Venture capital is usually described as a business of picking. The romance lives in the initial check: the founder meeting, the conviction bet, the contrarian call. Yet in most funds, a substantial share of capital is never deployed at first contact. It sits in reserves, earmarked for companies already in the portfolio. Where that money goes is decided with less ceremony, and often less rigour, than the original investment. Even so, that decision can determine whether a fund merely returns its capital or multiplies it.

The structural logic of staged financing is well established. Research on venture contracting, from Paul Gompers's analysis of staging to Josh Lerner's broader work on the venture cycle, shows that investors release capital in tranches to preserve the option to abandon. Each tranche turns uncertainty into a sequence of observable milestones, so that learning can discipline capital. In practice, the follow-on decision often inverts that logic. Reserves get deployed to defend positions, signal loyalty, or avoid admitting a mistake. They are rarely used to buy the next increment of information at a fair price.

This gap between design intent and actual behaviour is not a matter of individual weakness. It is an ecosystem problem, shaped by fund economics, syndicate dynamics, and the signalling externalities that insiders impose on outside capital. Addressing it requires three things. Investors need an honest assessment of what insiders actually know and a clear view of how prior commitment corrupts judgement. They also need a reserve architecture that treats each follow-on as a fresh investment decision.

The standard justification for follow-on investment is informational. Insiders sit on the board, read the monthly metrics, and know how the founders behave under pressure. Compared with a new lead who has done six weeks of diligence, the existing investor appears to hold a decisive edge. Akerlof would recognise the setup: an information asymmetry that should let the informed party buy good assets and avoid lemons. If that advantage were reliable, insider follow-ons would systematically outperform.

The reality is more conditional. Insider information is deepest on operational execution: burn discipline, hiring velocity, the founder's response to adversity. It is often shallowest on the variables that drive outcomes at later stages, such as how much of the market is actually captured, how competition evolves, and how the next round will be priced. Board proximity yields a granular view of the trees and can hide the shape of the forest. An outside lead benchmarking twenty comparable companies may understand relative positioning better than an insider who knows one company intimately.

There is also a contamination problem. Insiders typically receive information filtered through the founder, who has every incentive to present momentum. The relationship that creates access also creates dependency. Directors who championed the original deal become participants in the narrative rather than auditors of it. The advantage is therefore real but asymmetric. It is strong for detecting fraud, dysfunction, or a founder losing conviction. It is weak for judging whether a company has earned a valuation step-up.

This is why a credible outside lead matters so much. When an independent investor sets the price, the insider can pair private operational knowledge with a market-tested valuation. That is the configuration in which insider information genuinely improves decisions. When insiders price their own round without outside validation, the incentive to mark up the position is most likely to overwhelm the information advantage. Lerner's research on syndication suggested that seeking a second opinion is itself a way to improve selection, not merely to spread risk.

The practical discipline is to disaggregate. Before any follow-on, a partner should state which specific private information changes the decision. The partner should also say whether that information concerns execution or market outcome. If the edge reduces to 'we know the team is strong,' it is a weak signal that any competent outside lead has already priced. If the edge is 'we see retention cohorts the market has not yet observed,' that is an advantage worth acting on.

Takeaway

Insider knowledge is most valuable for detecting what is wrong and least reliable for judging what something is worth. Before doubling down, ask which kind of information is actually driving the decision.

Barry Staw's classic research on escalation of commitment found that people who were personally responsible for an initial investment were more likely to put more resources into a failing course of action than people who had inherited it. Venture capital is almost a laboratory for this effect. The partner who sourced the deal, argued for it at the partner meeting, and joined the board is usually the same partner who advocates for the follow-on. Responsibility and judgement are concentrated in one person.

Prospect theory sharpens the problem. When a company is struggling, the follow-on gets framed against the original cost basis as a reference point. That framing places the investor in the domain of losses, where risk-seeking behaviour becomes more likely. A bridge round becomes a chance to get back to even rather than a fresh bet with independent expected value. The question quietly shifts from 'is this a good investment at this price?' to 'how do we avoid writing this down?'

Fund structures amplify these tendencies. Interim marks influence fundraising for the next vehicle, so protecting a portfolio company's valuation has reputational value independent of its economic merit. Signalling risk adds another layer. When a known insider declines to participate, outside investors may read the absence as negative private information, which can impair the company's ability to raise at all. Insiders therefore feel pressure to participate defensively. Reserves become a form of reputational insurance rather than an allocation aimed at returns.

The contamination runs in the other direction as well. Investors who have been burned may under-invest in genuine breakouts. They anchor on the low entry price and treat a sharply higher valuation as evidence of overheating. Here the sunk cost acts as a ceiling rather than a floor. Some of the most costly errors in venture come from declining to buy more of the rare company that is compounding, simply because the new price feels expensive next to the original. Throwing good money after bad is not the only mistake.

Structural remedies tend to work better than exhortation. Some firms require a partner other than the deal sponsor to lead the follow-on analysis. Others evaluate every follow-on as though the firm held no existing position, then assess separately whether maintaining ownership has strategic value. The aim is to separate the investment question from the relationship question, so that each can be answered on its own terms.

Takeaway

Sunk cost bias in venture cuts both ways. It pushes capital into companies that should be released and holds it back from companies whose rising prices feel uncomfortable, and both errors come from judging the future against what you already paid.

Reserve strategy begins at fund construction, not at the moment of the follow-on. A fund that commits to defending pro rata across its entire portfolio is implicitly betting that outcomes are evenly distributed. That bet contradicts the power-law structure of venture returns. A small minority of companies typically generates most of a fund's value. Reserves are therefore most productive when concentrated on emerging outliers rather than spread across the portfolio as entitlements.

A useful framework treats reserves as a portfolio of options, each with explicit exercise criteria. At the initial investment, the firm models a reserve envelope for each company based on its expected financing path. That envelope is a ceiling, not a commitment. Whether the option is exercised depends on milestones defined in advance, such as product-market evidence, unit economics, and competitive position. These milestones should be set before the company approaches its next raise, while the partner's judgement is not yet exposed to the pressure of a live round.

Right-sizing is the second discipline. The relevant question is not whether to maintain pro rata but what the marginal dollar earns in this company compared with its next-best use. The alternatives include another follow-on, a new initial investment, or capital returned to limited partners. Each follow-on should be expressed as an expected multiple on the incremental cheque rather than on the blended position. This removes the averaging effect that makes weak follow-ons look acceptable.

Staging can also happen inside the firm's own process. Rather than making a single binary call at each round, firms can tier their participation:
• Super pro rata for companies that clear a high bar.
• Pro rata for strong performers.
• Partial participation where signalling risk is meaningful but conviction is moderate.
• A deliberate pass where the incremental economics are poor.
Communicating these tiers openly to founders and co-investors lowers the signalling cost of a pass, because the pass then reflects stated policy rather than an implicit verdict.

Finally, reserves should be dynamic. Regular reallocation reviews release reserves from companies that have stalled and redirect them to breakouts, so capital does not sit idle against stale assumptions. Opportunity funds and SPV capacity extend this flexibility, though they bring concentration risk that must be governed explicitly. The objective throughout is to make the follow-on decision a repeatable process rather than a negotiation conducted under the time pressure of a term sheet.

Takeaway

Reserves are options, not obligations. The discipline lies in defining exercise conditions before the round arrives and in measuring every follow-on by what the incremental dollar earns.

The follow-on decision exposes a tension at the heart of venture practice. Staged financing was designed to let learning discipline capital. Yet the proximity that generates learning also generates attachment, reputational exposure, and signalling obligations, and these pull in the opposite direction.

Resolving that tension is a design problem. Firms that perform well on follow-ons tend not to rely on superior individual judgement. They separate sponsorship from evaluation inside the institution, insist on outside price discovery where possible, and treat reserves as a dynamic pool rather than a set of promises.

For limited partners and policymakers shaping innovation ecosystems, the implication extends beyond individual funds. Healthy ecosystems need deep pools of independent later-stage capital, because insider rounds without external validation weaken the price signals the whole system depends on. Doubling down works best when someone else is checking the math.