This interview argues that investing in software companies requires a biologist's curiosity, not a physicist's formulas. Benchmark's Miles Grimshaw says the best companies don't want to be copies of existing versions—they want to be the best version of themselves. Key holdings discussed: Segment (switched from API-call pricing to per-user pricing, and dropped a big client), Benchling (started with academic users, walked away from a $1M client), and Lattice (bet all engineering resources on a second product at just $5-7M ARR—the product was 'crappy but effective').
At a Glance Benchmark partner Miles Grimshaw proposes evaluating software companies through a biological lens, emphasizing that early-stage investing requires attention to a company's "DNA" (e.g., investing in Segment, Benchling, and Airtable when they had fewer than 30 employees). Core views includ
Benchmark partner Miles Grimshaw (in his early 30s, who invested early in Segment, Benchling, and Airtable when each had fewer than 30 employees) proposes evaluating software companies through a biological lens, rather than applying SaaS physical formulas. Core thesis: The best companies do not want to become replicas of existing versions; they want to become the best version of themselves—investors should approach with curiosity and imagination, like Darwin stepping off a ship to discover new species, to uncover a company's unique DNA.
Miles Grimshaw argues that early-stage software investing should not rely on physicist-style fixed rules (SaaS metrics, comparable company templates) but instead adopt a biological perspective.
Key Analogy: YouTube did not want to be Flickr for video; Amazon was not just Barnes & Noble online; Shopify was not Demandware for the mid-market. Grimshaw emphasizes: "People think early-stage investing is about risk-seeking, but the biological perspective is about change-seeking."
Benchling Case: When Grimshaw first encountered Benchling (only 5 people, a single standalone tool), the classic perspective would have said, "This is Viva for R&D." Instead, he chose to be curious: Why are academic institutions using it? This hinted at a natural evolutionary path in R&D from academia to industry. Result: When Viva went public, it had about 150 customers, while Benchling already had 1,000—applying the template would have killed its entire potential.
Grimshaw proposes three dimensions for assessing the "DNA" of early-stage companies, arguing that these are better predictors of long-term success than short-term metrics.
Grimshaw believes the second product launch should happen earlier than expected, and "bad but effective" is better than "perfect but late."
Drawing on his experience at Segment, Grimshaw illustrates that blindly imitating successful API companies (Twilio, Stripe) can lead to fatal mistakes.
Grimshaw concludes: “We only later realized that we should have approached implementation, professional services, and support packages more like a vertical SaaS company.”
Addressing the widespread market skepticism toward pure-play API companies (using Twilio as an example), Grimshaw offers a different perspective.
Grimshaw offers two core recommendations for founders in the current environment:
1. Revisit operating assumptions: Should the aggressive support model of the past be maintained? With stronger market positions, capital is now harder to come by for the 15th competitor. Brex exited a certain market segment—the question of "whether to exit" is worth every team’s consideration.
2. Distinguish between core maintenance and growth investment: For companies with $100M+ in revenue, ask yourself, "What is the minimum cost to maintain the status quo?"—anything above that is an investment, and the question becomes, "What is the return on this investment?"
Key lessons on customer selection:
| Position | Guest Stance | Key Data |
|---|---|---|
| Segment | Bullish (Invested) | Shifted from API pricing to MTU pricing; Dropped Hotstar customers; Launched Personas development at ~$30M ARR |
| Benchling | Bullish (Invested) | Started with 5 people; Academic user base; Dropped Gen 9, a $1M ARR customer; Regeneron as key customer |
| Lattice | Bullish (Invested) | Third to market; Launched second product at $5-7M ARR; Dropped a large customer with 2,000 seats |
| Airtable | Bullish (Invested) | No specific data provided |
| Twilio | Neutral (Risk warning) | $4B revenue, 50% gross margin, NDR 120%; Free cash flow still negative; Uber is the only known large-scale in-house build case |
| Stripe | Bullish (Not invested but highly rated) | Key customer Shopify; Global payment complexity extremely high |
| Figma | Bullish (Not invested) | Elevated design department status; Cross-functional internal network effects |
| Viva | Neutral (As comparison) | ~150 customers at IPO; Launched Vault at ~$50-70M revenue |
| HubSpot | Neutral (As comparison) | Launched sales product around IPO (~$70-100M); Adopted "launch-re-launch" strategy |
| OnlyFans | Neutral (As comparison case) | Message-based "whale" customer dynamics; Subscription + pay-per-message |
| Patreon | Neutral (As comparison case) | Tiered subscription model; Lacks message-based payment feature |
| Salesforce | Neutral (As data point) | Customer base grew 47x in FY2007; Customer base grew 8-9x over 10 years in FY2012 |
| Greenhouse | Neutral (As comparison) | Recruitment data half-life of 3-6 months |
| Gong | Neutral (As trend case) | Collaboration tool around sales calls |
| Jasper | Neutral (As AI case) | Replaces high-value labor (marketing content) |
1. “The best companies don’t want to be replicas of existing versions; they want to be the best version of themselves.” (Miles Grimshaw) — Investors should avoid applying templates and instead ask, “What will make this company the best version of itself?”
2. “The half-life of information determines software stickiness.” (Miles Grimshaw) — Recruitment data has a half-life of 3–6 months (low stickiness), while performance management data has a half-life of 12–24 months (high stickiness). Early on, this can be used to infer customer retention potential.
3. “At $5–7M ARR, Lattice shifted all engineering resources from the core product to a second product — the product was terrible but effective, yet it completely changed the competitive landscape.” (Miles Grimshaw) — Second product launches should happen earlier than expected; “terrible but effective” is better than “perfect but late.”
4. “Segment once had a $1 million contract that was never implemented after 12 months — we mimicked the self-service model of API companies, but what customers needed was change management on the level of a database migration.” (Miles Grimshaw) — Do not blindly copy the pricing and delivery models of successful API companies; the implementation experience of vertical SaaS may be more applicable.
5. “Twilio’s NDR was still 120% at $4B in revenue — if it were truly fully commoditized, you wouldn’t see that kind of customer retention.” (Miles Grimshaw) — Market skepticism toward pure API companies may be excessive; the core issue is graduation risk (build vs. buy), not commoditization.
6. “We would rather have $3 million in revenue than $5 million if that $3 million is stronger and lays a better foundation for future success.” (Miles Grimshaw) — Early customer quality matters more than quantity; proactively walking away from bad customers is an important capability.
7. “Figma not only accelerates collaboration but also makes design work visible to everyone in the company — it elevates the status of the design department and creates a career-empowerment ‘meta-game.’” (Miles Grimshaw) — The value of software lies not only in its functionality but also in how it reshapes power and status structures within an organization.
8. “OnlyFans created a ‘whale’ customer dynamic — the message-based payment model allows superfans to invest deeply, making it more like a gaming economy than traditional subscriptions.” (Miles Grimshaw) — Product architecture determines the business’s genetic code; understanding the demand curve and designing low-friction upselling paths is key.
9. “Pricing and packaging are founder-level genetic setting issues — don’t hand them over to the finance department. Ask, ‘From the perspective of the long-term customer journey, what architecture maximizes value capture?’” (Miles Grimshaw) — First set the global maximization framework, then optimize locally; avoid starting with “how to charge a little more.”
10. “Software is a blend of art and science — the best product experiences have aesthetics and craftsmanship, and this is just as important in B2B.” (Miles Grimshaw) — Product “magicians” (e.g., Figma’s Dylan Field) create simple yet beautiful products while naturally locking in cross-functional usage and ecosystem building.