Perch · Role Models Internal · evidence layer · source-checked

How the others got their first hundred.

The companies whose playbook we're running, checked against the record instead of the press. What the design-partner motion actually looked like, and the two decisions it forces on us.

Start here

This is the proof behind the Perch Partner Program. Every claim below survived a three-vote fact-check; numbers that are reported-but-unconfirmed are marked est, and four widely-repeated claims we couldn't verify were thrown out. It's the evidence under the Growth Thesis, not a replacement for it.

A

The premise, corrected

Yes, $100M in ~24 months is real. The companies that hit it that fast were pure PLG. We run both engines. The partner program ignites the first hundred, which is Harvey-paced. Then our own PLG engine, the workbench where plans and specs flow plus the bid-surface loop, is what compounds. We get the trust of partner-led and the scale of product-led.

0 → $100M ARR: the two engines

Pure PLG is the fast road. We run partner-led to start, then PLG to scale.

The first hundred comes partner by partner, Harvey-paced. The workbench and the bid-loop are what bend the curve after.

CursorPure PLG · zero marketing spend~20 mo
ClayCommunity + templates · no design partners~2 yr est
HarveyDesign partner + land-and-expand · our ignition~3 yr

→ Harvey: ~$10M ARR end '23 · ~$50M '24 · ~$100M by '25. The partner start gets there in three years. Our PLG engine is the accelerant after.

B

The design-partner exemplars

Four companies that ran our exact motion. Read them as one composite: land one prestige logo, embed an engineer, time-box it, end in paid, expand inside.

Harvey

Legal AI · founded 2022
Proves

A single marquee design partner can become instant scale: Allen & Overy rolled Harvey out to 3,500 attorneys in early 2023. One logo, immediate land-and-expand.

The move

Land the most prestigious name in the category first, then grow seats and usage inside it. Usage-driven, not seat-by-seat selling.

Numbers

235 → 337 legal clients by Apr 2025. ARR ~$10M end '23 est → ~$50M '24 est~$100M by 2025. CNBC Aug '25 · Contrary

For Perch

Our ladder is the same shape: Built Exteriors → Zwick → a marquee GC. Land the prestige logo, expand inside it.

Sierra

Customer-service AI · 2023
Proves

The program structure works as a sales motion. Sierra targeted 4 design partners, secured 6, picked from a shortlist of 40 candidates.

The terms

Paid pilots (not free), time-boxed, and 100% converted to paying customers. The commitment filtered out tourists.

For Perch

This is the PPP: a small hand-picked cohort, real commitment, ends in a paid contract. And it's direct evidence that paid pilots work when the value is real. First Round Review

Legora (formerly Leya)

Legal AI · Europe · 2023
Proves

Deliberately partnered with elite firms, and ran a tight, founder-led co-build loop with them. Founders in the room, not a sales team.

For Perch

Prestige partners + founders embedded in the build cadence. The same loop the Growth Thesis describes: Brock in the room, not a rep. Rintala

Palantir

The forward-deployed-engineer origin
Proves

Embedding engineers inside the customer to build in their environment is a proven, 20-year-old model. Palantir had more forward-deployed engineers than regular engineers until ~2016.

For Perch

"One FDE per partner" isn't novel or risky. It's the proven co-build mechanic. The thing that makes co-development actually work. fde.academy

C

The loop, already run in our industry

We don't have to imagine the bid-surface loop. It was built once before, in construction, and it sold for nine figures.

BuildingConnected

GC↔sub bid network · 2012 → Autodesk $275M
Proves

A GC-to-sub bid network compounds. Started completely free, spent two years getting 1,000 active GCs before charging a cent, then grew to 1M+ subcontractors and 700k+ users.

How it grew

GCs seeded the network with bids; the sub side stayed free and crowd-sourced. The product was the channel.

Exit

Acquired by Autodesk for $275M, Jan 2019. Autodesk PR · Bricks & Bytes

For Perch

This is our bid-surface loop, proven in our exact industry: GC-seeded, sub-side free, monetize the GC, let the network pull in the next account.

D

The two decisions this forces

Decision 1 · Free vs. at-cost pilots

The advice splits. Amplify Partners says start a 3–5 partner program (✓ our number) but make pilots free by default. a16z says sell the services/implementation at cost, not for margin (✓ our ~80%-off, FDE-at-cost plan). Sierra ran paid pilots and converted 100%.

Paid-but-at-cost is the defensible middle: it filters tourists like Sierra's paid pilots, while covering only our costs like a16z prescribes. The one thing all three agree on: don't run it for profit. Decided: at-cost. Partners pay our hard costs only (FDE + tokens), no margin.

Decision 2 · The honest timeline

The design-partner road is ~3 years to $100M (Harvey), not the ~24 months the PLG rockets hit. If the goal is a 24-month number, the partner motion is the foundation, not the whole engine. The bid-surface loop is what has to bend the curve after the first hundred.

E

What we copy / what we avoid

+

Land one prestige logo first.

Harvey rode a single marquee name to credibility. Pick the partner whose logo opens the next ten doors.

+

Small cohort, real commitment, ends paid.

Sierra: 6 partners, time-boxed, paid, 100% converted. The ask is the filter.

+

Founder in the room; engineer in the building.

Legora's founder-led loop + Palantir's embedded FDE. That's what makes co-build real, not a survey.

+

Free the side that spreads you; charge the side with budget.

BuildingConnected seeded free, monetized the GC, hit a million subs. The product is the channel.

+

Run both engines, in order.

The partner program earns the first hundred. The workbench and the bid-loop, our PLG, compound from there.

+

Be honest about the clock.

Pure PLG hit $100M in ~24 months. Our partner-led start is slower; the PLG engine is the accelerant, not the opening move.

Sources: CNBC, Contrary, First Round Review, a16z, Amplify, Autodesk, Construction Dive, SaaStr, fde.academy, Rintala, Bricks & Bytes, nrich. Cut on review as unverifiable: Harvey "70-80% daily usage", Sierra "$165M ARR".