Introducing GPT Explorer

Today Cathy, Zain, and I open sourced GPT Explorer. Explorer is a power tool for GPT-3 experimentation with full historysharing, and the community’s best-practices built-in. If you’re just getting started with GPT-3 or don’t want to build out your own boilerplate codebase, try the hosted version: Explorer. The source is available here.

Screenshot of dashboard

A UI for experimentation

To get good results from GPT-3, you need to experiment. (Despite what the cherry-picked examples on Twitter will have you believe!)
This fundamental insight led to Explorer’s first feature: saved history. They say the difference between tinkering and science is rigor and reproducibility. We codified that transition by making all our prompts and settings automatically saved and shareable. Annotations further allowed us to build and improve upon previous results.

Long term, we envision an environment that enables the development of higher level APIs on top of GPT-3. A system for designing pipelines of prompts, complex scripts, logic to retrieve external information — that can all be published for use in products and research.
Building complex applications requires you to work at a higher level of abstraction than machine code. How can we build an open source operating system for GPT-3 at the dawn of Software 2.0

Building together

The community is continually discovering incredible methods for unlocking GPT-3’s potential. Without the right set of tools, most practitioners struggle to find and implement these new techniques.

That’s where we see the need for an open source project where we can all rapidly experiment and codify what we’ve learned. Here are some of the community’s methods and best practices that we’re excited to incorporate as features in Explorer:

Prompt chaining

  • This method skirts the token limit by “chaining” together prompts and their outputs.
  • Chaining breaks down larger tasks into component sub-tasks, before composing them back together in a single prompt to get your final result.

Context stuffing

This is a method of iterative prompt improvement that involves:

  1. Running an initial prompt
  2. Cherrypicking the best GPT-3 outputs,
  3. Appending them to the original prompt, to provide more examples for GPT-3 to use.

The Roadmap

In the coming months, with your help, we plan to build these features (and more!). As the private beta comes to an end, let’s enable new GPT-3 users to hit the ground running with the cumulative, collective insights of the community.

As we receive feedback and contributions from the community, this roadmap will continue to change and be reprioritized. You can follow a live version of it here.

Sharing and importing prompts

  • “GitHub for Prompts”. Let’s make it easy to share and build on each others’ work!

Prompt evaluation

  • GPT’s performance is very, very sensitive to prompt design. So iterating on prompt design is our highest priority.
  • With built-in tools for prompt evaluation, you can annotate, rank and sort results for the same task across different prompt designs.

Scripting & chaining

  • Enable commands which can trigger scripts.
  • These scripts can then: i. do computation, or ii. trigger API requests to do actions, get more information, call other prompts, etc.
  • In particular, this scripting interface can chain prompts together.

Higher level API interface

  • The OpenAI API is all about simplicity, it is a raw text-in, text-out interface, which is great to build on top of but does not make it easy to best take advantage of the tool.
  • We envision a higher level API which you can use to develop higher-level reasoning and operations on top of the low level API. Turn any prompt, which itself can be a pipeline of other prompts, into an API.
  • This enables a more broadly useful natural language interface for: chat, question answering, text transformations, multi-step labeling, and other pipelines of natural language operations

Try Explorer.
Contribute to the open source project.


Old Friends New Company

Last month, I started a new company with two old friends.

We’re building an intelligent screenshot manager. If you’d like to try it, sign up here:

Becoming the Personal Google.

Like I said, we’re building an intelligent screenshot manager. But that’s just the wedge to something much bigger.

Today, it’s shocking that there’s no unified way to effortlessly capture information, structure, and store it. Sure you probably have that one friend who meticulously categorizes their whole life in Workflowy, or their personal wiki. But what about the rest of us?

We’re building that tool: to instantly capture information from any source, across any device. What Google did to organize the world’s knowledge, we’re doing for all the knowledge in your life.

As one alpha user put it, “What Ilike about this thing is — I can be messy, but it’s still organized”.

Here’s Professor Dumbledore demoing our alpha

Beta launch: version 1.0.0

We’re launching our desktop app on April 13th.

On March 18th, we launched a closed alpha version of Relephant, inviting friends and users who paid for early access. It was the minimum feature set — a prototype. But it was enough to validate the app’s utility. We’ve hit on something special.

With these users’ insights and feedback, we determined the feature set for our next product launch: Relephant version 1.0.0. 

Version 1.0.0 will be a native desktop app. It will dramatically improve the user onboarding experience, and include native features like clipboard control and deep linking.

How we’re thinking about the next 6 months

Our top priority is to build a base of power users who love Relephant.


Our 6-month goal is to sprint towards product-market fit (PMF). This means building a small but devoted base of early Relephant users. The Paul Graham wisdom along this vein is to “make something that a few people love.” [1]

We’ve articulated our vision above, to become the Personal Google.

This is how we start. Building out a product that delights our users: enabling them to capture, store, and retrieve information that would otherwise get lost in the ether of their busy lives.

Metrics – Tracked

Our product goal informs our metrics.[2] We want to see high frequency use and strong retention.

To measure frequency of use, we’ll track sessions per weekly active user. To measure retention, we’ll track L30.[3]

Metrics – Surveyed

We’re planning to build for a relatively small user base at first. So we’ll supplement our purely quantitative metrics with richer survey data.

Following the Superhuman playbook, we’ll track the percentage of users who say they’d be “very disappointed” if Relephant ceased to exist.[4] We’ll presume to have reached PMF with those users.

This survey will also inform our product roadmap, as we identify the features required to bring more users into the PMF group.

Try it and give us your feedback!

As we prep for the v1 launch, we’d love for you to give our alpha a try. Sign into our Slack community here and I’ll get you set up.


We’ll offer the core Relephant for free: optimizing for adoption and usage data.[5] Once we’ve achieved PMF with a core group of users — whose User Personas and Core Use Cases we can clearly articulate — we’ll form our monetization strategy.

Backend Capital

Wanted to shoutout to the Backend Capital team. Dave and Lucy have been here for us in a way I’ve never seen before from investors. I won’t bore you by listing the myriad ways they proactively helped out. Suffice it to say they’re truly a part of our team, in the trenches with us.

Heads down in paradise.

Our team is holed up in the Santa Cruz mountains to ride out the pandemic. We’re living together in a large house with our significant others, overlooking a beautiful valley. It’s the ideal place for us to be right now: isolated from COVID-19, surrounded by peaceful nature, with little else to do than build Relephant, and bond with each other. I feel incredibly lucky to be together in this safe place with these amazing humans.


[1] “you can either build something a large number of people want a small amount, or something a small number of people want a large amount. Choose the latter” – Paul Graham

[2] “We would never release a feature without releasing the analytics for that feature and understanding what specific measurable result we wanted to create.”- Michael Seibel.

[3] Andrew Chen on engagement and retention metrics.

[4] Superhuman PMF article.

[5] Among the many great pieces of advice we got from you, our investors, a common refrain was to offer Relephant for free. A number of reasons for this: 1. Freemium is the most common monetization model for consumer productivity tools. Why innovate on that? 2. Early on, we need to optimize for usage — so that we can have a strong feedback loop, informing the product roadmap, and hastening our time to PMF.

Slack’s Missing Manual

Last week, I joined a new Slack workspace. It was particularly active. And I quickly found myself distracted by incessant notifications.

I would spend a few minutes every hour catching up on all the unread channels — scrolling through message upon message of irrelevant information — just to ensure I wasn’t missing the rare important message.

My autonomy over my attention was slipping. “Why don’t I feel this way in other Slack workspaces?” I wondered.

Slack is ubiquitous among startup teams. But it doesn’t scale automatically with teams. Instead, we need to architect Slack standard operating procedures (SOPs) that support our teams’ business operations. This takes intention and thoughtful design.

When you fail to architect Slack SOPs, you induce drag on your team. An information flood of useless notifications and noise saps your team’s attention.

Like your org chart, your codebase, and your financial models: you need to re-design your team’s Slack SOPs as you scale.

Macgyver-ing Slack

Slack is a messaging tool. That’s it! But we overload it — jerry-rigging it to act as many other things. Some examples that we use Slack for at Retriever:

  • ticket tracking
  • knowledge base
  • logging of outputs from web workers and cron jobs
  • a virtual water cooler for sharing cat memes

This Macgyver-ing Slack to your team’s needs seems unavoidable. We do the best we can with the tools we have. Your team’s jerry-rigged Slack uses are unique, as are ours at Retriever.

I’ll share the best practices and SOPs that have worked for our team, to help get you start building yours.

General Best Practices + Hygiene

1. Keep conversations in threads
In order to keep the channel clutter to a minimum, try to respond in threads so that messages are kept nice and neat.

2. Keep work out of DMs
Try not to post work-related items in DMs with other teammates. Use the appropriate channels, including internal client channels to discuss these items so that everyone has context and can find the info if needed.

3. Pin messages that need action
Many messages require action by a channel moderator. Pinning the message creates a queue for the moderator to triage, and makes it easy to see which messages require action at a glance in the channel.

4. Emoji reactions
Building on Slack’s own best practices, we’ve mapped particular meaning onto emojis. e.g.) making “read receipts” by reacting to a message with :eyes: to indicate that you’ve read it.

Here are all of our standardized reactions:

We keep these pinned to the top of our #team Slack channel for easy reference.

5. Assign a Moderator
Each channel is owned by a teammate. That channel owner is responsible for ensuring that channel-specific operating procedures are being followed.

Ticket Tracking in Slack

We have a #questions channel in Slack that operates as a just-in-time knowledge base.

When someone has a question whose answer they can’t find in Notion (which is Retriever’s central OS), they:

1. Post it in the #questions channel
2. Pin that post
3. Prefix their post with a 🔴, 🔵, or ⚪️ depending on the urgency of their question (see Tactic #1 above)

Next, the moderator of the #questions channel answers the question, or pings the relevant specialist who can answer.

Once the question has been answered to the satisfaction of the teammate who asked, that teammate will:
1. react to the original question with ✅ :check:
2. and, unpin the message (so that the moderator knows it’s out of their task queue)

Pin the Standard Operating Procedures

For each channel that’s being jerry-rigged for some non-standard business process, do the following:

1. Create a short description and set the channel topic
2. Write out the SOPs
3. Post them in the channel
4. Pin that post to the channel

Want an example?

At Retriever, #response-queries channel is where new teammates get practice drafting responses for our clients. They submit draft responses to be proof-read by an account manager before sending the responses out.

Here are the steps we used to set up the rules for this jerry-rigged channel:

1. Set the topic

2. Write out the SOPs

3. Post them in the channel & 4. Pin the post, so that any teammate can easily find it

Making the most of imperfect tools

Jerry-rigging Slack isn’t the most elegant solution to your business ops needs. But it’s fast and agile.

As long as you take the time to 1. spec clear standard operating procedures, 2. articulate them, and 3. ensure that they’re being enforced, you’ll be able to make the most of this woefully off-label use of this simple messaging tool.

How to live longer

The life we receive is not short, but we make it so. Nor do we have any lack of it–but are wasteful of it.

Seneca the Younger

Time is our most scarce and valuable resource.

Our calendar is the foremost tool that we use to manage and control our time.

Mastery over that tool gives us better control over the usage of our most scarce resource, time.

Spending our time

I get on a bit of a soapbox about time, and its premium in my personal philosophy and ethical system. Sometimes this can feel almost pathological. Why?

  • Over-emphasis on time can cause you to obsess over it — pulling you out of the moment, and keep you from being present. It can allow time to slip by.
  • And the dark side of the time value of money is that it reduces a moment with a loved one down to a dollar value. You can put a price tag on everything, including the hour you spent playing with your children. Was it worth it? At what price wouldn’t it have been worth it? Everything has a price! See how this kind of logic – this rational economics calculation – can pervert moments that feel valuable beyond reduction to a dollar amount? There’s a part of our moral compass that wants certain things be sacrosanct – pure, and off limits from economic principles.
  • Finally, if “time efficiency” is the utmost goal – for the sake of economizing that time, where’s the “human element”?. What inherent worth do I have as a human, whether or not I’m optimizing this scarce resource?

To all that I say, “yeah, ….but you have X more days to live”. [In my case, X ~ 22K]

And then nothingness.

So, how are you going to spend those days? Frugally? Wantonly?

How about thoughtfully and deliberately?

The core economic concepts of time value of money, opportunity cost, time discount factor, and efficiency all drive us towards a more thoughtful and deliberate use of our most scarce and valuable resource: time.


Want a huge leap towards mastery of your time? Learn to use your calendar’s shortcuts.

The difference is night and day. The small friction you feel every time you need to navigate your calendar dissolves. You’ll feel much greater ease of control over your calendar, and therefore, your time.

Google Calendar Shortcuts

T = brings you to current {day, week, month}
W = weekly view
D = daily view
M = monthly view
J = go forward (next {day, week, month})
K = go backward (previous {day, week, month})

Sample ‘Recipes’

Now, when I need to navigate my time, it’s so quick!

I reflexively — and therefore, effortlessly — navigate the UI of my time.

Example 1. “What am I doing tomorrow?”
=> type D (daily view), T (today),J (next day). boom!

Example 2. “What day did I meet up with X last week?”
=> type W (weekly view), T (this week), K (previous week). voila!

How to set it up

1. Go to settings:

2. Enable shortcuts:

3. That’s it! Take a few min to learn these, and enjoy greater autonomy over your most scarce resource ⏳



Official Resources for iCal and GCal:

More on GCal shortcuts from

Or, if you use iCal: iCal shortcuts

A/B Test Result Calculator

Here’s a simple tool that you can use to test whether the results of your A/B Tests are statistically significant. Happy growth hacking!


Plug in your two variations sample sizes (n1 and n2) and estimated success rates (p1 and p2), and scroll down to Interpreting The Results, to understand the results of your test.

The Hypothesis Specification explains how to formulate your A/B Test experiment.

Sample Size A:
Success Rate A (%):
Sample Size B:
Success Rate B (%):

Interpreting The Results

This section will populate after you complete the form above.

Hypothesis Specification

Null Hypothesis: Success Rate A ≥ Success Rate B
Alternative Hypothesis: Success Rate A < Success Rate B
Significance Level, ⍺: 5%

Reference Inputs and Computed Statistics

This section will populate after you complete the form above.

Sample distributions. Red: Variation A. Green: Variation B. The Null Hypothesis is that the true, population average of the distribution that generated the red sample is higher than the true, population average of the distribution that generated the green sample. The Alternative Hypothesis is that green’s underlying distribution has a higher average.

The Magic of Statistics

In statistics vernacular, we’re doing a test of “difference in proportions”, or a “two-proportion z-test”.

The data that we’re considering is analogous to a repeated coin toss. You flip the coin, and it either comes up heads or tails. Then you do it again, and again, …

The distribution that this sort of data follows is called a “Binomial Distribution”. It’s characterized by two parameters: sample size (denoted by the variable n, for number of coin flips), and probability of success on any given “coin flip” (denoted by the variable p, for probability of success).

Many business applications with a discrete outcome follow a Binomial Distribution:

  • ad click-through (n = number of ad impressions, p = probability of click-through),
  • email open (n = number of emails sent, p = probability of an email being opened),
  • website sign-up (n = number of website visitors, p = probability of a visitor signing up),
  • checkout conversion (n = number of users who go to the checkout page, p = probability of successful checkout)

As such, we often gather this data in the course of growing our businesses: optimizing our ads, websites, and funnels for conversion.

In order to improve, we iterate and run split tests on different variations of these funnels.

To understand the results of these split tests, we need to use statistical methods like the above difference in proportions test. That way, we can have confidence in moving ahead with the best variations for our sales and marketing funnels.

Here’s the stats theory:

We first take the two parameters that we need to characterize each sample’s distribution: the sample size (n) and estimated proportion of successes (p). What we’re doing is taking these two poorly behaved sample Binomial distributions, and merging them together to create a well-behaved statistic, the Z-statistic

We construct a well-behaved Z-statistic from our sample data (n1, p1, n2, p2). The Z-Statistic follows a Gaussian or Normal Distribution, ~N(μ=0, σ=1), from which we can determine the probability of witnessing the data that we witnessed, under the Null Hypothesis.

From there, we’re able to easily evaluate the probability that what we saw was due to chance (the “p-value”), and therefore, determine which variation of our split test had the higher Success Rate (p).

The further our Z-statistic is from 0, the lower the probability that the Null Hypothesis. is correct.

Luckily for you, the above tool will not require you to do even think about the Z-statistic! But it’s probably helpful to have some idea of what’s going on under the hood.


This gets much more complicated when considering one- vs two-tailed tests, and varying the hypothesis specification. So please, for simplicity, and to avoid errors in your analysis — make sure that you specify your hypothesis as formulated above.

Hope this helps! Happy growth hacking 😁📈

Further Reading:

Hypothesis Test: Difference Between Proportions
An Introduction to Hypothesis Testing [Youtube]

My User Guide

If we understand how our teammates communicate, work, and learn best, we’ll be more happy and effective as a team. Maintaining a User Guide allows people to understand how to work with you. That’s why I have every member of our team at Retriever build and maintain their own User Guide.

And there are substantial side benefits. It prompts introspection and fosters self awareness.

Below is the exact User Guide that I share w/ my team at Retriever. It’s tremendously helpful in quickly acquainting new colleagues with me and my working style — including my habits, quirks, and gotchas.

Note: this User Guide is very targeted to my role and my team at Retriever. Likewise, you should consider your context and audience, and target your User Guide accordingly.

Stedman’s User Guide, as Retriever CEO


  1. I process things visually. I need to see written notes to get my bearings, maintain attention, and keep track of where we’re going.
  2. I am a direct communicator. Please let me know if this style does not work for you and I will modulate it.
  3. Please explain things to me on a call, not on Slack.
    • I probably don’t have context for whatever you’re working on, and a 1 min call will allow you to give me the context that I need to help you out.
    • If it’s not worth getting on a 1 min call to discuss with me, it’s probably not worth asking me about anyway!
  4. I LOVE feedback. Critical feedback is great. There are things I’m messing up every day and if you don’t tell me, I probably won’t improve.

Things that you should know about me:

  • ⚠️ Communication style
    I’ve been told that I can be “intense”. If you feel I’m too direct, or intimidated, please let me know, and I will try to modulate during our discussions! That said, I am probably best suited to work directly with folks who respond well to my intensity. Please understand that even if I come off that way, it is coming from a place of enthusiasm and energy rather than aggression.

    Another key to my directness is that I think it’s the best way to get to the truth. Underpinnings of this view:
    Disagree and Commit
    Harmony vs Productive Conflict

    I also talk loudly. I grew up talking loudly because my mother is very hard of hearing and she needed us to speak up. If I’m speaking too loudly, you can let me know!
  • ⚠️ Take notes during our meetings 
    I don’t trust my memory — or yours. So I want a written record of the important stuff. This means, if we’re having a meeting, I want notes taken. If I can trust you to take good notes, I will be much more at ease, and you will get more out of me in meetings.
  • ⚠️ Context switching
    I’m bad at it. If you need something from me that’s urgent, ask to get on a quick call to explain it (not on Slack). Otherwise, make a task in Asana with a due date of Today and rest assured that I will look at it.
  • Convince me with data 
    I make complex decisions with data. e.g.) If we’re trying to decide how many people to hire next month, I need to see 1. a forecast of our growth, and 2. a forecast of our current team’s capacity to absorb that growth. The answer (how many people to hire) should be in line with those two things. The reason is that – when we’re making decisions under uncertainty – it’s important to articulate what we don’t know, and to try to be as precise as possible in the face of that uncertainty. When you start to layer uncertainty on top of uncertainty, it only takes small errors in each forecast to yield huge errors in the overall output. Back to the hiring example, if we’re a little over on our growth forecast, and a little under on our capacity forecast, then we can easily end up hiring too many people, which is very costly to us. Before making long-term strategic decisions, I need to see a model that justifies our decision.
  • Timing/Schedule
    • I like working at night. You’ll often get pings from me in the wee hours. There is no expectation that you respond to these at night or that you likewise develop nocturnal habits.
    • I’ve never been productive in the afternoon. Unless I have fixed engagements, I prefer to take a few hours in the afternoon to exercise, think about high-level strategy, or meet with people/take calls.
  • Closed-mindedness
    I’m conservative in the sense that I prefer to “stay the course”, and tend to be pessimistic about exciting new endeavors. I will often try to pick apart an idea even if I’m excited about it. Don’t let that discourage you; on the contrary! A good idea will stand up to the scrutiny.
    • In part, this conservatism is learned. We tend to succumb to “grass is greener” bias, and want to pursue the “shiny new” thing. I want to counteract that bias. This makes me err on the side of pessimism when considering new initiatives.
    • Furthermore, as CEO, my job is to keep this company aligned and moving in a unified direction.
      • Companies are like boats. Every time you change directions, you lose speed (focus, alignment, and subsequently, execution). Suppose we’re trying to sail north, and we find ourselves a few degrees off course. I’d rather continue slightly off course, and maintain our speed, rather than re-route to True North and lose speed in the process.
    • That said! Sometimes I’m unduly conservative. If you feel that I’m flippant or dismissive, please flag it. I don’t want to be negative or closed-minded. I just want to protect the overall company’s trajectory.
    • Finally, I can take a while to come around to new ideas/proposals. If you come to me excited with a new idea, don’t take my initial deadpan response as dismissal. I’m just trying to wrap my brain around it.
  • Process
    If you’re in charge of something — from a Client’s outbound campaigns to managing a team at Retriever — I want to see that you have process. We are a “Business Process Outsourcing” company: “process” is literally in the name! It’s crucial that we create deterministic processes in our company — rather than be running on habit, whim, and fickle memory.
  • Scale yourself.
    Follow these two rules:
    1. try to pass off as much responsibility to your direct reports as they can handle.
    2. If you’ve done that, and you’re at capacity, let your manager know.If everyone follows these two rules, we will be ensuring that everyone on the team is as highly-leveraged as possible.
    • Is there a meeting you don’t think you need to be part of? Ask your manager if you can be removed from it, or made optional. That’s why we have meeting minutes 🙂
    • Is something repetitive crowding out time you could be spending on higher ROI activity? Let your manager know, and you can work together to find an efficiency gain.
  • Getting in the Weeds
    When considering strategic questions, I often jump too quickly to implementation details. Try to help me stay high-level and abstract until we need to address implementation.
  • Language
    I use a lot of acronyms. I also have an annoying habit of using bigger words than necessary.
    • If you catch me saying a word that is unnecessarily esoteric weird, please let me know.
    • If I use an acronym that you don’t know, please ask me what it means. I’m probably using it because it’s important enough of a concept that we need to reference it often + quickly, so you should know it too. A short list:
      • ICP: ideal customer profile; target market
      • ACV: annual contract value; how much a year’s worth of something costs
      • ROI: return on investment
      • LTV: lifetime value of a customer; how much profit we make from a single customer
      • SG: sounds good
      • RN: right now
      • BTCHESA: “best team capitalism has ever seen assembled”; i.e. the Retriever team 🙂
  • Growth vs fixed mindset
    • I believe in my capacity to grow and improve.
    • As part of this User Guide, I listed some of my blind spots and shortcomings. But I’m still actively working on improving these things. So much so that I anticipate removing them from this User Guide once I’ve overcome them.
    • Likewise, I believe that you can grow and improve! If I flag an area of weakness for you, that’s not an eternal sentencing. I have confidence that you will be able to improve upon it — if you want it badly enough! Related:
  • Decision-making
    I often need a while to think through big decisions. Don’t expect me to make decisions on the spot. Let me talk/ramble through them plz , as that’s how I can best think through them.
  • Personal growth through feedback
    Personal growth is one of my 3 core principles (along with “creating impact” and “having fun”). So please, please, please don’t hold back the constructive criticism. It’s your greatest potential gift to me! I promise I will not be offended or upset, and I will be indebted to you for the insight. I’m particularly interested in feedback on my communication style: how can I be a more effective communicator? Am I using “um” too much? Do I come off as impatient? Is there something I’m doing in our interaction that makes you want to avoid interacting with me in the future?


Here are some other articles and resources that I found helpful while constructing my User Guide:

Questions to get you started

Questions focused on yourself:

- What are some honest, unfiltered things about you?
- What drives you nuts?
- What are your quirks?
- How can people earn an extra gold star with you?
- What qualities do you particularly value in people who work with you?
- What are some things that people might misunderstand about you that you should clarify?

Questions focused on how your interact with others:

- How do you coach people to do their best work and develop their talents?
- What’s the best way to communicate with you?
- What’s the best way to convince you to do something?
- How do you like to give feedback?
- How do you like to get feedback?

Hope this helps!


Why You Should Cultivate Buyer Empathy

After two years of pitching and selling, I found myself on the other side: as a buyer of a similarly complex, expensive, and strategic service — needing to make an immensely important purchasing decision.

I recently embarked on a Buyer’s Journey to find a marketing agency to help grow our company, Retriever.

This experience has helped me build empathy for “the buyer”: what goes through their heads as they do their diligence to survey the market of outsourced sales development services, to find the one that’s best for their business. Being the buyer is hard!

I wanted to share this insight with other sales leaders, so that you too can empathize with your buyer — better understand their psychology, and win more deals.

Here’s our Buyer Journey, and principles of effective sales tactics that I learned along the way.

Stage 1. Problem/Need Recognition

For us, the first step of the Buyer Journey was realizing that we had a need. We wanted to generate an inbound pipeline of marketing leads to supplement our sales efforts.

Since Retriever’s inception, we’ve grown by “dogfooding”: i.e. using our own service: outbound sales via LinkedIn.

And in less than a year, we’ve grown with outbound from a team of 3 to 20. 100% bootstrapped. I’m quite proud of what we’ve built with zero outside funding.

But as I began planning strategic initiatives for Retriever in 2020, I wanted to diversify our go-to-market — to generate inbound leads through paid ads and content. That was the first step in our buyer journey.

Stage 2. Consideration

Now that we knew we wanted to scale our marketing, we needed to answer two questions:

1. what marketing agencies are out there?

2. which agency would be the best fit for Retriever?

To answer the first question, one of our teammates researched and reached out to about 10 marketing agencies, scheduling calls with them to learn more about them.

Those calls offered amazing insight into how I can improve my pitch.

Here are some principles that I noticed:

Buyer Empathy Insight #1: Be clear and articulate

When people sound like they have their pitch nailed down, I feel like they’ve done it before, they’re professionals — veterans. “They know what they’re doing.”

Inarticulate explanations, bumbling, and rambling, sapped my confidence in them. One vendor said “page on time” referring to “time on page”, making me wonder, “is this person not used to discussing these metrics? I should hope not!”.

When the lead asks a question, answer first, then give context!

e.g.) cost? => not having a clear answer makes me think either 1. they don’t know (they’ve never sold this before!?) or 2. they’re being unnecessarily obscure (which breeds distrust)

Buyer Empathy Insight #2 Dig deep to understand your prospect

I felt more confidence in the vendors who asked detailed questions to understand my business model, my goals for growth, and other relevant constraints: my budget, my timeline, other go-to-market initiatives, specifics on our target customers, etc.

Shallow questions made me feel like they’re inattentive, not customized to our particular needs and constraints, and made me worry that they wouldn’t understand the business well enough to run marketing for us.

Further, I wanted a detailed proposal from these vendors. Working backwards, then, they needed to ask detailed questions about us, so that they could personalize their proposal for us.

Buyer Empathy Insight #3 Qualification is about the lead — not about you

One vendor spent most of his call with us figuring out whether we had authority and budget, and not asking specific questions to determine what we needed.

In BANT* speak, it was all “Budget” and “Authority” and no “Need” and “Timing”.

[*BANT = qualifying leads by determining their Budget, Authority, Need, and Timing.]

BANT is fine as a qualification framework. But it should startwith the Need. Understand why this person is on the phone with you? What pain are they feeling?

From the “Buyer Journey” perspective, what was the “Problem” or “Need” that they recognized? Dig into that before trying to tick every checkbox on your qualification list.

Buyer Empathy Insight #4 Be an Expert: Recommend a Specific Strategy

After asking specific questions: put your expert hat on.

I spend a bulk of my day leading people: leading my team in meetings, leading my prospects. As a buyer, I want to be led!

I want the vendor to expertly guide me through the process. Once they’ve asked all the questions that they need to create a strategy — a prescription, tailored to the prospect’s needs — I want it laid out perfectly for me.

High level: Help the prospect understand the lay of the land. Explain the universe of possible solutions, and what factors determine which is the best fit. This is sort of like a high-dimensional Gartner Magic Quadrant.

Low level: Zoom in on particulars of the strategy that you recommend. Draw upon your experience implementing that strategy.

What does the first week look like? The first month? What’s the time to ROI? Flesh out a picture of what this looks like, that the prospect can easily imagine for themselves.

Stage 3. Decision

By this point, your prospect likely has all the information that they need to make a decision.

They’ve solicited proposals from multiple vendors. They’ve gotten an understanding of the competitive landscape, their budget and expected ROI, and are deciding i) Which strategy to implement, ii) Which vendor is the best fit to implement that strategy.

By this point, there’s little you can do to further influence the decision.

So as a vendor, the best thing you can do is work backwards to win in the decision stage. In the consideration stage, you’ll need to make the best case that you can for (i) and (ii) above.

First: convince them that your proposed strategy is the best one (i).

Second: convince them that your company is the best fit to implement that strategy (ii).

If you do both of those things, you’ll maximize the probability of winning the deal in the Decision Stage.


Would love to hear your thoughts / feedback on this.

And if you want to talk more about sales, drop me a line: stedman at retriever dot co.

Debugging Churn / Diagnosing an Onboarding Problem

Fixing your high churn rate.

High churn is obviously a common problem for early-stage startups — and myriad causes behind that churn. 

I’d like to explore a subset of founders whom I often see struggling w/ high churn. So far, these founders have done everything right. They talked to their prospective users to figure out what they wanted before building

Q Is this an onboarding problem?

  • Do the users complain about not having features that you do have?
  • Do users churn w/out engaging w/ the product? How do you measure engagement? At what engagement level does churn drop off?
    • <Twitter> famously found that if a new user didn’t <follow> <X> people, they wouldn’t continue to engage with it. After that threshold, though, the new user retention rate increased dramatically. If you have a relatively complex SaaS product — particularly one that requires behavior change — your job is to figure out the analogous threshold, above which retention spikes and users remain happily engaged
  • What is the objective, measurable outcome that identifies that a new customer has been successfully onboarded?
    • NB much of my advice may sound very high-touch — infeasible if you’re only charging users $14/mo.
      I’m speaking mostly to products w/ annual contract values (ACVs) $10K or greater, who can afford that level of touch.
    • But as an aside, we notice that many software companies err too far on the side of fully automated onboarding, too quickly.
      It goes against the idea of “doing things that dont scale” and allows you to rationalize your way out of the hard, heavy lifting. Unless your automated onboarding is already optimized (if you’re experiencing churn, i argue the probability is very low), you’re losing out on valuable customer insight that you’d get if you onboarded manually.

What is the current situation?

  • Who onboards? Sales, CX/Account management, or self-serve?
  • $1M Question: What is the hurdle that new users need to get past before they have the “Ah ha!” moment?
    • Idk what this moment/threshold is. It’s unique to your product.
    • For one of our clients who provide lead generation, their “threshold” is the first time they get a meeting w/ a lead sourced through this tool. 
    • At Triplebyte, it was the first time a company chose to move forward after matching w/ a candidate that they were excited about.
  • More on the handoff b/w sales and {CX, AM, self-service}:

Companies will reasonably disagree on this. Here are a few variations that worked for others. 

Start with Variation C.

It involves a joint onboarding session between 3 parties: the customer, the salesperson, and the customer support rep or account manager (AM) who will own the customer relationship going forward.This has the highest cost, but is preferred for early-stage, high-ACV SaaS products for a number of reasons:

  1. trust / rapport from the existing relationship b/w the salesperson and customer
  2. salesperson has important context about the customer: constraints, goals, other considerations affecting the customers success w/ the product.
  • Airing this context during that onboarding session can: 
    • Give the customer time to elaborate and clarify
    • Give the AM the opportunity to understand those things, and ask clarifying questions
    • Ensure that everyone has shared context — where that context can have a meaningful impact on the customer’s success w/ the product.
    • It also makes the customer feel pampered / well-understood, and taken care of.
  • It’s hard to overstate the importance of all this
    • <cite doctor bedside manner’s impact on likelihood of malpractice suit>
  • It can all feel like over-communication. But — until you’ve solved your churn problem — think of it like an insurance policy.
  • Once you’ve solved churn, you can transition to Variations B or C — both of which require less human resources per new customer onboarding.


Rules-based Public Policy FTW

Rules-based policymaking allows us to find common ground

TLDR: Rules-based policymaking allows us to find common ground.

In my Vulcan-esque political fantasy, we use rules-based public policy in place of many of our fixed policies.

The reason is: it’s easier to agree on ideal states of the world than it is to agree on how to get there. And when we miss this crucial step, and jump straight to policy positions, we lose the opportunity to find common ground.

Rules-based policy in practice: monetary policy

One of the most prominent examples of rules-based public policy today is monetary policy. The Federal Reserve sets monetary policy via their choice of the interest rate. (This conversation glosses over a ton of nuance. If you have beef with that, follow the rabbit hole here.)

How do they determine the interest rate? The generally accepted decision model is known as the Taylor rule, which explains the Federal Reserve’s interest rate decision as the result of their effort to jointly keep unemployment low and inflation under control.

The rigorous mathematical formulation of the Taylor rule specifies a loss function with respect to both unemployment and inflation — and target levels of both inflation and unemployment.

The further inflation or unemployment get from the Fed’s targets, the greater the “loss”. The Federal Reserve thus sets policy (the interest rate) to minimize the output of the loss function — i.e. to keep inflation and unemployment as close as possible to target levels.

Why is this helpful?

I think it’s easier to common ground. We could avoid becoming quagmired in partisanship by debating about the loss function instead of debating about the particular policies.

The reason for this derives from a very popular method of negotiation, popularized in the book, “Getting To Yes”.

In short, the negotiation strategy is to focus on interests, not positions.

This shift from “positions” to “interests” dramatically improves our ability to find common ground, and thereby reach compromise in the negotiation that leaves everyone feeling understood, and like their needs were addressed.

Let’s translate this strategy into the framework of public policy.

“Interests” = your loss function.
“Positions” = your proposed policy for achieving your interests.

Interests vs positions, in practice

Consider again the case of a monetary policy. Suppose we determined monetary policy by debating our “position”, what the should the interest rate be. I imagine the Federal Reserve’s policymakers shouting numbers across the table…

“ We need to raise the interest rate 10 basis points.”
“10 basis points? You’re insane! No more than two basis points!”

Not very substantive!

Moving from “positions” to “interests” allows the policy-makers to instead debate questions like:
1. What is our target unemployment rate?
2. What is our target inflation rate?
3. How do we want to make trade-offs when we can’t jointly achieve both at once?

Again: It’s easier to agree on ideal states of the world than it is to agree on how to get there. And when we miss this crucial step, and jump straight to policy positions, we lose the opportunity to find common ground.

What public policy would you like to see moved from fixed to rules-based?

Perfectionism & Degrees of Freedom

Done Is Better Than Perfect

Perfectionism is one of my most pernicious attributes, keeping me from being the ruthlessly efficient and effective human that I aspire to be. When solving a problem, I feel the allure of the minutia. Plumbing the edge cases, and searching for the true global optimum — even on a relatively flat surface of alternatives.

I’m not sure how this is related to pedantry. It feels similar. Yet, I scorn pedantry, and still applaud undo rigor. Not “rigor qua rigor”; this is where i see pedantry. Rather, I see it as a thorough explanation to find the truth.

In academia this seen as a good thing. Getting published in my field almost always required a closed-form, analytical proof of a new methodology’s validity. I now see this as undue rigor. But I struggle to shake it in business.

Confidence In Uncertainty

I feel a need to fully understand everything before diving in. This initially hindered my ability as a web developer. Upon entering a new codebase, I’d try to understand the entire structure before I felt comfortable contributing. Painstakingly following every function call back up to some root state…

I’d see an unfamiliar class, method, or variable, and alarm bells would go off. I’d grep the whole codebase to see where it was defined — completely derailing my current path of inquiry.

Finally, I learned to be comfortable looking things up “just in time”. I became confident in uncertainty. I trusted variable names and used them to make educated guesses about what was happening. And only if some missing piece of knowledge became i) a blocker to my understanding, or ii) showed up multiple times s/t I knew I’d be dealing w/ it directly, would I stop to learn more about it. In either case, I already had more context to know what I was expecting. (APIs are a good example of this.)

Degrees of Freedom

Vacillating on decisions is another place where perfectionism stymies efficacy.

The key is to realize that there’s no perfect answer. Or, if there is, the cost of the search for it might outweigh the benefit once found. 

This leads to the imperative to make a decisions and commit to them.

Business partners are a good example of where I struggle on this. No business partner is perfect. You and I certainly aren’t. But those partnerships need to be steadfast in the face of extreme challenge. It’s easy to second guess a business relationship before you’ve raised money, hired employees, etc. (It’s even easy after you’ve done those things!) But without a foundation of unwavering commitment — if you always have a foot out the door, feeling FOMO for all the other business partnerships that could’ve been — you’ll be operating way under capacity and are far less likely to succeed.

In cases like these, then, it’s important to have some way to commit to your decisions.

I think of this like degrees of freedom. Take a simple system: x + y = 1. If I asked you to solve for x and y, you’d give me an infinite sequence of pairs, all satisfying the above system. One equation, two degrees of freedom => no single solution.

If, however, I told you that y = ½x or y = 0.5, now you can easily identify a unique solution for x. We’ve reduced the system’s degrees of freedom, and can now find a tractable solution.

We took this thing which was a variable, y, and turned it into something immutable. From that foundation, we were able to solve for x.

So, when you’re waffling on a decision that you’ve already made, and that you don’t have any compelling reason to change: reduce the degrees of freedom. Adjust your mental model to think of that thing not as a variable, but as something immutable.

It will give you a stronger foundation to solve the other problems downstream of that one.

When to Waffle?

It’s clear that there will be times when it makes sense to change these immutable things. “When the facts change, I change my mind. What do you do, sir?”

That’s beyond the scope of this discussion, but i’ll give two rules that can help.

  1. Give yourself some other condition that must be hit for you to reconsider this immutable variable. e.g.) an amount of time, an objective milestone, etc.
  2. Realize that as a perfectionist you likely skew way too far towards wanting to rethink these immutable decisions, and therefore should try to overcorrect for them.