Identifying Investor Thesis Drift in Active VC Funds

A fund's website said it backed founders decarbonizing the built environment. That was the whole pitch: clean language, a specific vertical, the kind of thesis a founder building industrial heat pumps or low-carbon concrete would read and immediately shortlist. A partner who has spoken three times in the last six months about supply chain software is showing you the thesis regardless of what the About page says. She was passed immediately. The fund's last twelve deals told a different story: four logistics and supply chain software companies, three manufacturing ops tools, two EV charging networks, two financial infrastructure plays, and one actual decarbonization company. The thesis page described a fund that, by its own recent behavior, had mostly stopped existing.

Why VC thesis pages lie, by design, not intention

A review of 8,665 active VC fund websites against eighteen months of real deal activity found that most funds show a genuine mismatch between what they say they back and what they actually fund, and the mismatches run deep, not shallow. Funds quietly stretching their stated focus understates it: their current portfolios bear little resemblance to the category on their homepage. None of this points to deception. The more accurate explanation is lag: thesis pages get written once, usually by a junior associate at launch, and then sit untouched for long stretches while the partners who actually write checks respond to deal flow, follow their winners, and answer to LPs. Picture a generalist seed fund that backs twenty companies in its first fund. A handful land in logistics almost by accident. Those few perform well, word gets around, and the fund's network starts routing more logistics deals its way. Within a year or two, the portfolio has tilted hard toward logistics while the website still reads "generalist," because nobody circled back to rewrite it. A second mechanism works in the opposite direction: funds sometimes update their public language to match what LPs are reading in the press, well ahead of any real change in what they're funding, or the reverse happens and the deal behavior shifts while the language stays frozen. A third mechanism is simpler still: a partnership tells its LPs it's moving upstream or narrowing focus, finalizes the strategy internally, and never gets around to updating the website that founders are reading. None of this is malicious. All of it leaves founders working from a document that describes a fund that no longer fully exists. A founder who spends a week customizing a deck, building warm intros, and shaping an outreach plan around a thesis page is often spending that time on a fund that stopped matching its own description months or years earlier, and the cost lands entirely on her side of the table.

The four patterns thesis drift follows

Sorting these mismatches by type shows that four patterns account for almost all of them, and once a founder knows the categories, drift changes from a vague suspicion into something she can name on sight. Stage drift is the most common pattern, responsible for 31% of all thesis mismatches according to VC Boom's analysis, and it looks like a fund whose website still says "pre-seed and seed" while its actual recent checks have landed almost entirely in Series A and B. Sector drift runs close behind: a fund's site promises a narrow category like fintech, but its recent deals cluster in logistics, proptech, and infrastructure software with a payments angle, which usually means the fund's real thesis has quietly become something closer to "software with transaction revenue," an evolution the website never bothers to spell out. The AI reorientation accelerated sector drift across the industry, as funds that once held generalist or non-AI mandates felt pressure to shift their language and their deal flow toward AI once valuation premiums for AI companies climbed sharply, producing public thesis drift that ran well ahead of what those funds were actually funding. Geography drift follows a similar logic: a fund's site claims a focus on the West Coast and Mountain West, but the bulk of its recent deals have landed in Texas and Florida, because founders have moved and websites haven't caught up. The fourth pattern, model drift, is the one founders feel most directly: a fund's site promises that it leads rounds and takes board seats, but its recent deals are mostly follow-on checks with no board involvement, often because partnership bandwidth has tightened or because the fund has decided that smaller checks into competitive rounds beat the risk of leading cold ones. A founder looking for a lead who pitches a fund in this state gets a check that looks like a lead on paper but can't anchor the round, won't set the price, and won't offer governance support. Beyond these four, a smaller number of funds show double-drift, where stage and sector shift at the same time, or full thesis abandonment, where a mid-fund strategy pivot leaves the website behind entirely. Those cases exist, but the four core patterns cover the large majority of what a founder will run into.

Which funds to pitch, skip, or re-frame for

Venture capital's current structure has turned thesis drift from an occasional byproduct of busy partners into something closer to a survival tactic, because funds that fail to reorient toward where the money is concentrating face real pressure from their own LPs to do so. Capital concentration hit an extreme in 2025: a very large share of all US venture dollars went to a tiny fraction of companies by valuation, AI premiums at Series D and later climbed to levels that dwarfed non-AI comparables, and only a small fraction of total capital reached the bottom half of the market. Funds that once ran sector-agnostic mandates now face LP scrutiny the moment their portfolio shows no AI exposure, and that scrutiny pushes them to update their public thesis language toward AI even when their actual check-writing hasn't caught up yet. A parallel dynamic, sometimes called the "SaaSpocalypse," the expectation that AI will gut traditional SaaS business models, reshaped the strategy and public language of a large part of the industry, generating a wave of website-level thesis updates that moved faster than the investment decisions behind them. Sequoia offers the clearest large-scale illustration: its portfolio shifted heavily toward early-stage deals in the AI era, yet its rate of leading those investments dropped at the same time. A firm's seed-stage thesis language can describe a fund that shows up to rounds without leading them. A check like that doesn't anchor a round, doesn't set the price, and doesn't send the signal of conviction that other investors look for when deciding whether to join. If anything, it can complicate a raise by suggesting the big name passed on leading. Multistage firms more broadly have picked up real pricing power at the seed stage while deploying wide early-stage coverage with less intent to lead, which reads as thesis alignment on paper but behaves very differently once a founder is trying to build a cap table around it. Meanwhile, the one channel that used to give founders a reliable alternative, the emerging manager market, has shrunk fast: first-time fund formation fell from 461 closed funds in 2021 to just 44 in the first half of 2025, with 106 for the full year, a fraction of what the market looked like in 2021. Fewer new funds mean fewer fresh, clearly stated theses entering the market, leaving founders with fewer easy, legible targets to pitch.

Reading what a fund backs: a four-step behavioral audit

The only reliable account of what a fund is actually backing right now is its recent deal activity, not its thesis page, not its "about" section, and not the blog posts its partners write. Building that account takes tagging every mismatch by type, which reveals that four patterns account for the large majority of cases, so founders can recognize drift before it wastes their time rather than after. The first step is pulling the fund's recent portfolio, not its full history, because older deals say nothing about what the fund is writing checks for today. Filtering by recently closed deals through whatever investor database is on hand gets a founder most of the way there, and checking the thesis page's own age, through page source metadata, a "last modified" tag, or the Wayback Machine, tells her how long that page has gone untouched. If a thesis page hasn't moved in years and sits next to a portfolio that's clearly moved on, you should treat it as stale information, not a live description of the fund. The second step is tagging each recent deal by stage, sector, geography, and role, meaning whether the fund led the round or simply joined it. This step only works if a founder ignores the fund's own labels and pulls the real round type, the real check size range, and the real answer to who led. The third step is cross-referencing that pattern against the published thesis to name the type of drift at work, running the four-part taxonomy from the previous section against what's been tagged: is the gap in stage, in sector, in geography, or in deal role. Each answer points to a different response. If a fund shows stage drift, it usually can no longer write a check the size a founder needs. If a fund shows model drift, it usually can't anchor the round, even if it writes a check. One shortcut makes this step faster: specialist funds with genuinely narrow mandates, something like "only AI infrastructure" or "only femtech," show almost no drift, because their lane is too tight to wander out of, and their thesis pages tend to be some of the most accurate on the market. The fourth step is checking the GP's recent public signals: conference talks, LP letters where they're public, posts on Twitter/X, and portfolio company announcements. Funds typically rewrite their published thesis only every three to seven years, timed to a new fund raise, but individual partners tend to talk about where their head actually is well before the website catches up. A partner who has given three talks on supply chain software in the last six months is telling a founder exactly where the fund's attention sits, regardless of what the About page still claims. New fund announcements carry the clearest signal of all: a fund raising a Fund III under a new name or a visibly different focus is almost always describing a real shift in mandate, and the LP pitch deck behind that raise, sometimes visible through press coverage, lays out what the fund is actually telling its own investors it will do.

Pitch, skip, or re-frame: what the audit reveals

Running this audit on a real list produces three distinct groups: funds to pitch exactly as their website describes them, funds to remove from the list entirely, and funds worth re-approaching with a different frame, and that third group is where outreach starts converting again. One founder ran this audit against a list of eighty funds and found that more than half showed meaningful drift. She re-pitched nineteen of them with messaging built around what they'd actually funded in the past year instead of what their website claimed, cut thirty-three from the list outright, and watched her response rate climb enough to close her round in eight weeks. The mechanism behind that jump is straightforward: a pitch built around a fund's real recent behavior tells the partner reading it that the founder did actual homework, which sets it apart from the flood of generic pitches quoting the fund's own thesis page back at it. Every email sent to a fund that's moved two stages up or drifted two sectors sideways is outreach that can't convert and that still eats into the limited bandwidth available for following up with funds that can. For funds showing stage drift, the decision tends to be binary: if a fund's recent checks run a stage above where a founder is raising, it comes off the list, because no amount of reframing changes what the fund's fund economics will allow it to write. The one exception is a fund whose recent portfolio shows it taking small follow-on positions at earlier stages purely to protect existing bets. That fund might still write a small check to keep access to the company, but it won't lead. Sector-drifted funds are where the real opportunity sits: if a fund's actual, current thesis has moved toward something adjacent to a founder's category, the move is to pitch that evolved thesis directly rather than the one still published on the site. A founder running a logistics SaaS company with a payments angle, pitching a fund whose website says "fintech" but whose recent deals cluster around software with transaction revenue, should lead with the transaction revenue story and leave the logistics framing in the background. Funds showing model drift, the ones that say they lead but have mostly been following lately, call for a different kind of honesty: a founder who needs a lead investor should stop counting that fund in the lead column and instead weigh whether a smaller check from a recognizable name still carries enough signaling value to other investors to be worth taking.

AI screening and the cost of pitching the wrong thesis

Many funds now run AI screening on incoming pitches before a partner ever opens the deck, and a pitch built around a fund's stale, published thesis can fail that automated filter before it has any chance to reach a person who might read it differently. That shifts behavioral thesis-matching from a nice-to-have into something close to a prerequisite just to get a human in the loop at all. The tooling behind this has moved fast: basic fit filters in 2024 gave way to alignment scoring in 2025, and by 2026 funds are running closed predictive models that match pitch decks, founder backgrounds, and business models against what a fund has genuinely backed, not against the language sitting on its website. Ivan Nikkhoo, Managing Partner at Navigate Ventures, said these tools can screen and score pitches at a scale and speed no human reviewer could match. The audit described in this piece, pulling recent deals, tagging them, cross-referencing against the published thesis, and checking public signals, is the same information an AI model is scoring a founder's materials against, on the other side of the submission. Platforms such as Metal, an investor intelligence and pipeline platform for startups that parses real deal activity and partner signals to show founders which funds are actually writing checks in their category, exist precisely to close this gap before a pitch goes out. Why the current market produces drift faster than founders realize