Scuttlebutt & Seed

One of the best tools I’ve found to understand a market or company opportunity is the cross-sectional view, what Phil Fisher calls “scuttlebutt.” How do competitors, customers, investors, current & former employees all perceive a company and its products? As one spends more time building these cross-sectional perspectives, the understanding of a space compounds on itself.

Seed investing can often be a mode of scuttlebutt research for markets in their formation stage. Markets tend to bubble up with companies arriving in clumps and clusters — this year, it was permutations of Airtable and Plaid — and going through the very motion of early stage investing over a few months or a year can land you in front of competitors vying for a similar prize, building on a dimensional view. The tradeoff made with receiving this lens into new markets, technologies, and products is potentially a lack of ability to make the most of it. The window of opportunity for seed investments is tight with limited bites at the apple.

I joke with my friends in traditional venture or growth that they have the luxury of time to let a market play out a bit, but I’ll concede the competition they face is stiff. I may have taken a view towards investing that hinges on consistently picking category-defining or market-dominant companies, when there’s less pressure to do so as AUM decreases. Yet, the possibility of being involved with those businesses that grow for years and define generations is the allure of the venture business — there’s always an opportunity cost with missing, regardless of stage.


“Only slight knowledge and simple machinery went into the manufacture of the first Ford. No trained engineers were used. The frame of the car was moved manually; it helped that it could be lifted by two men. The modern auto factory, hi contrast, is itself a complex and closely articulated machine. Nothing is done by muscular effort. Computers control the flow of parts and components. Only the hideousness of the product, on occasion, reminds us that human beings are involved. We see here one reason for the increase in capitalization of Ford from £30,000 to £2,200,000,000.” - JK Galbraith, Reith Lectures 1966: The New Industrial State

The maturation of the automobile, both as a technology and an industry, introduced increased complexity, planning, and organizational bureaucracy into the production of individual cars. As it became a more legible technology, layers were added in, and the process of development and assembly became highly mechanical. Any sense of craftsmanship and dynamism was all but lost. 

In a piece from the beginning of this year, Ben Thompson points to the full bloom of the software paradigm, writing, “In other words, today’s cloud and mobile companies — Amazon, Microsoft, Apple, and Google — may very well be the GM, Ford, and Chrysler of the 21st century.” (The whole post is worth reading). While we are far from seeing the totality of impact of software on our lives, I have been thinking quite a bit about the state of software business models recently. While my lens does not stretch back many years, there are a few things worth noting that could support this claim:

  1. Beyond Equity-Based Funding Models -- In Public to Private Equity in the United States, Michael Maboussin and his team note, “buyouts of software firms, a sector with above-average multiples, rose from 6 to 17 percent of all deals over the past decade.” Buyout arrangements are reserved for industries with hard assets, and financing software with debt demonstrates a sophisticated understanding of the recurring nature of enterprise software. 

  2. Coalescence of Capital -- Venture investors seem to have collectively agreed on funding a few categories -- collaborative (and horizontal) applications, infrastructure software and developer tools, fintech rails, and vertical SaaS + marketplaces. I kid slightly, but the level of pre-emptive investment in these categories signals to me that there many assumptions baked in about market size, runway, and margin structure at steady state. 

  3. The “Markets Driven” Investor -- This hypothesis is far from fleshed out, but I tend to see more specialist investors winning competitive deals (again, not a translatable signal to eventual return) vs. generalist investors. One can point to the prevalence of platform funds (which host a cadre of specialist GPs), founder sensitivity to the cap table (picking an investor who shares POV and understands the space), and network effects stemming from being in a portfolio cluster. In some eyes, the prevalence of top-down thinking signals the “end of cycle.”

The last argument needs a bit more development, and I’m not exactly sure I’m right yet, but it’s something I’m watching. Another discussion worth having is if venture is “less risky” than previously anticipated and returns are compressed due to capital in the asset class, how should large allocators treat it given their portfolio models?

I highlight all these points to say that it feels as if startups have become more mechanical in the past few years. Some categories even seem formulaic i.e. “Building an OSS business? Here are steps 1, 2, and 3 to world domination.” I’m guilty of propagating this myself. There’s certainly merit to guiding heuristics and frameworks, as founders should not have to make easily avoidable mistakes. But when the steps for idea generation, reaching product-market fit, and company building become too concrete and almost algorithmic, its worth questioning. 

The beauty of startups is in their dynamism. They benefit from uncertainty, while big companies falter from it. A startup benefits and captures values from small dislocations and changes that lead to massive overhauls in adoption and behavior. The inspiration for this post came from a re-reading of Elad Gil’s “10X Your Business.” He writes, “ Do x, then y, then z, each of which has a linear increase in the value of the product or company.  I think it is good to periodically get out of that way of thinking and ask what sorts of deals, adoptions, or customers would completely change the game for the company. These things should be at least borderline realistic.”

Our mental model fantasies have relegated us to the fact that Google’s structural, product-driven network effects always contributed to its dominance with limited to no hiccups. History gets re-written. We rarely talk about some of the initial distribution deals that jump started the engine. While software continues to mature and that force feels insurmountable, I would love to see more of this thinking Elad pointed to in our ecosystem now. One recent example was Kustomer willing to buy out its competitor Zendesk’s contracts as a means of customer acquisition. It was brash, bold, counterintuitive, and unexpected. It’s too early to tell how that exact chess move will pan out, but I applaud the strategy.

Product Areas - September 2020

Over the long weekend, I thought it might be worth underscoring a few product areas that are top of mind.

Data Integration for Regulated / Legacy Industries

I’m interested in the Fivetran model, a data infrastructure company that’s grown significantly recently. They create and maintain pre-built connectors for over a hundred common data stores, applications, etc., so customers can centralize operational data with Fivetran pipelines. Fivetran’s competitive advantage widens as they build more connectors, as well as their own internal infrastructure for quickly spinning up new integrations, hoping to take advantage of their own scale.

Most of the Fivetran connections are with stores like Snowflake or S3, which can be served across industries. Yet, there are a host of vertical-specific integrations that tend to be quite tedious and gnarly, that both initial integration and upkeep contribute to. Regulated industries also tend to have special requirements around auditing & compliance, along with software delivery — either on-prem or in a single tenant cloud. Given the scope and size of some of these industries, I sense schlep-work-as-a-service to be a source of opportunity here within life sciences, industrial, energy / utilities, government, automotive, and more. One could also wedge into building broader vertical-specific data infrastructure platforms (orchestration tool, operational database, catalog, audit log, migration & replication).

Developer-First Product Analytics

A customer is no longer singularly represented as a row in a relational database — they are instead represented as a collection of events across transaction stores. Arguably, two of the most important companies that feed into this for product analytics and events are Amplitude and Heap. Both of these tools track metrics about product usage, guiding product development and pushing users through the customer journey.

When it comes to what’s next in the market, some point to the increasingly prevalent role of engineering practices across the product development journey, especially for smaller startups going through fast iteration cycles. For example, given its emphasis on speed, integrations, and keyboard-first design, Linear is a project or task management tool built for engineers, not PMs or managers. A similar thing could happen in the product analytics space with an open source-first approach — PostHog is an example. However, primarily inspired by Segment’s Config API, I’m interested in an approach for product analytics that replaces the graphical interface with an API. In Segment’s case, they noticed customers wanted extreme extensibility, so they exposed all their data models (users, access tokens, sources, destinations, etc.) as an API. I think it’s brilliant. If developers are the ones building product, why shed one’s opinion about the web interface for a product analytics tool? Just hand over an API.

Automations Library

What comes in the automation universe after RPA and process mining software? I see a bit of a bimodal split. On one end, there is a developer-centric approach — give developers the libraries and frameworks to write all-powerful scripts that can automate different pieces of business software. This is what Robocorp is doing (and to a lesser extent, in test automation). On the other end, work with CIOs top-down to set up an app store of common automations within a company. Given their involvement with large financial institutions, this is the approach Instabase is taking. In either case, a “library of automations” is a unifying primitive.

Gruntwork is a company I find fascinating. They offer companies access to a library of common Infrastructure-as-Code templates, along with common playbooks or recipes given the profile and characteristics of a customer. With tools like Robocorp or Instabase that allow users to design automations, is there room in the market for a vendor-agnostic automations library?

If you’re building in or interested in chatting about any of these spaces, feel free to send me a note at aashay [at]

App Clips

In the last edition of RFS 100, I touched upon my interest in App Clips, the new SDK from Apple that enables fast and ephemeral applications in iOS 14. New development primitives are worth paying attention to, as they augment existing use cases (for App Clips, renting a scooter or paying for coffee) and introduce new user paradigms that expand and contour the market in interesting ways. With integrated payments & notifications, most existing native mobile apps actually fit the App Clip form.

App Clips are lightweight installations of apps, accessible without having to download a standalone app. Users can obtain them by scanning a visual code / NFC tag in a physical space or through a mobile web banner. Additionally, developers can integrate Apple Pay for quick payments and push notifications in an 8 hour window.

Apple emphasizes reduced functionality in their development guidelines, presenting App Clips as a return to the original vision of iOS apps, which were really meant to be utility-like supplements for PC applications. App Clips do have some detractors, as critics position them as a way for Apple to bring developers into a closed ecosystem, for use cases that were emerging on the “open” mobile web. I’ll spare you that debate for now.

I have a few ideas for folks interested in building in this space:

  • API Abstractions — Apple wants the App Clip development experience to be straightforward if a developer has a pre-existing app, but I still think there are design nuances for this new form factor. For example, RevenueCat has layered a simple abstraction, analytics, and cross-platform compatibility (React Native, Flutter, etc.) for a different iOS SDK, In-App Purchase.

  • Marketing & Attribution — While Apple would prefer that App Clips are not exclusively used for marketing, they still have merit as a distribution channel. Tools that help creators and developers understand how to design App Clips conversion, create funnels, and track conversions will be useful. They can also play nicely with an API like Radar.

  • Issuing Infrastructure for Tags — As noted earlier, App Clips are activated by scanning unique tags, either physical or digital. Apple still has to release detailed information on how to create these tags, but I think there are parallel opportunities with credit card issuing platforms like Marqueta i.e. programmatic infrastructure for App Clip tags (production, fulfillment, shipping all under the hood).

These are early days for the App Clips ecosystem, but I’m excited about what could materialize. If you’re building in the space, feel free to send me a note at aashay [at]

Short Run / Long Run

There’s always something.

Currently, early-stage investors are complaining about (yet participating in) a supposedly frothy funding environment, catalyzed by a surge of interest in software infrastructure. During the early days of the pandemic, people were preparing for the worst, but the split between the digital and real economy has spun a different story.

Pre-seed and seed investments are quickly marked up with subsequent financings happening within a matter of months. It’s easy to call bubble, but Galbraith explains the difficulties associated with that:

The euphoric episode is protected and sustained by the will of those who are involved, in order to justify the circumstances that are making them rich. And it is equally protected by the will to ignore, exorcise, or condemn those who express doubts.” — A Short History of Financial Euphoria (1990).

At first glance, the markups seem only positive for early investors. The initial investment is worth more than what you paid for it! Great. Nevertheless, I can’t help some level of paranoia, and I wanted to highlight some possible second-order effects of markups.

  • Illiquidity — Josh Wolfe, of Lux Capital, sounded the bells on this a few years ago, where he commented on the danger of late-stage liquidity preferences and ratchets on unicorns that could wipe out early-stage investors. In the 2018 paper, “Squaring Venture Capital Valuations with Reality,” the authors argue that private technology company valuations could be overstated by 50% when controlling for the worth of the individual share classes. I haven’t seen the recent data yet, but I also have a hunch we’re living a very founder-friendly environment, so the late-stage ratchets may not be as strong, which would diminish this.

  • Fight for dollars — Ultimately, ownership at exit and reserve management are major factors in outstanding venture fund performance. If other sources of capital are backing into companies at variable rates, it may be difficult for seed managers to even maintain pro-rata, much less put more dollars to work and play reserves accordingly.

  • Overcapitalization — There is a fear that if you invest too much in a business before one knows where the truth growth levers are (and it has an inherently capital-light business model), you will contribute to aimlessness. Ilya Sukhar, of Matrix Partners, presented an interesting counterargument, citing the need for significant resourcing against product development to build an enterprise-grade business in a B2B software market today.

These are all disquieting in their own ways, but to me, the real danger markups pose for early-stage investors is in acting as proxies for truth; behaving as short run noise, when long run signal matters more. Being a great early stage investor does not equate to being right 100% of the time, but blind faith in personal ability based on portfolio markups is also not the way.

There have been plenty of examples of cohorts of companies that raise lots of money, but plateau in the $500M - $1B valuation range (the no-man’s land of venture). D2C brands and fraud / risk prevention startups come to mind. Their stories and narratives of world domination suddenly change when growth inflects the wrong way.

Loading more posts…