VC Portfolio Monitoring: What to Flag and What to Ignore
VC portfolio monitoring should flag a change only when it could alter something the fund has to decide: whether to call a founder, what to say at the board, how much to reserve, or how to mark the position. In practice that means five kinds of flag: runway closing in on financing lead time, a sustained miss against the approved plan, a real shift in a core metric, a leadership or capital event, and a report that did not arrive. It should stay quiet on single-period noise, approved spending, large percentages on small numbers, and anything already open. The rest of this post sets out those rules, how to calibrate them, and how each flag gets closed.
Why does a smaller flag list work better?
Because repeated flags teach people to stop responding, and the fund has limited attention to spend.
The closest evidence comes from medicine, where alert overload has been measured directly. Ancker and colleagues studied 112 primary care clinicians from January 2010 to June 2013. The likelihood of accepting a reminder dropped by 30% for each additional reminder per encounter, and by 10% for each five-point rise in the share of repeated reminders. Roughly a quarter to a third of alerts were repeats for the same patient in the same year. The amount of work did not explain the result. Neither did newly deployed alerts wearing out over time.
This is an analogy, not venture evidence. It is an observational study of one electronic health record system, and a clinician's alert is not a portfolio flag. But the design lesson transfers: the problem was less the raw workload than alerts that said nothing new. A fund can act on that directly by never raising the same issue twice.
Attention is also worth protecting. Bernstein, Giroud and Townsend studied 22,986 venture-backed companies and 3,158 lead VC firms from 1977 to 2006. When a new airline route cut a lead investor's travel time, patents rose 3.1%, citations per patent rose 5.8%, and the probability of a successful exit rose by 1.4%. Their survey of 306 VCs reported a median of six visits to a portfolio company per year. The paper measures on-site involvement rather than monitoring software, and the authors caution that their survey sample is unlikely to be perfectly representative. The point here is narrower: an investor's time is a limited input that appears to matter, so a monitoring system should spend it deliberately.
What should portfolio monitoring flag, and what should it ignore?
Each signal needs a trigger, an exception and a decision it informs. If no decision follows, it belongs in a dashboard, not in a flag.
The thresholds below are working defaults, not research findings. Set your own per company.
| Signal | Flag when | Do not flag when | Decision it informs |
|---|---|---|---|
| Runway | Months of cash fall below expected fundraising time plus a three-month buffer | A financing is signed and closing is scheduled | Reserves, bridge, introductions |
| Revenue against plan | More than 20% below the board-approved plan for two periods running, and above an absolute floor | One period missed with a named timing cause, such as a contract signed days after period end | Board agenda, plan reset |
| Net burn | Above plan for two periods running with no board-approved reason | Prepayments, one-off costs or hiring the board already approved | Founder conversation, runway recalculation |
| Retention or gross margin | A sustained shift, tested as described below | One month's movement inside the normal range | Product or pricing question for the board |
| Customer concentration | Notice of loss from a customer above 10% of revenue | A planned contract end already reflected in the plan | Forecast revision, valuation file |
| Leadership | CEO, founder or finance lead leaves or gives notice | No exception | Immediate partner conversation |
| Capital events | New financing, term sheet, debt covenant problem or legal claim | No exception | Follow-on decision, valuation file |
| Reporting | A report is overdue, a metric disappears, or a definition changes | The company is inside its agreed grace period | Request to the company |
Leadership has no exception because oversight is most needed at that moment. Lerner's 1995 study found that venture capitalists' board representation increases around CEO turnover, while the number of other outside directors stays constant.
How do you set thresholds that hold up?
Set them per company at budget time, then leave them alone until the next approved plan. A threshold changed after a bad month is no longer a test.
Three calibration rules prevent most false alarms.
Tie runway to financing time, not a fixed number. Take an illustrative company with $3.0 million in cash and $250,000 of monthly net burn, which is 12 months of runway. The fund expects its next raise to take six months, so with a three-month buffer the flag fires at nine months, or $2.25 million in cash. That is three months away. If net burn rises to $300,000, the flag fires at $2.7 million, which is only one month away. Because the rule is tied to the raise rather than a fixed number of months, it moves with the burn.
Pair every percentage with an absolute floor. Monthly revenue falling from $20,000 to $14,000 is a 30% decline and a $6,000 change. For a company at that scale, a percentage rule alone will fire on one customer's invoice timing. A rule that also needs a $10,000 movement stays quiet until the change is large enough to mean something. Pre-revenue companies should skip revenue rules altogether and track cash and agreed milestones.
Require persistence before calling something a shift. Statistical process control separates ordinary variation from a change that needs investigating. A control chart guide on an NHS trust's site gives two usable rules: eight or more consecutive points on one side of the centre line is a shift, and six or more consecutive points all rising or all falling is a trend. These suit mature, stable monthly measures such as retention or gross margin. They do not suit fast-growing revenue, which has no stable centre line, or quarterly data, where eight points takes two years. For those, use the plan variance rule above.
What should happen when a report is missing?
Treat missing data as its own flag, separate from performance. Do not fill the gap.
First establish what the fund is entitled to, because a report cannot be late without a due date. Nixon Peabody's summary of the NVCA Investors' Rights Agreement describes annual statements delivered within 90 to 120 days and quarterly statements within 45 days. Morrison Foerster's guidance puts annual delivery at 90 to 180 days, and notes that typically only Major Investors receive information rights and that monthly statements are negotiated only sometimes. Terms vary, so record what each signed agreement says and ask counsel where it is unclear. Where the fund has no formal right, record the reporting cadence the founder agreed to and treat a miss as a prompt to ask.
Then apply four rules:
- Show every metric with its reporting period and its date received.
- Never carry a previous value forward without labeling it as stale.
- Keep blank and zero distinct. A missing cash figure is not zero cash, and it is not last quarter's cash.
- Flag a dropped metric or changed definition. A metric that disappears from the pack, or improves because its definition moved, deserves a question.
This is a process rule rather than a prediction. A late report justifies a question to the founder, not a conclusion about the company.
How does a flag get closed?
Every flag needs one owner, one open record per issue, and a recorded reason for closing it. Without that, flags pile up and the list stops being read.
Use this lifecycle:
- Opened. Record the company, signal, value, source document, reporting period and date.
- Verified. The deal lead or board member checks the figure against the source and adds context. False alarms should end here.
- Decided. The owner records the decision: no action, founder call, board item, reserve review or valuation note.
- Closed. Pick one reason: explained, acted on, false alarm or superseded.
Two habits keep the list useful. New data updates the open record instead of raising a second flag, which removes the within-issue repeats that the Ancker study points to. And once a quarter, count how each rule performed. A rule that fired for four straight quarters without changing a single decision should be rewritten or retired.
Closed flags also feed the quarter-end valuation file. The IPEV Board published its 2025 Valuation Guidelines on 11 December 2025. KPMG's summary says calibration is required at each measurement date and that strictly unchanged valuation assumptions should be the exception. We relied on that summary rather than the guidelines text, so confirm the requirement against the guidelines and your own valuation policy. If your fund follows them, a closed record of what changed at each company gives the valuation owner a starting point, and it supports the review workflow behind LP reporting accuracy.
Where can AI help, and where should it stop?
AI suits the mechanical steps: pulling defined metrics out of inconsistently formatted updates, comparing them with the plan, listing missing fields and drafting questions. It should not close a flag. The verification step stays with a person who knows the company, and any decision about reserves, marks or the board stays with the fund. For a broader view of the stages involved, see our guide to AI tools for VC portfolio management.
Thresholds and review logic differ by fund. Sonexa Labs builds this as a commissioned engagement through Portfolio Intelligence, one of our Tailored Solutions: a system could flag changes or missing reporting fields against thresholds your team defines, with your portfolio owner verifying context before anything is escalated.