Strategy

Technical due diligence for VC portfolio companies

Editorial · Reveneau · July 27, 2026 · Updated November 4, 2026

Technical due diligence for VC portfolio companies

Technical due diligence is the review an investor runs before investing, to learn whether the company's technology is worth the price being paid for it. It covers four core areas: architecture, code quality, the engineering team, and delivery risk. It also checks three things that change the price of a deal: who owns the code, which open-source licences the product uses, and what the system costs to run. Reveneau runs these reviews and also builds software. AI writes all of its code, and every change has to pass an eval suite written from the specification before it is released.

The review asks what a clean demo and a clean deck leave out: who wrote the code, and what breaks first at ten times the current usage.

What does technical due diligence check?

A technical due diligence review for a VC or PE deal covers four areas: system architecture, code quality, engineering team strength, and delivery risk. The goal is a plain-language answer to whether the product can scale, whether it can be maintained without the original builders, and whether the roadmap the founders are pitching is realistic for the system that exists today.

A security audit, a full penetration test and a line-by-line code review are separate jobs. They answer different questions, and they take more time than most deals allow. Diligence measures risk so it can be included in the price. The most useful output is often a short call where an investor can ask "so what worries you" and get a direct answer, with the written report behind it.

Architecture: will it keep working as the company grows

A product can work well in a demo and still fail under real customer usage. Reviewers look at how the system is structured, where the slowest parts will cause problems first, and whether basic decisions around data, security, and scaling were made deliberately or skipped under deadline pressure.

We follow the path of a single request. Where does it query the database, how many times, and is that number growing with each new customer or staying flat. A system where every page load runs ten queries against one shared table is fine at a hundred users and painful at a hundred thousand. That alone is no reason to reject a deal. Many early products have a slow step like this, because the founders were right to release first and make it faster later. The question for an investor is whether the fix is a two-week job or a rebuild, and whether the team knows which one it is.

We also look for the decisions that are expensive to reverse. A poor choice of hosting provider can be replaced in a two-week work cycle. A data model that mixed customer records together with no clean way to separate them is a more serious problem, especially for a company that will later need to sell to enterprise buyers who ask about data isolation. The architecture review sorts what it finds into "cheap to fix" and "structural." Investors can accept a long list of the first kind. The second kind changes the price.

Code quality: can someone else maintain this

Code that only one person understands is a liability. A thorough review checks whether the codebase has reasonable structure, test coverage, and documentation, or whether critical logic is understood by only one engineer. This matters most right after a deal closes, when the team composition often changes.

Test coverage is the signal people check first, and it is useful, but a raw percentage can mislead. A codebase can report high coverage and still have no test on the one payment flow that earns revenue. So we read what the tests protect as well as how many there are. The better signal is often less obvious: can a new engineer clone the repository, follow the setup notes, and get the product running locally in an afternoon. When the answer is no, that same difficulty will slow every hire the company makes for the next two years.

More early codebases now contain large amounts of AI-generated code, and that changes what "only one engineer understands it" means. The code may run and look tidy, yet no human on the team can fully explain why a given module does what it does. That is a maintainability risk even when no founder has left, because nobody on the team ever held that knowledge. We check for it by asking the team to explain the parts of the system they trust least. A team with a clear answer knows where its risk is. A team with no answer is the warning sign.

Team: who actually does the work

Due diligence should identify who on the engineering team does the real work, who reviews it, and how dependent the company is on any single person. A team with one strong technical co-founder and no other experienced engineers has a different level of risk than a team with several senior engineers who can each own a critical system.

The version-control history shows this better than any org chart. When most meaningful commits over the last year come from one account, the company has a key-person risk that no term sheet mentions. That alone is no reason to reject the deal. Plenty of strong companies depend on one exceptional builder in the early days. But an investor should know it, include it in the price, and plan for it, because the plan after closing usually gives that person a bigger role.

We also look at how work moves through the team. Is there code review, or does everything go straight to production. Is there a way to roll back a bad release, or does a failure mean a rushed, unplanned fix. These answers decide whether a team can add its next five hires without problems or slows down as it grows. A healthy early team can explain, in one or two sentences, exactly what happens when something breaks at 2am.

Delivery risk: is the roadmap realistic

The roadmap in a pitch deck is a plan. A technical review checks that plan against the team's actual track record and the codebase's actual state, so an investor knows whether the next twelve months of promised features are achievable or optimistic.

The gap to look for sits between the roadmap and the existing system it depends on. A deck promises three major features and an enterprise tier by year end, but the current system cannot support more than one customer's data cleanly, or has no way to add user permissions without touching every screen. The features can be built, but they depend on work that is in neither the plan nor the timeline. A good review makes that hidden work visible, so the investor prices the roadmap the company can deliver.

The careful assessment usually sits somewhere in the middle. Rarely is a roadmap completely unrealistic, and rarely is it free of risk. What an investor needs is the honest breakdown: which promised features are straightforward given the team and the code, which ones carry real technical unknowns, and which ones silently assume a second engineering team the company has not hired yet.

Does technical due diligence check who owns the code and its open-source licences?

Yes. Beyond the four core areas, a review checks whether the company owns its code, whether its open-source licences fit how it sells the product, and what the system costs to run. Each one can change the price.

Who owns the code. Under US copyright law, the author of a work is ordinarily the person who created it. "Works made for hire" are the exception: work an employee creates as part of the job belongs to the employer. For commissioned work, the US Copyright Office's Circular 30 (revised August 2024, read 4 November 2026) lists nine kinds of work that can be made for hire under a signed written agreement, and computer programs are not among them. So a company that wants to own the code an outside contractor or agency wrote needs a transfer of the copyright, and 17 U.S.C. 204(a) (read the same day) says a transfer "is not valid unless" it "is in writing and signed by the owner of the rights conveyed". A review asks for that signed agreement from every contractor and agency that wrote code, and from each founder for code written before the company employed them. The guide on who owns code a partner or an AI writes covers the same rules for a company that is about to hire a partner.

Open-source licences. Each open-source package in a product carries a licence, and the licence decides what the company must do in return. Section 5(c) of the GNU GPL version 3 (read 4 November 2026) says that a company giving others a modified version "must license the entire work, as a whole, under this License to anyone who comes into possession of a copy." The licence counts any step "that enables other parties to make or receive copies", and it leaves out running a program or changing a private copy. The GNU Affero GPL version 3 (read the same day) adds a duty for software used over a network: its section 13 says a modified version "must prominently offer all users interacting with it remotely through a computer network" a way to get its source code. So a review lists the licence of each package and checks it against how the company delivers its product: an app installed on customers' machines, a hosted service, or both. The guide on open-source licence due diligence explains how to run that check.

Running costs and security. A review reads the hosting bills against usage, to see whether the cost of serving each customer rises or falls as the company grows; the guide on cloud cost due diligence explains what to ask for. It also checks security at the depth a deal needs: how access is controlled, how customer data is kept apart, and what happened in past incidents, as set out in the guide on assessing security and compliance.

How Reveneau runs this review

We assess architecture, code quality, team, and delivery risk, plus code ownership, licences and running costs, and give you a plain-language assessment with the tradeoffs that matter, sized to your deal. None of this requires a six-week audit. For an early-stage deal a focused review can be finished in days. A seed round and a growth round deserve different depth, and paying for a deep audit on a small deal is its own kind of waste.

We also work with portfolio companies after the deal closes, adding senior teams where a company needs to act quickly on what the review found. A review that only produces a list of problems leaves the hard part undone. Following it with people who can fix the structural issues, or add experienced engineers to support a single founder, turns the finding into progress. See how we support VC and private equity portfolios end to end.

The same kind of evaluation applies before a hiring decision. See how to choose the right external development team for the founder-side version of this review.

Good diligence tells you, in plain words, exactly what you are buying and what it will take to grow it.

Related guide: Technical due diligence for investors.

One area we now assess as standard: how much of the codebase was generated, and how it was checked when it was. We ask because it changes what "only one engineer understands it" means. Generated code on its own is a neutral finding. Generated code that nobody checked shows up in diligence as a codebase the team cannot fully explain. We are direct about our own position here too: AI writes all of our code, and every change has to pass an eval suite written from the specification, which is part of why we know where to look.

Sources

Common questions

What does a technical due diligence review check?

A technical due diligence review covers four core areas: system architecture, code quality, engineering team strength, and delivery risk. It also checks who owns the code, which open-source licences the product uses, and what the system costs to run. The goal is a plain-language answer on whether the product can scale, whether it can be maintained without the original builders, and whether the roadmap is realistic for the codebase as it exists today.

Why does code quality matter for an investment decision?

Code quality shows whether the codebase can be maintained once the deal closes and the team around it changes. A thorough review checks for reasonable structure, real test coverage, and documentation, since code that only one person understands is a risk that appears the moment that person leaves or the company tries to hire people to work on it.

What does a technical review check about the engineering team?

It identifies who on the team does the real work, who reviews it, and how dependent the company is on any single person. A team with one strong technical co-founder and no other experienced engineers carries a different level of risk than a team with several senior engineers.

How long does a technical due diligence review take?

A focused technical due diligence review, scoped to the deal size, can be finished in days and still give a plain-language assessment before signing. Paying for a deep, six-week audit on a small deal wastes time on both sides, so the depth of the review should match the size of the investment, with a seed round and a growth round deserving different levels of scrutiny.

Is technical due diligence the same as a security audit?

Technical due diligence answers a narrower, faster question than a security audit, a full penetration test, or a line-by-line code review. Those exist to certify a system in detail and take more time than most deals allow. Diligence exists to measure risk so it can be included in the price, within the timeline of a deal, giving an investor a plain-language assessment of architecture, code quality, team, and delivery risk before the money is paid.

What are the biggest technical warning signs for investors?

A data model that mixes customer records together with no clean way to separate them, critical logic that only one engineer understands, and a roadmap that silently assumes a second engineering team the company has not hired yet. We sort findings into cheap to fix and structural, and it is the structural ones that change the price.

Should a technical review happen before or after the deal?

A focused review before signing gives an investor an honest assessment of what they are buying and includes the risk in the price of the deal itself. Reveneau also works with portfolio companies after the deal closes, adding senior teams to fix the structural issues the review found.

How does AI-generated code affect due diligence?

More early codebases now contain large amounts of AI-generated code, so the code may run and look tidy while no one on the team can fully explain why a module does what it does. We check for it by asking the team to explain the parts of the system they trust least. A team with no answer to that question is the warning sign.

How do you assess whether a roadmap is realistic?

We check the roadmap against the team's actual track record and the codebase's actual state, since a deck can promise features the current system cannot support. That gives an investor the honest breakdown: which promised features are straightforward, which carry real technical unknowns, and which silently depend on work, or a second engineering team, that is not in the timeline.

Who should run technical due diligence?

Someone who can read the architecture, the version-control history, and the tests, then explain the tradeoffs in plain words. Reveneau runs these reviews as a team that also builds software: the same reading of a codebase that tells an investor what they are buying is the reading that tells an engineering team what it has to fix.

Does technical due diligence check who owns the code?

Yes. The US Copyright Office's Circular 30 lists nine kinds of commissioned work that can be made for hire, and computer programs are not among them, so code from an outside contractor ordinarily belongs to the person who wrote it. A transfer of copyright must be in writing and signed, so a review asks for those signed agreements.

Does technical due diligence check open-source licences?

Yes. Each open-source package carries a licence, and some licences, such as the GNU GPL version 3, require the whole work to be licensed under the same terms when the company distributes a modified version. The GNU Affero GPL version 3 also requires a modified version to offer its source to users who reach it over a network.

Start a project