

AppLovin = AXON (the buyer) + MAX (the signal)  →  Ours = the cross-channel buyer + the realized-profit signal.

~80%+
Software gross margins — not arbitrage spread
30–50
Anchor DTC advertisers we can light up from our network
~10
Of our own brands as live training data — customer zero
$10B
The target: a tech multiple, not a 5–8× arbitrage cap
00 · The shape of the company

One platform. Two faces. A single compounding loop.

Face B makes Face A smarter than anyone's buying. Face A generates the data that makes Face B impossible to copy. That loop is the company.

Face A — The Buyer

Cross-channel performance engine

Programmatically buys native + paid-social + open-web for DTC advertisers and bids to realized retained profit — rebills, refunds, chargebacks netted out — instead of front-end conversions. Revenue and spend volume live here.

Face B — The Truth Layer

Neutral incrementality + retained-ROAS

Proves true incremental profit per channel via continuous geo-holdouts and real backend truth (Checkout Champ / Shopify / Stripe). This is the moat — and the multiple.

AppLovin = AXON (buyer) + MAX (signal)   |   Ours = the buyer + the realized-profit signal

Why capital + your network flips the earlier verdict

The two reasons the honest red-team killed the bootstrapped version — both were constraints of being broke and alone. With money and the network, they invert.

The two killers (bootstrapped)

  • Data won't generalize past our own vertical — a model trained on our rebillers can't predict another category.
  • Channel conflict — rival advertisers will never pool churn data with a competing operator, so the network can't form.

How money + network invert them

  • Breadth, day one — onboard 30–50 advertisers across verticals (subsidized). The methodology generalizes; the cross-vertical benchmark dataset becomes the asset.
  • We're neutral infra, not a network — we make their existing spend more profitable. No conflict. Your relationships land the anchor logos that pull everyone in.
01 · The product, concretely

Four engines that feed each other.

No tool today combines true cross-channel incrementality + real backend retained-LTV + closed-loop automated bidding. Dashboards exist. Experiment tools exist. Nobody closes the loop.

Data Spine · build #1

One schema per advertiser: creative → spend → click → session → checkout → rebill 2/3 → refund → chargeback → LTV. Network APIs + analytics + backend via server-side CAPI.

Measurement Engine

Continuous geo-holdouts yield incremental retained-profit-per-dollar by channel, campaign, audience, creative. A "True ROAS" advertisers trust — because it matches their bank deposits.

Prediction + Bidding · the AXON

Per-opportunity model predicting realized retained LTV — not p(click). Auto-bids and reallocates budget across every network via their existing Target-ROAS APIs. We ride their auctions; we don't build supply.

Creative Engine

We productize our own marketing edge: generative + performance-tested creative tied to the same realized-profit loop, so the system learns which angles drive retained customers. Engineers who never bought a click can't fake this.

02 · Why we win

Unfair advantages, ranked.

1

The network

We can light up 30–50 anchor DTC advertisers fast. Cold-start solved. Benchmark breadth solved. Logos that pull the rest in.

2

Customer zero

Our ~10 brands give live, full-funnel, realized-profit data to train on before a single external customer signs.

3

Operators, not vendors

We buy media and live or die on ROAS. The product is built by the people who need it to work — it out-performs tools built by engineers who never bought a click.

4

Capital

We subsidize onboarding to acquire data, out-hire on ML, and move faster than any bootstrapped competitor.

5

Neutrality

No channel conflict. Everyone shares with the layer that makes their existing spend more profitable.

The combination

Networks have data but no neutrality. Tools have neutrality but no closed loop. Agencies have neither data science nor scale. We have all three.

03 · The flywheel

Why it compounds into a tech multiple.

DATA FLYWHEEL Moreadvertisers Richerdata Betterbids BetterROAS Sharpermodels
  1. More advertisers join the neutral layer (your network seeds it).
  2. More cross-channel realized-profit data flows into one schema.
  3. Sharper benchmarks & LTV predictions than any single-network optimizer.
  4. Our bidding beats in-platform optimizers — on profit, not clicks.
  5. Advertisers move more budget through us; results attract the next cohort.
  6. The dataset is the moat: networks see only their own traffic, attribution tools have no closed loop, agencies have no data science. No incumbent can assemble it.
04 · The plan

Four phases, each gated on proof.

We raise against milestones, not vibes. Every round is priced on demonstrated lift and spend-managed growth — so the multiple narrative stays clean.

PHASE 0
0–4 months

Prove it on ourselves

Build the spine + measurement on our ~10 brands. Run holdouts. Produce a defensible True ROAS vs. network self-report.

Success gate≥15–25% retained-profit lift reallocating our own spend. This is what we sell and raise on.
PHASE 1
4–9 months

Design partners

8–12 anchor advertisers from your network. Subsidized, white-glove — our marketers run it for them.

Success gate8+ advertisers with holdout-proven lift, signed case studies, verbal expansion intent.
PHASE 2
9–18 months

Productize + scale

Self-serve + managed tiers, onboarding automation, hardened channel integrations. Land 50–150 advertisers.

Success gate$X00M+ annualized budget flowing through the platform, net spend retention >120%, software-trending margins.
PHASE 3
18–36 months

Category dominance

Become the default profit layer for DTC. Expand channels, geos, verticals. Selective M&A of point tools — data + talent, not supply.

Success gate$1B+ spend managed, clear data-moat lead, tech multiple.
05 · Team & capital

Who we hire, and what we raise against.

The #1 hiredo this first
ML / bidding lead with real ad-tech pedigree — ex-AppLovin / Meta / Google / The Trade Desk / Moloco. The single most important person in the company.
Core build team
Data engineering (spine + network integrations) · applied ML (LTV + incrementality models) · measurement-science lead (causal inference / geo-experiments).
Go-to-market
A small, elite sales team selling into your network · a CS / managed pod staffed by our own marketers — the differentiator competitors can't hire for.
The base
The brand holdco keeps running as the cash engine and live lab — it funds Phase 0 and trains the models.

Milestone-gated capital ladder

Money is available — so we deploy it against proof, not ahead of it.

Internal
PHASE 0

Funded from Acentecom cash. No dilution. Build the proof.

$15–30M
SERIES A

On the Phase 0 proof: platform + bidder, design partners, key ML hires.

$75–150M
SERIES B

On Phase 1/2 traction: scale GTM, channels, geos, subsidize the network land-grab.

$300M+
GROWTH

At Phase 3: category capture + selective M&A.

06 · The model & the path to $10B

The AppLovin model — you fund a budget, we buy the impressions.

Exactly like AppLovin: an advertiser funds a budget, our engine spends it buying impressions across native, paid-social and the open web, and they see the CPM each creative set paid. We sit in the flow of funds — we buy media, our realized-profit engine wins the cheap, high-retention impressions, and the spread between what we pay for media and the budget they fund is our revenue. That's a principal ad-platform model with ~75%+ software-like gross margin — not an agency cut of someone else's spend.

1 · Fund a budget. Advertiser deposits spend into the platform — one balance, all channels.

2 · We buy impressions per creative set. The engine bids to realized retained profit and reports the CPM each creative set actually paid.

3 · We keep the spread. Win impressions below the value we deliver → the gap is high-margin platform revenue that compounds as the model sharpens.

$300–650M

of gross profit (the spread) is roughly what an ad-tech multiple needs to clear $10B — which means single-digit $billions of budget flowing through the platform at AppLovin-class margins, with >120% net spend retention as each advertiser scales. A realistic ceiling for the default DTC profit layer, given the size of native + paid-social DTC budgets.

07 · Risks & how we neutralize them

The honest objections — and the answers.

Networks restrict API access as we scale
We're demand they can't afford to lose. Diversify across many channels so no single API is fatal — and the measurement layer works even if one closes.
Incrementality claims get challenged
(cf. AppLovin's short-seller fight.) Holdout-validated, advertiser-auditable methodology is the brand — rigor is the product, not a liability.
Data won't generalize across verticals
Solved by breadth (the network) + methodology transfer. Fall back to measurement-only where prediction is thin; monitor per-vertical model quality.
Incumbents add backend data
Our lead is the closed loop (measure + bid) + the cross-channel realized dataset + operator DNA. Keep widening it with capital and speed.
Privacy & regulation
Server-side, consent-based, first-party by design. This is a tailwind against pixel decay — not a threat.
08 · The next 90 days

Concrete moves, starting now.

Stand up the data spine on our own brands — one unified schema across every channel + the Checkout Champ backend.

Run the first continuous geo-holdout on our biggest brand; produce the first True-ROAS vs. network self-report.

Make the #1 hire — the ML / bidding lead.

Pick the 8–12 design-partner advertisers from your network; draft the subsidized offer.

Package the Phase 0 proof into the Series A narrative.

Open items to decide with Uzi

  1. Run both faces (buyer + measurement) as one platform, or sequence measurement-first?
  2. Which 8–12 advertisers in the network are the anchor design partners?
  3. Do we run geo-holdout incrementality testing anywhere today, or is it net-new?
  4. Green-light Phase 0 build + the ML/bidding lead search?