Should startups make bespoke product and deployment commitments to win early customers?

Christian Kleinerman · Sep 22, 2023

Software businesses that assume 'build once, resell many times' usually degrade into services businesses because of customization demands

His airline scheduling startup was premised on building it once and reselling it, but required so much customization it became a services business

4:45 20VC: Are Foundation Models Becoming Commoditised? Do OpenAI and Anthropic of the World Have a Sustaining Moat? Why Smaller Models May Work Better? Why Incumbents with Data Power Win the AI War with Christian Kleinerman, SVP Product @ Snowflake

Sanjit Biswas · Dec 8, 2023

A product can fail because it doesn't fit the business model even when the technology works; products that require heavy custom installation aren't viable if your model depends on a plug-and-play, rinse-and-repeat recipe

Samsara's machine vision camera for assembly-line inspection required custom lighting and product-specific knowledge, so it couldn't be repeated at scale

Scope: specific to Samsara's plug-and-play business model

13:26 20VC: $18BN Market Cap and $1BN in ARR in 8 Years; Samsara | How to Find Product Market Fit Reliably | How to Create a Multi-Product Company | The Pros and Cons of Serial Entrepreneurship with Sanjit Biswas, Founder & CEO @ Samsara

Sean Murray · Jan 24, 2024

Founders must actively balance 'sales debt' — customization demanded to win deals — and know when to walk away from deals

Moving upmarket, every enterprise wants to be treated as special and demands bells and whistles; occasionally a marquee logo is worth the debt, but unbounded customization is the greatest source of friction a founder will face

Scope: occasionally worth taking on debt to land a name-brand logo

33:38 20Sales: How to Scale Into Enterprise Effectively and the Biggest Mistakes Made When Making the Move From PLG to Enterprise, Why Discovery Today is F***** & The Biggest Lessons on How to Do Sales Team Compensation with Sean Murray, CRO @ Greenhouse

Aatish Nayak · Apr 11, 2025

The advice 'never build anything custom' is too blunt: the discipline is teasing apart which demanding-customer requests are truly specific to them versus which others will also want, saying no to the former and using the latter as a forcing function.

Scale's very custom Reddit-summarization labeling product for OpenAI's early RLHF work on GPT-2 was reused two years later when everyone else caught on to RLHF.

Scope: requires tremendous discipline to separate the two categories

7:50 20Product: How Scale AI and Harvey Build Product | Why PMs Are Wrong: They are not the CEOs of the Product | How to do Pre and Post Mortems Effectively and How to Nail PRDs | The Future of Product Management in a World of AI with Aatish Nayak

Sean Murray · Jan 24, 2024

The real cost of taking on too much sales debt is the irrecoverable product and engineering time, not overpromising to the customer

He once had product and engineering spend a year trying to build for a marquee logo and it didn't work; for founders that time is too precious to lose

35:06 Sales debt cost is lost engineering time not broken promises

Ron Gabrisko · Aug 4, 2025

Before moving into enterprise, founders need real sales expertise on the team and must resist enterprises' custom requirements while treating security and governance as non-negotiable

Big customers wield a big stick to push custom engineering, but no enterprise will adopt tech without security, and a mainstream provider must have it

19:00 Resist custom engineering demands but treat security and governance as mandatory

Chris Degnan · Nov 10, 2023

Making bespoke product and deployment commitments to win early customers is necessary pre-IPO but off-limits once public.

As a public company those arrangements amount to side letters and you have to stay within bounds.

51:11 Necessary pre ipo off limits once public

Aatish Nayak · Apr 11, 2025

In frontier markets, companies should get deep with — and even overfit to — the best frontier customers, because those customers effectively define the market and everyone else follows.

At the cusp of a new market, frontier customers' demands become the general requirements later; Scale's custom pedestrian-labeling work for Nuro was soon requested by everyone else.

6:32 Overfit to frontier customers who define the emerging market