The First 100 Customers Course #37: Merge Spoke With 100 Companies Before Its First 10 Customers
A sourced course on how Merge used roughly 100 buyer conversations, a readiness queue and hands-on pilots to find its first customers, with separate gates for interviews, deployments and payment.
By Lead Scorer 19 min read The First 100 Customers Course · Lesson #37
TL;DR
Merge spoke with roughly 100 prospective companies before building its unified API. Cofounder Gil Feig says about 10–15 of those companies were ready when the product existed, and the SaaS Club interview describes the first ten customers. That is a remarkable early pipeline, but it is not a verified 10% paid conversion rate. Cofounder Shensi Ding later explained that the first three customers received about two free months. Keep conversations, pilots, deployments and payment in separate columns.
The useful mechanism is discovery as a named buyer queue. Merge did not treat a research call as a completed task. It selected companies with a painful integration problem, learned who owned it, kept track of when they could adopt, built toward the earliest useful deployment, and stayed close enough to support it. This course turns that sequence into a ten-account exercise for a B2B founder without a large audience. The exercise thresholds are proposals for your sprint, not historical Merge benchmarks.
What you will build
You will make a discovery-to-deployment register, not a list of people who said they liked an idea. Choose one buyer group and ten companies. For each, record a real operating problem, the person who can approve a pilot, the system or workflow involved, a dated next step, an adoption constraint, and a payment state. The finished output is one of three decisions: continue the wedge, revise it, or stop. You do not need 100 calls to run the first test.
This fits a B2B product whose buyer can explain a costly, repeated workflow and whose first useful deployment can be scoped. It fits poorly when the buyer group is vague, the value is entirely speculative, or you cannot safely support the first customers. Merge had two founders who had lived the integration problem at prior companies. That domain access was part of its advantage; a spreadsheet alone cannot replace it.
The evidence snapshot
| Stage | What the sources support | What they do not prove |
|---|---|---|
| Problem | Both founders had seen customer-facing integrations slow sales and engineering at previous employers. | That every B2B company had the same urgency. |
| Discovery | About 100 prospective-company conversations, mainly with product and engineering leaders. | That 97 were committed buyers. |
| Readiness | Gil recalls roughly 10–15 companies ready when a product was available. | A measured 10–15% paid conversion rate. |
| First use | Shensi says Drata and two others got about two months free while Merge proved reliability. | Ten paid contracts on launch day. |
| Later paid base | Gil reported about 40 paying customers in a July 2021 interview. | That all 100 organizations reported on the platform in May were paying. |
The first four rows come from the Gil Feig interview and a later interview with Shensi Ding. The paid-base distinction comes from TechCrunch's May 2021 report and Gil's July 2021 interview. The first report says about 100 organizations were on the platform; the second reports about 40 paying customers. Different dates and definitions prevent a clean conversion calculation.
The model: a conversation must leave a trace
Merge's founders saw the same type of integration burden in two different employers. Gil had watched a recruiting product lose opportunities because it could not support every applicant tracking system. Shensi had seen ticketing integration work at a security company. Their initial question was therefore narrow: would B2B software teams pay to stop rebuilding and maintaining customer-facing integrations? Shensi's First Round account confirms that the two spent months talking to companies before committing full time. They needed a solution and a buyer, not applause for a concept.
The causal chain was lived pain → matched buyer → discovery record → readiness date → first deployment → paid decision. Each arrow can break. A prospect can confirm the pain but not control a budget. A buyer can approve a pilot but need a platform integration you cannot deliver. A deployment can work but not earn enough trust to be paid. The register below makes those breaks visible before you mistake a full calendar for traction.
Step 1 — Pick one expensive repeated task
Write a problem statement with an actor, a current workaround and a cost. Merge's early buyer was a B2B software team building embedded integrations for its own customers. The repeated cost was not only writing an API connector. Gil describes the later work: real customer data exposes edge cases, engineers return to old integrations, and support has to explain failures. That is why an internal engineer who could build one connector was not automatically a buyer for Merge.
Your worksheet: “For [buyer role] at [company type], [workflow] repeatedly interrupts [valuable work] because [specific failure]. Today they use [workaround]. A successful first deployment would change [observable event].” Write down the evidence for every bracket. A named complaint or an observed workflow counts. A founder's belief about what buyers should want does not. If you cannot name the actor and workaround after ten relevant conversations, narrow the segment before writing more code.
Step 2 — Build a company list around the task
Gil says he looked for product and engineering leaders at companies that plausibly needed Merge, then asked mutual contacts for introductions. On some days the founders spoke with about five companies. He was describing a discovery motion, not saying every first customer came from a warm introduction. Shensi gives a different, specific acquisition story: she says she sent a cold LinkedIn message to Drata's CEO and was introduced to its CTO. Keep both paths in the model. Source of interview and source of eventual deal are separate fields.
Start with ten companies whose public product, job ads or documentation show the task. Record why each belongs. Split the list into warm-route and direct-route candidates, but keep one problem definition. Ask a mutual contact for an introduction when a legitimate relationship exists. Otherwise write a short direct note about the observed problem and ask for a research conversation. Do not claim a connection, a data access right or an outcome you do not have. Your first pass condition is ten reasoned accounts and one plausible buyer per account. An unverified list of 100 names fails the gate.
Step 3 — Turn interviews into a dated queue
A good interview reveals both pain and timing. Gil recalls prospective buyers saying they had whole teams doing the work and might pay substantial sums if the founders could deliver. He also says many interested companies had to put a new integration on their roadmap. That is the gap between “we need this” and “we can start.” Roughly 10–15 of the approximately 100 discovery companies were ready when Merge had a product, according to Gil's estimate. The sources do not supply a prospect-by-prospect ledger, so this is a directional founder account.
After each call, record five observations: the exact task, the economic or operational impact, who owns the decision, what must be true before a pilot, and the next date. Ask “what happens if this remains unsolved for another quarter?” and “what will you stop using or building if our test works?” End by asking permission for a specific follow-up. Put a company in the ready queue only when it has a named owner, a scoped test and a date. Keep “interested, no timeline” in a separate column. If most calls end without an agreed next action, revise the problem or buyer hypothesis rather than sending generic follow-ups.
Step 4 — Build the smallest honest deployment
Merge spent about six months building. That duration reflects an infrastructure product with sensitive HR and payroll data; it is not a universal MVP schedule. The founders needed enough integrations and security proof to make a first customer's workflow possible. Gil says they pursued SOC 2 before customer launch. A contemporaneous TechCrunch report confirms a security emphasis and SOC 2 status by May 2021, while the exact early certification timing remains founder reported. The first customers then exposed bugs and support needs that a demo account did not.
Choose one ready company and write a pilot agreement in plain language: supported workflow, data access, owner on each side, expected result, response time and a decision date. If the product handles sensitive data, security review is part of the first useful version, not a marketing badge to add later. Use a sandbox or non-sensitive test data when possible. Do not promise “production ready” to win a call. The pass condition is an observed task completed for a real user under agreed conditions. A polished demo without a supported first deployment remains an experiment, not an activated customer.
Step 5 — Separate proof, payment and expansion
This is where the edited story becomes misleading if read quickly. Gil says Merge charged from day one but also describes short free betas. Shensi says Drata and two other early customers received about two months free so the team could prove reliability, and that Drata's first product feedback changed a visible onboarding component. Those statements can coexist if “charged from day one” describes the commercial model rather than cash from each first account. We cannot verify every contract or exact payment date. For this course, the lesson is to log the distinction rather than choose the tidier version.
Put an explicit conversion meeting on every pilot calendar. Show the agreed result, unresolved defects, operating cost and what a paid agreement would cover. Record four dates: first contact, pilot start, first useful outcome and first payment. Do not call a free account “paid,” and do not hide a pilot that did not convert. If the buyer cannot name the value received by the decision date, fix the deployment or stop; an automatic free extension rarely repairs a weak problem. Only then ask whether another team or company has the same need.
What failed, and what is easy to copy badly
Three companies reportedly told Merge the idea would not work. Gil says they signed later, after seeing competitors offer many integrations through Merge. That is a later credibility event, not evidence that every objection should be ignored. Other prospects insisted they could build integrations themselves. Merge sometimes let them try and return after discovering the maintenance burden. This is safe only when you can afford the wait and can point to a concrete cost, not when you need to win every deal immediately.
The two founders' accounts also warn against copying a single acquisition channel. Warm introductions helped book discovery calls; Shensi's cold message reached the first named customer. The relevant constant was a precise buyer problem. A broad “networking” campaign would miss it. So would a huge cold list sent before the founder can explain the current workaround. Finally, the 50,000 engineers mentioned in a later Shensi interview were recruiting prospects. They were not 50,000 customer leads and do not belong in the acquisition funnel.
Your seven-day discovery-to-deployment sprint
- Day 1: define the wedge. Write the actor, task, workaround and failure event. Collect three public examples that show the task exists. Output: one problem statement and an explicit exclusion rule.
- Day 2: select ten accounts. Name a likely owner and the evidence for fit. Split warm introductions from direct messages. Output: a ten-row account register; stop if fewer than six have a grounded reason to be there.
- Day 3: prepare five conversations. Ask about a recent incident, who fixed it, cost, timing and approval. Output: interview guide and five reviewed invitations. No invented familiarity.
- Day 4: conduct or book interviews. Record the buyer's words as notes, not a confidence score alone. Output: problem, workaround, owner and next date for each completed call.
- Day 5: sort the queue. Mark each company as ready, timed later, unsuitable or unclear. Output: one candidate pilot brief with observable value and known constraints.
- Day 6: scope the first useful test. Specify environment, data, support owner, security check and acceptance condition. Output: a pilot plan; stop if you cannot support the promised workflow.
- Day 7: make the decision. Compare actual conversations with the starting hypothesis. Output: continue, narrow, or stop, plus the next ten-account batch only if the first batch taught you something specific.
A small sprint does not need to hit Merge's reported volume. Its job is to make the next unit of work obvious. Record denominators: accounts researched, invitations reviewed, conversations held, dated follow-ups, pilots started, useful outcomes and paid agreements. Never divide an eventual paid count by people who merely viewed a landing page and call it an interview conversion rate.
Run the research and draft loop with Lead Scorer
This module fits a narrow B2B discovery motion when there is a clear buyer role and public fit signals. It does not fit a mass self-serve launch or a product whose buyer and use case are still unknown. Connect the Lead Scorer MCP and use its ICP, discovery, scoring, enrichment and campaign-draft skills as an execution surface. A founder must still inspect sources, conduct interviews, approve credit use and review every outbound message. The product does not supply Merge's domain knowledge or guarantee ten customers.
Phase 1: define the product and buyer. Invoke the ICP and offer-context skill with your one-sentence wedge, current product state, geography, buyer role, exclusion rules and only proof you can show. Copyable prompt: “For this product and task, define a narrow buyer group, explicit disqualifiers and a ten-account research hypothesis. Label every assumption.” Keep the discovery cohort in its own list; do not mix it with newsletter subscribers, current customers or a campaign-ready list. Pass when another person can say why each account belongs.
Phase 2: find and score before spending credits. Use the daily vertical prospecting or sales-navigator lookalikes skill to discover public company and person candidates. Use the ICP scoring rubric to score them before contact enrichment. Read the evidence behind each score: role, current workflow, relevant product and dated signal. A useful first gate is to review the top ten and reject anyone whose fit depends only on a job title. Set a credit budget before invoking enrichment or contact discovery; enrich only the keepers above your chosen threshold, and confirm any spend that crosses the product's approval gate. Output: a scored research list with source URLs and a separate approved enrichment list.
Phase 3: draft one respectful ask. Invoke signal research for each keeper, then the first-touch and outreach QA skills to produce a short note about a verified task. A warm introduction request and a direct message need different copy. Copyable prompt: “Draft a first message for this person using only these verified signals. Ask for a short research conversation about the named workflow. Do not claim we can already solve it.” Create a multichannel campaign only as a draft if the motion warrants it. Review the recipient, evidence, promise and timing of each message. A draft is not permission to activate a campaign or send anything.
Phase 4: turn replies into the queue. Classify each actual response as an interested buyer, later timing, wrong person, objection or refusal. Record the next action and date in the discovery register. An objection such as “we can build it ourselves” may reveal a missing cost explanation, while a security objection may reveal a real product requirement. Feed repeated questions into Content Studio as sourced notes and useful draft content, rather than treating a reply as permission for more outreach. If a public post attracts an engaged audience, capture only the visible people and score them before adding them to a new signal list. Keep that audience separate from the original ten accounts.
Stop the agent workflow when evidence is thin, the buyer asks not to be contacted, the credit budget needs approval, or a campaign needs human review. Your success measure for this phase is a better buyer queue with real next steps. The paid-customer count stays zero until a contract and payment actually exist.
Saveable checklist
- One buyer role, one repeated costly task, one observable first outcome.
- Ten accounts chosen for evidence of that task, with warm and direct routes marked.
- Every interview logged with owner, constraint, next date and source of contact.
- Ready buyers separated from interested-but-later and unsuitable companies.
- One supported pilot with security, data, support and payment decision gates.
- Separate counts for conversations, pilots, active deployments and paid accounts.
Sources and limits
The primary source is SaaS Club's full interview with Gil Feig. Cross-checks are Shensi Ding's First Round retrospective, her later podcast interview, TechCrunch's 2021 reporting and Gil's 2021 paid-customer interview. Founder interviews support the sequence and approximate counts; they do not provide a raw CRM, billing export or controlled channel test. The podcast's summary calls ten initial customers paying, but the cofounder says the first three received free months. Accordingly this course does not claim ten paid contracts at launch, a 97% positive-interview rate, or a universal conversion benchmark. Its ten-account plan is an editorial exercise designed to test the mechanism honestly.
Frequently asked questions
Did Merge get ten paying customers from 100 interviews?
The SaaS Club host describes ten early customers, and Gil Feig estimates that 10–15 discovery companies were ready to start. Cofounder Shensi Ding says the first three received about two months free. The sources do not prove ten paid accounts at launch.
Were Merge's first customers all warm introductions?
No. Gil describes using mutual connections to book many discovery calls. Shensi says she reached Drata, the first customer, through a cold LinkedIn message. These are different stages of the early funnel.
What should a founder measure after customer interviews?
Keep separate counts for relevant conversations, agreed next steps, time-bound pilot commitments, active deployments and paid contracts. An enthusiastic interview is not a purchase.
Can Lead Scorer reproduce Merge's result?
No tool can reproduce a founder's product, trust or deal outcomes. Lead Scorer can help research a narrow buyer group, qualify contacts, enrich keepers and prepare reviewed outreach drafts while the founder owns interviews and approvals.