---
title: "Head of SignatureAPI at Superposition Labs, Inc."
company: "Superposition Labs, Inc."
position: "Head of SignatureAPI"
period: "October 2026 - Present"
contract: "Full-Time"
description: "I own the whole SignatureAPI product, from the roadmap to engineering, infrastructure, security and compliance. My main focus is growth, measured by how many customers keep signing documents every week."
website: "https://superposition.company/"
url: "https://rolandoandrade.me/career/head-of-signatureapi/"
author: "Rolando Andrade"
---

# Head of SignatureAPI at Superposition Labs, Inc.

> I own the whole SignatureAPI product, from the roadmap to engineering, infrastructure, security and compliance. My main focus is growth, measured by how many customers keep signing documents every week.

Period: October 2026 - Present. Contract: Full-Time. Company website: https://superposition.company/

# From first engineer to owner of the product

In November 2023 I joined SignatureAPI as its [first engineer](/career/software-engineer-signatureapi/). For almost three years my job was to build the platform and keep it running. In October 2026 the job changed. I became Head of SignatureAPI at Superposition Labs, the company behind it.

Now I own the whole product. That means the roadmap and the priorities, the engineering, the infrastructure, security and compliance. I make the product decisions, and I answer for the results.

The title says "Head", but the daily work is closer to a product engineer or a forward deployed engineer. I talk with customers, join their integration calls, decide what to build and build it myself. I do it with a full team of AI agents.

## Why the change happened

Superposition Labs runs several projects. SignatureAPI was one of them, and it had to share attention with the others. Decisions waited. Development slowed down. The product needed one person whose full time job was SignatureAPI.

I had built most of the platform and worked with its customers for years. So in September I started to make the product decisions in practice. In October the role became official.

The first thing I learned is that owning a product is a different job. As an engineer I asked "how do we build this?" As the owner I ask "should we build this at all, and how will we know it worked?"

## A team of agents

One person cannot cover a whole product alone. So I built a harness for AI agents: protocols, skills and rules that tell each agent how SignatureAPI works and how to do one kind of job well. A skill is a short written playbook an agent reads before it starts.

Each agent has a clear job:

1. **Support.** A support skill works through the ticket queue in Plain. It finds the customer's account from the failing request and follows our support playbook.
2. **Production debugging.** An investigation skill looks into production problems. It sends its queries to a read replica, a copy of the database, so it never touches live data.
3. **Growth.** A growth signals playbook turns account data into a list of who needs attention and why.
4. **Compliance.** One compliance skill routes every SOC 2 task, from evidence to vendor reviews, to the right procedure.
5. **Building.** Coding agents write code with me. A context router gives each one only the parts of the system it needs.

<AgentTeam caption="Each agent has one job. The work comes back to me before it reaches customers." />

Agents write fast, so review matters more, not less. Every change passes an automated change gate. The depth of the review depends on the size of the change, and release pull requests are written by a bot. I wrote about this on the SignatureAPI blog: [how we keep every change reviewed when agents write the code](https://signatureapi.com/blog/how-we-keep-every-change-reviewed-when-agents-write-the-code).

<MergedPrsChart caption="Pull requests merged each month at SignatureAPI. September 2026, my first month owning the product, reached 103, far above the 17 to 29 of a normal month a year earlier." />

My job moved from writing every line to deciding what to build, giving agents good context, and checking what they ship.

## Retention is the output

I wanted a method, not a list of opinions. I studied the growth programs from Reforge. One idea from them became the base of everything:

**Retention = activation + engagement**

Retention means customers who keep using the product. It is the result you want, but you cannot change it directly. You change the things that create it:

1. **Activation.** A new customer reaches real value for the first time.
2. **Engagement.** That customer uses the product often and deeply enough to keep it.

<RetentionFormula caption="You cannot move retention directly. You move the two inputs that create it." />

Customers who leave can also come back, so bringing them back is part of the work too. But activation and engagement are where most of my time goes.

## One number to follow

A retention metric is only useful if it measures real value. Logins do not measure value. Sign ups do not either.

For SignatureAPI, real value is a signed document. So our retention metric is **LCE3/4**: a customer completed at least one live envelope in 3 of the last 4 weeks. An envelope is one package of documents sent out for signature. "Live" means production mode, with real documents and real signers, not test mode.

<WeekStreak caption="One empty week is fine. A customer who completes live envelopes in 3 of the last 4 weeks counts as retained." />

If a customer is in LCE3/4, signing has become part of how their business works every week. That is the habit we want to grow.

## Two kinds of customers

Before changing anything, I studied who our retained customers really are. I used the Customer Retention Canvas, a tool from Reforge. It describes each use case by five things: the problem, the persona, why people choose you, their alternative, and how often they need you.

Two use cases came out. Together they bring 94% of our monthly revenue.

The first is **signatures inside another product**. A developer or product lead works at a platform in fintech, SaaS, real estate, HR or health. They want their users to sign inside their own app, without emails from a third party. These customers sign every day.

The second is **a company sending its own agreements**. The owner or operations person of a small business wants contracts made from their own Word template and signed without manual work. These customers sign every week.

The canvas also showed who is not our customer. About 26% of sign ups come from people who only signed a document once and were curious. Almost none of them pay. Building for them would pull the product in the wrong direction.

## Frequency decides retention

One finding changed how I see churn. Customers who send 10 or more live envelopes a month stay in LCE3/4 about 91% of the time. Customers who send 1 to 9 stay only 27% of the time.

<FrequencyBars caption="The same kind of customer, doing the same job. The ones who sign often almost always stay." />

It is the same person doing the same job, just less often. When signing is rare, people forget about the product. And for the second use case, the biggest competitor is not DocuSign. It is doing nothing at all. About two thirds of the customers who cancelled told us they were simply not using it. Very few left for another service.

## Where new customers get lost

Then I worked backwards from LCE3/4 to find the steps a new customer goes through. I call it the activation chain:

1. **Setup.** They send an envelope with their own document, from their own code.
2. **Aha.** They come back on another day and run their own document again in test mode. Their flow works from start to end.
3. **Purchase.** They pay and move to production.
4. **First real value.** A real document leaves their system and comes back signed by someone else.
5. **Habit.** They complete live envelopes in 3 of their first 4 weeks.

The chain showed where the problem was. Only 27% of new accounts reach setup. Only 32% of those reach the aha moment. Almost all the loss happens in those first two steps. Someone who skips setup almost never reaches the aha moment (0.9%).

<ActivationChain caption="Almost all the loss happens before the aha moment, in the first two links of the chain." />

Customers also told us what the aha moment feels like, in their own words: "I tried it and it works." "I finished the integration." "I had what I needed in 3 hours." And when we lost people near the end, it was often at payment. They could not find where to pay, a payment link had expired, or the price was not clear.

## What I changed first

With the chain in hand, the work became clear. Every change below targets one step.

**A guided first envelope.** New accounts now get a short guide to their first envelope. The signing ceremony (the screen where people sign) is embedded right in the dashboard. The test API key is right there on the first screen, so nobody has to search for it.

<FlowSteps caption="The guided first envelope. Every stage after the welcome can be skipped." />

I wrote the full story, with the data from the first week, in [what a week of onboarding data taught me](/blog/time-to-first-value-developer-onboarding/).

**Nudges at the right time.** Someone signs up but sends nothing. They get an email on day 1 and another on day 3. Each email has a one click unsubscribe. Every step of onboarding now records an event, so we can see exactly where people stop.

**The first API call.** The share of new customers who make their first API call went from 42% to 70%.

**Unpaid accounts.** I built automatic handling for subscriptions that stop being paid. There is now a grace period with clear notices. Several accounts came back to good standing, and others were deactivated. When a subscription is cancelled, the account moves back to a trial. The details are in my post on [involuntary churn and the subscription grace period](/blog/involuntary-churn-subscription-grace-period/).

**Knowing where customers come from.** We now store the first source of every sign up and its country. With it, we can compare channels by the customers who stay, not only by the sign ups they bring.

**Signals for engagement.** A daily lifecycle job now looks at every account. It emails owners when something needs their attention, and it records growth signals, like an account sending less than before. It respects a frequency cap, so nobody gets too many emails.

The early numbers are moving in the right direction, but only slightly. It is too soon for conclusions. We are still building a baseline we fully trust, and I would rather wait than celebrate the wrong number.

## Agents as a channel

More and more developers build with coding agents like Claude Code, Codex and Cursor. If the agent does not know SignatureAPI, the developer will not find it. So agents became an acquisition channel.

MCP is a standard way for AI assistants to use outside tools. In my first weeks in the role, we turned our MCP server on for every account. It now has 24 tools, and it is listed in the official MCP Registry. We prepared plugins for ChatGPT, Claude and Codex.

I also built the SignatureAPI CLI, a command line tool that developers and agents can both use. It can set up a project for 11 different coding agents and forward webhooks to a local machine. We published agent skills that teach agents how to integrate SignatureAPI correctly. We also made our docs easy for agents to discover, and built demos for embedding signing in native mobile apps with React Native.

To know if any of this works, I built a small harness that asks coding agents about electronic signatures and checks whether they mention SignatureAPI.

## Running the whole product

Growth is my focus, but the rest of the product does not stop.

**Qualified signatures.** In Europe, a qualified electronic signature (QES) under the eIDAS rules has the same legal weight as a handwritten one. It must come from a certified provider. We released the Signature Worker protocol, which lets customers connect those providers through the Cloud Signature Consortium (CSC) standard. It includes a hosted sandbox, a self service connection, and qualified timestamps. QES is in private beta today.

**SOC 2 Type 2.** A SOC 2 Type 2 audit checks that security controls actually worked over a full period, not only on one day. I am leading the audit for the October 2025 to October 2026 period. I handle the evidence requests and the follow ups with the auditor. The report is still pending. To make evidence easier, I added a monthly change report and audit logs that cannot be changed without anyone noticing. The change gate from the agents section also counts as evidence.

## What I am learning

As an engineer, I measured my work by what I shipped. Now I measure it by what customers do after I ship it. That changes which work I choose.

I also learned that the most useful work is often not a feature. A metric we can trust and a clear picture of our customers came before any change. Without them, I could be building for the 26% of sign ups who will never pay.

The goal for the next months is simple to say and hard to do: more customers who sign documents with us every week.
