Long-Range Plan vs. Annual Budget vs. Rolling Forecast: Three Horizons, One Set of Numbers

The board approved a five-year plan in January that assumes 30 percent headcount growth by year three. The annual budget, locked in March, staffs this year at 18 percent growth. The rolling forecast, updated last week, is tracking to 12 percent because two reqs got frozen in a hiring slowdown nobody put in either of the other two documents. All three numbers are “current.” None of them agree, and the person who has to explain the gap to the board next month is the same person who built all three models and still can’t say, with a straight face, which one is right.

This is not a discipline problem. It is a structural one. Finance runs three planning horizons simultaneously (a long-range plan spanning three to five years, an annual budget locked for the current fiscal year, and a rolling forecast that moves every period), and almost every company builds them as three separate artifacts, usually by three different owners, in three different files. They start from the same business. They drift apart because nothing forces them to share a foundation.

Three horizons, three jobs

Each of these documents answers a different question, and confusing the questions is where the trouble usually starts.

Horizon	Time span	Question it answers	Typical owner
Long-range plan (LRP)	3-5 years	Where is this business headed, and does the strategy add up?	CFO / strategy
Annual budget	12 months (fiscal year)	What are we authorized to spend, department by department?	FP&A + department heads
Rolling forecast	12-18 months, moving	What do we currently believe will happen, updated as facts change?	FP&A

The long-range plan is directional. It underwrites fundraising conversations, board strategy sessions, and multi-year capital decisions (hiring a VP a year before revenue justifies it, committing to a facility lease, sizing a Series B raise). It is allowed to be wrong in the details because its job is to be right in the shape.

The annual budget is a commitment device. Once locked, it is the number department heads are measured against and the number that authorizes spend. It should not move every time a driver assumption shifts: that defeats the point of having a commitment at all.

The rolling forecast is the opposite: it exists specifically to move. It rolls forward every close, dropping the period that just happened and extending the horizon, so finance always has a current best estimate rather than a nine-month-old guess dressed up as one. For the mechanics of how that roll actually works period over period, see rolling forecasts on a fixed cadence.

These are legitimately different jobs. The mistake isn’t running all three, it’s building them as three unrelated models instead of three views over one reconciled set of actuals and drivers.

Where the three horizons quietly diverge

Divergence rarely happens because someone made an obvious error. It happens through a dozen small, defensible-in-isolation decisions that compound.

The long-range plan gets built once a year, often in a planning offsite, using a headcount ramp and revenue curve that made sense at the time. Nobody revisits it when the annual budget gets tighter three months later: the LRP isn’t “due” again until next January, so it just sits there, increasingly disconnected from what the company actually knows.

The annual budget gets built from a snapshot of assumptions in Q4 (a growth rate, a hiring plan, a pricing assumption) and then locked. That’s correct behavior; a budget that moves constantly isn’t a commitment. But the lock means the budget’s assumptions freeze at a point in time while the business keeps moving.

The rolling forecast updates constantly and correctly reflects the newest actuals. But it’s usually built in its own spreadsheet, by whoever owns the forecast that month, referencing whatever version of “current headcount” or “current pipeline” they happened to pull. It rarely gets reconciled back against either the budget it’s supposed to be tracking or the long-range plan it’s supposed to be feeding.

Three months in, the LRP still says 30 percent growth, the budget says 18, the forecast says 12, and each number was defensible when it was set. The failure mode isn’t that any one document lied. It’s that nothing tied them to a shared, evolving set of facts, so they drifted at three different speeds and nobody noticed until the board asked.

The dependency that gets skipped

The three horizons aren’t independent. Each one should constrain the next.

The rolling forecast should be the most current read on actuals and near-term trajectory. The annual budget, while locked for spend-authorization purposes, should be checked against that forecast every close, not to change the number, but to know how far reality has moved from the commitment, and why. The long-range plan should absorb what the annual budget and rolling forecast have learned each year, so a five-year plan built in year one isn’t still running on year-one assumptions in year three.

In practice, this dependency chain gets skipped because each horizon lives in its own file with its own driver tree. Reconciling the LRP against current actuals means someone manually pulling numbers from the forecast model into the LRP model, mapping line items that were never structured the same way, and hoping the definitions of “headcount” or “gross margin” match across both. That reconciliation work is tedious enough that it mostly doesn’t happen until someone is forced to do it, usually right before a board meeting or a fundraise, under time pressure, which is exactly when reconciliation errors are most likely and least likely to get caught.

For the underlying distinction between “budget,” “forecast,” and “plan” as concepts (a commitment versus an estimate versus a strategic frame), see budget vs. forecast vs. plan. This piece is about the time-horizon version of the same problem: three documents built on the same underlying business, drifting because they don’t share a base.

What sharing one foundation actually means

The fix isn’t collapsing the three horizons into one document: they serve different purposes and should stay distinct. The fix is having all three reference the same reconciled actuals and the same driver definitions, so “current headcount” or “gross margin” means the same number in the LRP, the budget, and the forecast, because it’s pulled from the same place rather than retyped into three models.

Concretely, that means:

• One driver tree, three horizons layered on top. The assumptions that drive revenue, headcount, and cost should live in a single governed structure. The LRP extends those drivers out five years with wider bands of uncertainty; the budget locks them for twelve months; the forecast updates them every period. All three are reading the same tree, not maintaining three copies of it.
• Actuals flow up automatically, not by re-entry. When a close happens, the rolling forecast should update from reconciled actuals without someone manually re-keying numbers. The budget stays locked (correctly), but the variance between budget and actuals should be visible without a side reconciliation project.
• The long-range plan gets refreshed against what the other two horizons now know, at a defined cadence, not just once a year in a vacuum. If the rolling forecast has known for two quarters that hiring is running below the LRP’s ramp, the long-range plan should show that gap explicitly rather than silently carrying a stale assumption into a board deck.

None of this requires merging the documents. A five-year strategic plan and a twelve-month spend authorization are different instruments and should stay that way. What changes is that every number inside them traces back to the same reconciled base, so when someone asks “why does the plan say 30 percent and the forecast say 12,” the answer is a specific, visible driver difference, not a shrug and a promise to reconcile the spreadsheets before the next meeting.

The takeaway

A long-range plan, an annual budget, and a rolling forecast are supposed to disagree a little: they’re different instruments measuring different things at different points in time. What they shouldn’t do is disagree because nobody can trace where they diverged. If your three-year plan and this month’s forecast can’t be reconciled without a multi-day spreadsheet exercise, the problem isn’t your planning cadence. It’s that the three horizons were never built on the same foundation to begin with.

Rexfin keeps one reconciled model of actuals and drivers underneath every horizon you plan against, so the annual budget, the rolling forecast, and the long-range plan are three views of the same facts instead of three competing spreadsheets. Book a demo to see how a driver change shows up consistently across all three.

For how the budgeting cycle itself gets locked and tracked, see budgeting season. For the definitional groundwork behind the long-range plan itself, see the long-range plan glossary entry.