Non-Consensus
The received wisdom of VC is that you have to be "non-consensus".
It's particularly important if you're not in one of the brand-name platforms like Sequoia, a16z, and so on. There, you can afford to be consensus: building the "conviction" to back Brett Taylor (former Facebook CTO, Chairman of Open AI) when he was starting Sierra is not hard - there probably wasn't much overthinking to do. The job is to get into the deal, and without the storied brands behind you, that is very, very difficult.
For the rest of the market, non-consensus is the only viable route.
Firstly, consensus investing isn't really actionable - at least not in the really good stuff. Even if you manage to tag into hot deals with a tiny cheque (a 'logo deal'), there is very little chance of driving fund-level returns. Secondly, the chances are that the consensus is wrong (or at least overpriced relative to the risk-adjusted return potential). Evidence abounds that the 'average' VC fund is not a great generator of returns, and by definition the average is a close cousin of the consensus:
"In the 2017 vintage, the median net TVPI through the end of Q3 is 1.76x, and the 75th percentile is 2.27x. The median fund from 2018 has a net TVPI of 1.38x, while the 75th percentile is 1.97x." (Carta)
Said another way, for every dollar invested in a median VC fund in 2017, the value of that dollar as of the end of Q3 2025 was $1.76. A dollar invested in the S&P 500, meanwhile, would be worth roughly $2.96, and that excludes dividends paid out along the way. So we can safely conclude that non-consensus has to be the way. That's the theory, at least. But it's hard.
It's hard to follow through on a hunch when a lot of very smart people have looked at the same opportunity and said no. It's particularly hard as a small fund where you need the market to come towards you before the company runs out of runway. It feels like doing star jumps on a knife edge and hoping you stick the landing every time. It is by no means comfortable.
But that, ultimately, is the job.