An executive with $30M of vested RSUs and a $2M annual spend asks a straightforward question: when do I need to sell? The honest answer involves guessing about tax rates, personal expenses, future equity grants, the company's performance, and life events the executive has not told the advisor about yet. Most firms answer by suggesting a 10b5-1 plan that sells 10% per year, which is a defensible default and often wrong.
Liquidity forecasting for concentrated holdings is where generic planning tools break down hardest. Retirement calculators assume stable income and diversified portfolios. Neither assumption holds for an executive with a concentrated position, variable vesting, and unpredictable liquidity events. Predictive models are starting to close this gap — not by predicting markets but by predicting the client's actual cash needs.
What liquidity forecasting actually models
Three interacting variables, each with its own dynamics:
Cash outflows. Lifestyle spend, taxes, planned purchases, philanthropic commitments, education costs. For most households, spending is surprisingly predictable month over month. Large, episodic expenses — home purchase, business investment, gift to children — are what break the forecast.
Cash inflows outside the concentrated position. Base salary, bonus, dividends from other holdings, distributions from alternatives, real estate income. Often more stable than people assume, though bonus timing matters.
Constraints on selling the concentrated position. Blackout windows, 10b5-1 plan terms, tax consequences of selling specific lots, 83(b) election implications, Rule 144 volume limits for control persons, optics concerns for senior executives.
The useful output is not a single number. It is a probability distribution over when liquidity is required, coupled to the optimal selling schedule given tax and constraint parameters.
LSR = (Projected cash inflows + Available liquid assets) ÷ Projected cash outflows, over a rolling 24-month window. An LSR below 1.2 triggers a structured selling recommendation. Below 1.0 triggers immediate action.
Where AI earns its keep
Most of the forecasting math is not hard. Deterministic models can project predictable cash flows adequately. Where machine learning adds value is on the irregular, judgment-laden inputs:
Life-event detection from client communications. A client mentions a child starting college next fall in a meeting. A client's spouse is looking at houses in Jackson Hole. A client asks about charitable remainder trusts out of the blue. Each is a signal of upcoming liquidity need. NLP applied to meeting transcripts and advisor notes surfaces these signals automatically and feeds them into the forecast.
Equity grant modeling. For senior executives, future equity grants are a major inflow, but grant size and timing vary with company performance and role changes. ML models trained on historical grant patterns — industry, title level, company performance — produce more realistic projections than either "assume current" or "assume zero."
Correlation modeling between position and spending. Clients with concentrated positions often have spending correlated with the position's value. When the stock is up, they buy the vacation home. When it is down, they defer. Modeling this correlation rather than treating spending as independent produces substantially better liquidity forecasts.
The output that actually changes decisions
A good liquidity model produces three things an advisor can act on:
A liquidity sufficiency trajectory. Does the client have enough liquidity through the next 24 months without selling the concentrated position? Through the next 60? The answer over multiple horizons tells the advisor whether this is a today problem or a five-year problem.
An optimized selling schedule given constraints. Given the LSR trajectory, blackout windows, tax lot structure, and any regulatory limits, when should sales happen and in what size? The output is a 10b5-1 plan draft, not a vague "sell gradually" recommendation.
Sensitivity to the parameters that matter. If the stock drops 30%, what changes? If the client adds a $500K annual spend? If the next grant is 50% smaller? The advisor should be able to stress-test in seconds during a client meeting.
| Approach | Typical output | Adaptability |
|---|---|---|
| Static rule of thumb | "Sell 10% per year" | None |
| Standard retirement calculator | Probability of running out of money at 95 | Low — assumes diversified portfolio |
| Deterministic cash flow model | Point forecast of liquidity need | Medium — handles known constraints |
| Predictive liquidity model | Probability distribution + selling schedule | High — updates with new information |
Where this breaks down
Two scenarios where even good models struggle.
Illiquid concentrated positions. A founder with a pre-IPO stake in a private company has most of the same liquidity pressures as a public-company executive, but no public market to sell into. The model can tell the client the liquidity need is coming; it cannot manufacture a transaction to meet it. Planning in this case is about secondary sales, loan-against-stock structures, and timing the IPO or sale window.
Sudden life changes. Divorce, disability, death of a spouse, unexpected business failure. None of these are in the model. The model should be rerun when they happen, not expected to predict them. Advisors who treat the forecast as fixed between meetings miss these shifts.
For firms building liquidity forecasting into their planning stack, the financial planning capability model maps liquidity forecasting against adjacent capabilities like tax planning, estate planning, and risk management — which helps scope the data integration work required before the models become useful.