Technical due diligence at Series B: what the audit shifts to

Technical due diligence at Series B is the point where the questions move from judgment to evidence. At seed we assess how the founders decide, as we laid out in technical due diligence at seed stage. By Series B the product exists, customers pay for it, the team has thirty or forty engineers, and the round is underwriting a plan that multiplies all of it. The audit stops asking whether the thing can be built and starts asking whether what was built can carry the plan.

Can the architecture carry three years of the plan

The Series B model usually assumes a multiple of today’s volume and at least one product line that does not exist yet. We test the architecture against that plan rather than good practice in the abstract. Which component has been rewritten once already, and which is next. Which part of the system is a single database that every feature touches. What happens to the largest customer’s workload if it doubles, and does the answer come from a load test or from hope.

The method is the scalability assessment: find the ceiling, then establish whether lifting it is a defined intervention or an unplanned rebuild. The difference is roadmap months, which are what the round is buying.

An engineering organization built to scale, or a large team

A seed team is a group of people in a room. A Series B team is an organization, and we assess it as one. We look at who owns each service, who is on call for it, and who decides when it changes; at how work flows from a customer request to production and how long it takes, measured rather than described; and at the hiring plan and whether the current managers have run a team that size.

The usual findings sit between the people and the structure. The platform still depends on two early engineers whose knowledge is not written down, a pattern we price as key person risk. Or the team has doubled and the process has not, so output per engineer has fallen and nobody noticed because the headline shipping rate held. Both need a budgeted line in the plan.

Technical debt with a price and a date

At seed we tell investors to ignore technical debt, because borrowing is the right call before product-market fit. At Series B the loan comes up for review. The question is whether the team knows what it borrowed and has a schedule for repaying it, or whether the debt is a mood that surfaces every time a feature slips.

We ask for the inventory. A good team has one, formal or not: the three components everyone is afraid to touch, the migration postponed twice, the test suite that takes an hour and gets skipped. We put a cost to fix and a cost to carry next to each item, the way we price technical debt in an acquisition, and check whether the roadmap leaves room for the work; one that assumes no repayment over three years has a quarter missing.

The AI layer, defensible and measured

Most Series B companies we audit now have an AI layer, as the product or as the feature behind the last year of growth. At seed we ask whether the thesis can be built; at Series B, what it costs to serve, whether anyone measures its quality, and what happens when the next model release absorbs the feature.

The evidence we want is operational: an eval set that runs before each change, production logs that show failure rates by use case, a cost per request that someone owns. A company whose AI feature is a prompt and a vendor key, with no evals and no view of its margin, carries a product risk and a margin risk at once. A company that can show the numbers has what the next model release will not take away, its data, its workflow, and the discipline to know when to change the model.

Security and compliance as a sales blocker

At seed, a missing security program is normal. At Series B it is a sales finding before it is a risk finding. The plan assumes enterprise customers, and procurement there asks for SOC 2, a recent pentest, single sign-on, and a data processing agreement the architecture can honor. We check which of these exist, which pipeline deals are waiting on them, and whether the architecture can deliver what sales already promised in questionnaires. What a certification proves is a separate question; what matters here is whether its absence is stalling revenue the model already counts.

The deliverable ends in a list rather than a verdict. Each finding carries a cost, an owner and a quarter, so that the scaling work, the hires, the debt repayment and the certification become the first hundred days of the investment. Private equity diligence goes further, with full inventories, contract review and an integration plan. Series B sits between the two; scoped as either, the report is too thin to protect the check or too heavy for a company still moving fast. We run it as an audit sized to the round, closer to the two-week audit than the three-day read, with a list the board can act on in its first quarter.