I've been in screening rooms where a startup founder pitched so well that the review team wanted to fund them on the spot — before checking anything. I've also seen the aftermath: six months into a cohort, the same team is foundering because the due diligence was skipped, the red flags were there all along, and nobody looked. Due diligence isn't glamorous. It's the work that keeps your cohort intact.
Most accelerators have a checklist. Most of those checklists are wrong — not in their criteria, but in their order of operations. They ask founders to prove traction before establishing whether the team can execute, or they evaluate market size before checking whether the founders actually understand the market they're claiming to address. The result is a lot of activity that looks like diligence but produces a shortlist that looks like luck.
Here's the framework we use at Brinc. It's built for high-volume programs where you can't spend three hours on every finalist. If you're doing due diligence on fewer than 50 companies, you can do more. If you're doing it on 300+, you need this.
Why Generic DD Checklists Fail at Accelerator Scale
A standard venture DD checklist assumes you're evaluating three to ten deals. You've done the sourcing, the initial screening is already done, and the companies in front of you have already cleared some bar. You're deep in a single opportunity, looking for reasons to pass — or reasons to proceed.
Accelerators do the opposite. You start with 3,000 applications. By the time you're doing due diligence, you've filtered to 200. You're looking for reasons to accept, not reasons to pass, which changes the psychology of the review. Every candidate has gotten past the initial screen, so there's an implicit bias toward yes. The checklist serves as the only counterweight to that bias.
The generic checklist fails because it treats all companies as if they require the same scrutiny in all dimensions. A pre-revenue deep-tech hardware company and a $50K MRR B2B SaaS business have wildly different DD profiles — not because one is better, but because the risk dimensions are different. The checklist should flex to your program's thesis, not apply uniformly regardless of context.
The 4-Tier Screening Funnel
We think about due diligence as four distinct tiers, each with a different purpose and a different depth of review. Skipping tiers or doing them out of order is where programs get into trouble.
Tier 1 — Auto-reject gate. This is the fastest possible pass. Does the team exist? Is it full-time? Is there any obviously disqualifying context — litigation history, co-founder disputes that are publicly visible, fundraising claims that contradict public data? We're not evaluating quality here. We're filtering out the 20% of applications that have a fundamental problem that makes everything else irrelevant. AI systems handle this tier effectively. We've seen error rates below 2% on auto-reject on well-configured systems.
Tier 2 — Quick scan. If it clears Tier 1, the next 10 minutes should tell you whether this is a real opportunity or a pitch deck in search of a product. Team background check, market size plausibility check, traction signal verification. No deep dive. You're looking for the signal that says "this is worth 30 more minutes" or "this is clearly out of scope for our thesis."
Tier 3 — Deep dive. This is where the time goes. You can only do 40-60 of these per cycle with a reasonable team. For each candidate: founder reference calls (not provided references — ones you find yourself), customer reference calls (or for pre-revenue companies, prospect discovery calls), thesis fit documentation, and a detailed financial model review if revenue exists. This is human work. Nothing automates it well enough to skip it.
Tier 4 — Partner review. Final stage before an offer goes out. By this point you have a file: the application, the Tier 3 notes, the references, the financial review. The partner review is a judgment call — does this company belong in our specific cohort, with our specific mentors, against our specific thesis? The checklist material is the context. The decision is a program-level choice.
What Each Tier Actually Checks
The temptation is to make every tier check everything. The result is a checklist so long that reviewers spend 45 minutes on a Tier 2 candidate that should have taken 10. The tiers exist precisely to avoid that dilution.
Team checks — Do the founders have direct experience with the problem they're solving? Not adjacent experience, not aspirational experience. Direct, lived experience of the pain point. We check LinkedIn history, prior company exits or relevant operators, and whether the founding story in the application matches what we can verify independently. Reference calls at Tier 3 are the real signal here — we ask references the same three questions across every candidate so we can actually compare.
Traction checks — Revenue and MRR get verified against bank statements or payment processor data when available. User numbers get verified against product data or third-party analytics access. Growth rates get checked against actual cohort data when the company has been operating long enough to have it. The most common lie at this stage is growth rate inflation — not revenue fabrication, but treating lifetime revenue as MRR or stacking one-time contracts into a recurring number. Verification takes 15 minutes if you know what to ask for.
Market checks — We validate the TAM claim by looking at how the number was constructed, not the number itself. A bottoms-up calculation with source citations is worth more than a $10B TAM pulled from a generic market report. We also check whether the founders have done primary research — customer interviews, win/loss data from sales conversations — or whether they're relying entirely on desk research. Programs that invest in their thesis typically have founders who've done original market work. Programs that haven't done that work show up in the application language.
Thesis fit checks — This is where most programs are weakest, and it's where the most consequential mistakes happen. A great team in the wrong sector for your program is a drag on both sides: the founders get less value than they could elsewhere, and your cohort mix drifts from your thesis. We score thesis fit on a defined rubric that maps to our portfolio priorities. The rubric isn't secret — it's shared with every applicant. We're evaluating fit against stated criteria, not against an unstated preference for whatever feels most exciting in a given cycle.
Where Technology Fits — And Where It Doesn't
AI handles Tier 1 and the early part of Tier 2 with high reliability. At Brinc, we've configured our screening system to auto-reject on defined criteria — active litigation, founders not full-time, claimed revenue that contradicts founding date and team size. The error rate on those specific triggers is low enough that we're comfortable letting the system handle the gate without human review. This frees your team to spend their time where it actually matters.
The part AI cannot replace — and will not be able to replace for a long time — is Tier 3. Reference calls, customer discovery conversations, and the qualitative judgment of whether a team will thrive under the specific pressures of your program require human judgment. This is where your program's value is actually created: in the deep dive where you catch the thing that would have gone wrong six months later.
The failure mode I see most often is programs that over-automate Tier 3 because it's expensive and time-consuming. They know they should do reference calls. They're under time pressure. They do a LinkedIn check and call it done. The company that seemed great on paper turns out to have a co-founder dynamic that no application form could capture. You find out three months in. The fix is to protect the Tier 3 time ruthlessly. If your program cannot do reference calls on every finalist, you're accepting candidates on incomplete information.
For programs running at real scale — 1,000+ applications per cycle — tools like DealForge handle the automated enrichment and routing that makes the tier structure executable without a massive team. You still do the reference calls. You still make the partner-level judgment. The system just makes sure you're doing it on the right candidates.
Building Your Own Checklist
The checklist you use should reflect your program's specific thesis, not generic startup evaluation criteria. Here's the framework I recommend:
Define your minimum bar before you design the checklist. What are the two or three things that will always result in a pass or decline regardless of everything else? For us, it's team commitment (full-time, aligned equity split) and thesis fit (this company is in a sector or geography we actively invest in). Everything else is scored relative to those two anchors.
Assign weights, not just criteria. A list of criteria without weights is not a rubric — it's a menu. "Team, market, traction, product" means different things to different reviewers. "Team is 40% of the score, market is 25%, traction is 20%, product is 15%" means something consistent you can apply across every application and track over time for pattern recognition.
Document every decline with a one-line reason. This sounds administrative, but it's the mechanism that prevents cycles from starting from scratch. When you're three cycles in, the data on why you declined companies becomes a signal about what your thesis actually is versus what you say it is.
Test the checklist against your last cohort. Take your current due diligence process and apply it to the companies you accepted last cycle, before you knew how they'd perform. Does the checklist score correlate with actual outcomes? If it doesn't, you have a calibration problem — your criteria aren't actually predicting what you think they're predicting.
If you're moving from informal due diligence to a structured process — or from a generic checklist to one calibrated to your program — this is the right moment to make that change. For the full screening framework that leads into due diligence, including how to structure the early tiers so your DD time goes to the right candidates, see the screening guide we published last month. The two frameworks are designed to connect.