The First 100 Customers Course #14: How Endra Turned Prelaunch Trust Into $1.5M ARR in One Month

TL;DR

Endra founder Niklas Lindgren reported that the company closed $1.5 million in enterprise ARR in January 2026, the same month its first commercial pilots converted. He described the early ACV only as roughly $300,000-plus, so the often-repeated count of four or five customers is an inference, not a published customer ledger.

The mechanism was not “spend $150,000 on a website.” It was user truth → narrow workflow → prelaunch trust → targeted entry → bounded proof → competitor spillover. Endra researched MEP engineers in Reddit communities, built three or four complete workflows, translated the product into a credible category story, contacted engineers inside a small set of consultancies, and let successful pilots create urgency among competitors. Every revenue and channel figure in that sequence remains founder-reported.

What you will build

You will build a prelaunch trust system for one expensive B2B workflow. The outputs are a market adjacency map, ten user conversations, one workflow card, a three-part category story, a 25-account list, a pilot scorecard, and a competitor-proof rule. The goal is to make an unfinished product safe enough to evaluate without pretending it is finished.

Who this is for

Use this course when you sell high-value software into a concentrated vertical where practitioners share methods, firms watch one another, and the old tool can remain in place during a test. It fits AI products that automate a bounded professional workflow and can return evidence to the buyer's existing system.

Do not use it for a low-price self-serve app, a horizontal product with no narrow operator, or a workflow where one output error creates unacceptable safety or legal risk. Endra also had unusual resources: four co-founders, adjacent-domain experience, $300,000 to $400,000 of founder capital, and later venture funding. Your proof must fit your own risk and cash position.

Case snapshot

Stage	Reported fact	Evidence limit
Summer 2024	Four founders assembled around MEP design automation	Founder chronology
Spring 2025	Reddit DMs, calls, and threads used for user research	No interview count or response rate
Mid-2025	Prototype showed three or four complete workflows	Founder is uncertain whether the count was three or four
September 2025	Rejlers publicly announced an Endra pilot	Does not prove paid revenue
December 2025	$20M seed announced before the planned product launch	Funding is not customer validation by itself
January 2026	$1.5M enterprise ARR closed at roughly $300K-plus ACV	Founder-reported; exact account count unpublished
June 2026	Notion reported 600 companies from 90 countries on the waitlist	Investor-reported later traction
June 2026	$50M Series A led by Andreessen Horowitz	Independently corroborated by Axios

The model: make trust arrive before the finished product

Enterprise buyers normally ask a young vendor for three forms of proof: does the team understand the job, can the product work inside the current stack, and will the vendor survive the deployment? Endra attacked those questions in order. Its founders had run a security-systems business with an internal MEP design team. They used that adjacency to enter the market, Reddit to learn operator language, and a narrow prototype to show technical truth.

The website and product trailers then packaged that truth into a future the industry could repeat. Targeted outbound carried the story into large consultancies. Pilots supplied account-level proof. When one firm appeared able to design faster, its competitors wanted the same conversation. Trust compounded because the market was concentrated and the cost of waiting became visible.

Step 1: enter through an earned adjacency

Endra did not begin with “construction is large.” Lindgren and co-founder Anton Juric had spent eight years installing enterprise security systems. Their team produced drawings, calculations, and schematics inside the same building-design ecosystem. That gave them access to a real workflow and a reason to notice the larger MEP problem.

Build an adjacency map with four fields:

• Work observed: which job have you watched at close range?
• Data handled: which files, tools, and approvals have you touched?
• Access earned: which practitioners can correct your assumptions?
• Constraint known: what makes the obvious solution unsafe or incomplete?

Pass condition: three practitioners accept that you understand the current job. Stop condition: your market thesis depends only on size, funding activity, or an AI capability looking for a buyer.

Step 2: mine user rooms before approaching budget owners

In spring 2025, the founders used Reddit DMs, thread conversations, and calls with MEP engineers. The channel was good for users and weaker for senior decision-makers. That distinction matters. A user room gives you language, exceptions, screenshots, and workflow order. It does not automatically give you budget authority.

Run ten conversations around one job. Ask for the last completed project, not a feature wishlist: “Show me the input, every handoff, the point where work is repeated, and the file that proves it is done.” Record repeated steps and contradictory methods. Do not hide commercial intent or imitate a community member. Endra's founder described early Reddit activity as shadow marketing; copying that behavior can destroy the trust you are trying to build.

Output: one workflow map with inputs, operators, exceptions, current tools, and the final proof artifact. Pass condition: seven of ten users confirm the same expensive bottleneck. Stop condition: the apparent pain exists only in your wording.

Step 3: complete a few jobs instead of previewing a platform

Endra deliberately limited the early product to three or four workflows that worked end to end. It also integrated with Revit instead of demanding immediate replacement. The current Endra product site still describes native Revit integration, model generation, calculations, and documentation.

Choose one workflow first. Define its minimum complete loop: accepted input, transformation, review, exception path, and output returned to the existing system. “Generate a design” is too broad. “Produce this model and documentation from these inputs, with these constraints, then write the approved result back” can be tested.

Pass condition: one practitioner completes the job without a founder repairing the output off-screen. Stop condition: the demo looks impressive but the result cannot be reviewed, exported, or used in the status-quo tool.

Step 4: turn workflow truth into a category story

Lindgren said Endra spent about $150,000 on its early website, visuals, motion, story, and interface work. The site framed the current state, Endra's intervention, and the future of MEP design. It made a product that barely existed feel coherent enough for engineers to discuss and share.

The budget is not the lesson. Build the smallest story with three panels: today, the exact costly sequence; bridge, the narrow workflow you can already prove; and after, the changed operator job if that proof holds. Remove market adjectives, invented ROI, and capabilities outside the prototype.

Today, [operator] moves [input] through [tools] to produce [proof artifact]. We automate [bounded job] while keeping [approval and legacy system]. If the pilot reaches [metric], the team can change [specific operating result] without replacing [system].

Pass condition: five practitioners repeat the same before-and-after in their own words. Stop condition: they remember the visual identity but cannot say what job the product completes.

Step 5: enter 25 named accounts through the right actor

Endra targeted large consultancies in the Nordics and UK. The founders contacted engineers inside those firms and used the prelaunch story to start discussions. No public source gives the number of messages, reply rate, or exact copy, so this course cannot manufacture a funnel benchmark.

Build three separate lists: users who feel the workflow, champions who can run a pilot, and buyers who own risk and budget. Select 25 accounts only when the workflow is repeated at meaningful scale. For each account, find two dated signals, one likely operator, one likely approver, one legacy-system constraint, and one reason to reject the account.

Ask for a workflow review, not belief in the roadmap: “We mapped how [peer group] completes [job]. Your [dated signal] suggests the same handoff may exist. If the operator owns this process, I can show one bounded version and the proof criteria. If not, I will close the file.”

Pass condition: five qualified accounts introduce the operator and three accept a proof-design conversation. Stop condition: meetings come from curiosity but no one will name the current workflow, owner, or risk.

Step 6: write the pilot contract before the demo

Endra's January pilots converted quickly because the product addressed an urgent competitive job and could show ROI. A public Rejlers release had already documented a 2025 evaluation, while an AFRY announcement later described an April 2026 collaboration. Neither source proves the January revenue, but both show that enterprise testing was real.

Define the pilot on one page: named workflow, input set, output artifact, human reviewer, baseline, proof threshold, security boundary, duration, implementation owner, commercial decision date, and failure rule. A pilot should test whether the system works in context, not whether people like a presentation.

Pass condition: the buyer signs the scorecard before access begins and the output clears the threshold without hidden founder labor. Stop condition: success means only “users were excited,” or every exception becomes unpaid custom work.

Step 7: let proof travel between competitors

Lindgren said outbound started the motion, then word of mouth took over as consultancies noticed competitors testing Endra. This works only in a concentrated market where firms compete for similar contracts and the operational advantage is visible. It is not a license to invent customer names, leak confidential pilots, or manufacture fear.

After a successful proof, ask what can travel: a named reference, an anonymous metric, a workflow benchmark, a public integration, or a customer-authored explanation. Give the customer veto power. Track referred conversations separately from founder outbound. The compounding channel exists only when peers repeat the proof without the founder pushing every introduction.

Pass condition: one approved proof creates two qualified peer conversations. Stop condition: urgency disappears when the first founder stops mentioning the pilot.

What failed, what is risky, and what not to copy

• Brand spend before evidence: $150,000 was a high-risk allocation made by founders with prior capital and domain access. A polished story cannot rescue an unverified workflow.
• Reddit shadow marketing: the founder acknowledged the tactic. Hidden commercial behavior can trigger community rejection. Research transparently and contribute before asking.
• Rounded account math: $1.5M divided by roughly $300K-plus suggests four or five accounts, but no exact count is published.
• Public pilot timing: Rejlers announced a pilot in September 2025, while Lindgren called the January tests the first pilots. Treat the former as design-partner work and the latter as the commercial conversion sequence.
• Missing retention evidence: no public source provides churn, renewal, expansion, gross margin, or implementation-time data for the January cohort.

Your seven-day implementation plan

1. Day 1: write the adjacency map and reject markets with no earned access.
2. Day 2: recruit five users from one practitioner room and map the last job.
3. Day 3: complete the workflow card and define the legacy-system bridge.
4. Day 4: write the today, bridge, and after story using only proven capability.
5. Day 5: build 25 accounts and separate users, champions, and buyers.
6. Day 6: draft five account-specific asks and one pilot scorecard.
7. Day 7: review every claim, send manually, and log the reason for each reply.

The week passes when five operators validate the same workflow, three accounts accept a proof-design conversation, and one buyer edits the scorecard. It fails usefully when the job, actor, or proof threshold changes across every account. Do not respond by buying a better website.

Lead Scorer implementation

This motion fits Lead Scorer when the target vertical and workflow actors can be identified from public evidence. It does not replace community participation, product testing, security review, or the founder's approval. Keep three lists separate: users for workflow language, champions for pilot design, and buyers for risk and budget. Mixing them produces generic messages and false qualification.

Phase A: store the product truth and score before enrichment

Start with the ICP and offer context skill. Record the workflow, acceptable company size, legacy-system requirement, supported regions, disqualifiers, proof event, and claims you may use. Then run the ICP scoring rubric skill on a small backlog. Give workflow repetition and visible transformation pressure more weight than company fame.

Define an ICP for [workflow]. Reject accounts without [repeated job], [named operator], or [legacy compatibility]. Score fit from 1 to 10. Do not enrich or find contact data below 7.5.

Output: 25 scored accounts with explicit rejection reasons. Credit gate: review the keepers before any paid enrichment or contact discovery. Stop condition: more than half the score depends on company size alone.

Phase B: build evidence dossiers for the correct actors

Use the signal research dossier skill on users, champions, and buyers separately. Require two dated public signals per person or account. A hiring post for MEP engineers may support workflow scale. A platform migration may support urgency. A generic funding announcement does not prove the job exists. Skip honestly when no signal is available.

When a public LinkedIn post, event, or search exposes relevant practitioners, create an audience source with create_audience_source, then prequalify before enrichment. Do not import private community membership or imply consent that does not exist.

Output: five complete dossiers, each with actor, workflow, two signals, source links, objection risk, and one reason not to contact. Pass condition: the message can refer to an observable job, not a guessed pain.

Phase C: draft a bounded proof ask, then keep approval human

Use contact discovery only for approved keepers, then the AI-authored campaign skill to create drafts. The campaign should remain inactive. One message asks for one workflow review; it must not claim that a peer is a customer unless the reference is approved.

For each approved account, draft one message under 120 words. Use one dated signal, name the workflow hypothesis, offer a bounded proof-design call, and include one clean exit. Keep every message in draft for human review.

Run the outreach QA audit before sending. Set a practical threshold such as 85/100, then review factual claims, tone, recipient role, and ask. No agent should activate the campaign, spend unapproved credits, or send on the founder's behalf.

Phase D: turn objections into the next proof asset

Apply reply triage to classify responses as interest, timing, wrong person, objection, or no. Aggregate objections by workflow and actor. If three qualified accounts question the same integration, security boundary, or proof metric, return that issue to the product and Content Studio. Publish useful evidence only after approval, then use the resulting visible engagement as a new signal audience.

Pass condition: objections change the workflow card, pilot scorecard, or evidence asset. Stop condition: the team answers every objection with more message volume. Lead Scorer can make research, qualification, drafting, and feedback systematic. It cannot create the proof that makes competitors pay attention.

Saveable checklist

• I can name the adjacent workflow that gave us access.
• I have user evidence before a buyer pitch.
• The product completes one job end to end.
• The legacy system can remain during the proof.
• The story separates today's job, our bridge, and the possible after state.
• Users, champions, and buyers live in separate lists.
• The pilot has a baseline, threshold, reviewer, and decision date.
• No customer name or metric travels without approval.
• Referred conversations are measured separately from outbound.
• Every material claim carries a source or an uncertainty label.

Sources and limits

The main source is the August 2026 Product Market Fit Show interview with Niklas Lindgren. The $1.5M January ARR, roughly $300K-plus ACV, $150K brand spend, Reddit activity, outbound sequence, and word-of-mouth explanation come from that interview and are not independently audited.

Notion Capital corroborates the $20M December 2025 seed and reports more than 20 user interviews in its diligence. Its Series A note reports a 600-company waitlist across 90 countries and later commercial conversion. Axios independently corroborates the $50M June 2026 financing. Investor statements support timing and market interest, not an audit of January revenue.

Current product claims come from Endra's website. The current platform is broader than the January version, so no present capability is projected backward. There is no public exact customer count, January customer list, outreach activity, win rate, churn, retention, gross margin, or expansion rate. The useful asset is the sequence and its decision gates, not a cleaner success story than the evidence allows.