A license forecast that is off by 15 percent in either direction costs real money: over-forecast and you carry shelfware for the full term, under-forecast and true-up bills plus shortfall penalties add 10 to 20 percent to the contract. Forecasting is the quiet discipline that decides whether every other negotiation lever works. The discount you negotiate is applied to a quantity, and that quantity comes from the forecast. Get the forecast wrong and a good discount on the wrong number still overspends.
Why the forecast is the foundation
Every commitment-based software contract, from a Microsoft Enterprise Agreement to an Azure consumption commitment to a SaaS seat block, prices a quantity over a term. The buyer commits to that quantity and pays for it whether or not it is used. The forecast sets the quantity. A forecast built on optimism inflates the commitment and creates shelfware. A forecast built on last year's number ignores growth and triggers in-term true-ups at undiscounted rates. The forecast is the negotiation's center of gravity, which is why our software contract negotiation guide treats it as the first step rather than the last.
The demand drivers that move license counts
License demand is driven by a small set of measurable factors, and five of them account for most of the variance. Headcount change is the largest for per-user licensing. Project pipeline drives demand for specialist tools and for cloud capacity. Mergers and divestitures move counts sharply in both directions. Cloud migration converts fixed license demand into variable consumption. Product adoption curves, such as a phased Copilot rollout, ramp demand over quarters rather than all at once.
| Driver | Affects | Forecast input |
|---|---|---|
| Headcount plan | Per-user subscriptions, CALs | HR hiring and attrition forecast by quarter |
| Project pipeline | Specialist tools, cloud capacity | Portfolio roadmap with go-live dates |
| Cloud migration | Consumption commitments | Workload migration schedule |
| M&A activity | All metrics | Corporate development pipeline |
| Adoption ramp | New products (Copilot, analytics) | Phased rollout plan |
Forecasting methods
Three methods cover almost every program in practice, and mature teams blend them. Top-down forecasting scales last year's number by a growth rate, which is fast but blunt. Bottom-up forecasting builds demand from the drivers above, business unit by business unit, which is accurate but data-intensive. Driver-based modeling links license demand to a leading indicator such as headcount or transaction volume, which lets the forecast update automatically as the indicator moves. The strongest approach builds bottom-up for the next 12 months and uses driver-based modeling for the back end of a three-year term.
A working forecast model
A defensible forecast separates three layers: committed baseline, expected growth, and a small confidence buffer. The baseline is what you will use for certain. Growth is the demand the drivers predict. The buffer is deliberately small, because the contract should commit the baseline and most of the growth, leaving genuine upside to be added through true-up at a pre-negotiated rate rather than over-committed up front.
| Component | Year 1 | Year 2 | Year 3 |
|---|---|---|---|
| Committed baseline (seats) | 10,000 | 10,000 | 10,000 |
| Forecast growth | 800 | 1,700 | 2,400 |
| Commit at signature | 10,500 | 11,200 | 11,800 |
| Add via pre-priced true-up | 300 | 500 | 600 |
The principle is to commit the high-confidence portion and reserve the uncertain portion for pre-priced true-ups, so the buyer never pays for capacity that does not arrive while still locking the discount on the growth that does.
The true-up alignment lever: The most expensive forecasting error is committing to forecast growth as baseline. If growth does not materialize, the committed seats become shelfware that cannot be removed until renewal. The disciplined move is to commit only the high-confidence demand and negotiate a fixed true-up price for the rest, so unrealized growth costs nothing and realized growth is added at the locked rate. Estates that adopt this split cut committed shelfware by 15 to 25 percent without losing any discount on real growth.
Measuring forecast accuracy
A forecast that is never scored never improves, so the strongest programs measure accuracy each quarter against actuals. The standard measure is the absolute percentage error between forecast and actual demand by product line. Tracking it builds a feedback loop: products that consistently over-forecast get a tighter commitment next cycle, and products that under-forecast get a larger buffer or a pre-priced true-up. The accuracy band also tells the negotiator how much buffer is justified, because a product line that forecasts within 5 percent needs almost no buffer while one that swings 25 percent needs a true-up mechanism rather than a committed quantity.
| Forecast accuracy band | Interpretation | Commitment approach |
|---|---|---|
| Within 5 percent | Stable, predictable demand | Commit to forecast, minimal buffer |
| 5 to 15 percent | Normal enterprise variance | Commit baseline, pre-price growth |
| 15 to 30 percent | Volatile demand | Commit conservative floor, true-up the rest |
| Over 30 percent | Unpredictable | Avoid long commitment, use flexible terms |
Scenario planning
A single-point forecast hides the risk, so build three: a base case, an upside case, and a downside case. The base case is the expected demand. The upside case captures an accelerated project pipeline or an acquisition, and tells you what true-up headroom to negotiate. The downside case captures a hiring freeze or a divestiture, and tells you how much committed quantity is at risk of becoming shelfware. Sizing the commitment to the downside case while pre-pricing the path to the upside case gives the buyer protection in both directions. This is the same discipline that governs cloud commitment sizing, where the downside case prevents a shortfall penalty.
Forecasting cloud consumption
Cloud consumption commitments carry a different risk than seat licensing: unspent commitment is usually forfeited, and falling short of a commitment can trigger penalties of 5 to 15 percent of the unspent amount. Forecasting cloud spend means mapping the workload migration schedule to a consumption curve and sizing the commitment to the conservative end of that curve, then capturing additional discount on overage through a separate mechanism. The consolidation effects on the curve are covered in our cross-vendor consolidation guide, and the discount mechanics in credits and incentives negotiation.
Forecasting for renewals versus new purchases
The forecast input differs by deal type. A renewal forecast starts from actual deployment and usage data, so the question is how much of the existing estate is genuinely needed and how it will grow. A new-purchase forecast has no usage history, so it leans on driver-based modeling and on benchmarks from comparable estates. The common error is forecasting a renewal from the prior committed quantity rather than from actual usage, which carries shelfware forward into the new term. Grounding the renewal forecast in usage data is the same baseline discipline that drives a true-down, and it ties directly into the budget optimization cycle.
Governance and tooling
A forecast is only useful if it is owned and refreshed. The strongest programs assign the forecast to a named owner in IT sourcing, refresh it quarterly against actual deployment, and reconcile it to the license baseline used for compliance. Software asset management tooling helps by providing the deployed-versus-owned data that grounds the forecast, and the same data feeds the compliance checklist and the renewal plan. The forecast, the budget, and the renewal should run on one set of numbers, not three.
Common forecasting errors
Four errors recur across enterprise estates. The first is anchoring to last year's number and applying a flat growth rate, which ignores the real drivers. The second is committing to optimistic growth as baseline, which creates shelfware when the growth does not arrive. The third is forecasting seats without forecasting the access path, which misses indirect and external users. The fourth is treating the forecast as an annual budget exercise rather than a living model refreshed each quarter. Avoiding all four turns the forecast from a compliance formality into the foundation of a quantity-led negotiation.
Connecting the forecast to the budget
A forecast that does not reconcile to the budget is an academic exercise, so the two should be built from one set of numbers. The license forecast produces a quantity by product and term; the budget converts that quantity into committed and variable cost at the contracted rates. When they are built separately, the budget tends to inherit last year's number plus an across-the-board increase, which reintroduces exactly the shelfware the forecast was meant to remove. Tying them together means every dollar in the software budget traces back to a forecast driver, and any variance prompts a question rather than a quiet overspend.
The reconciliation also exposes the timing mismatches that catch finance teams. Enterprise Agreement anniversaries, cloud commitment terms, and SaaS renewal dates rarely line up with the fiscal year, so a forecast that ignores contract timing produces a budget that is right in total but wrong by quarter. Mapping each commitment to its renewal date and its true-up date lets the budget show cost in the period it actually lands, which is what a CFO needs to manage cash and to plan the year. The same map feeds the renewal calendar and the negotiation timeline.
Finally, the forecast-to-budget link is what makes the savings durable. A one-time true-down at renewal reduces cost for a year, but only a forecast that is owned, refreshed, and reconciled to the budget each quarter prevents the shelfware from creeping back as the estate grows. The discipline is modest, a quarterly review against actuals, but it is the difference between a saving that holds and one that erodes over the term.
Ownership and the quarterly refresh
A forecast without an owner decays within two quarters. The strongest programs name a single accountable owner in IT sourcing, give that owner access to the deployment and usage data, and schedule a fixed quarterly refresh against actuals. The refresh is not a full rebuild; it is a check of the drivers that changed, a comparison of forecast to actual demand, and an update to the commitment plan where the variance is material. Thirty minutes a quarter per major vendor is enough to keep the forecast live.
Ownership also fixes accountability when the forecast is wrong. A forecast that no one owns produces no learning, so the same errors repeat each cycle. A forecast with an owner and a scored accuracy history improves, because the products that consistently miss get a tighter commitment or a true-up mechanism next time. Over a few cycles this turns forecasting from a guess into a measured capability, and a measured capability is what lets a buyer commit confidently to the quantity that earns the best discount.
The payoff is compounding. A forecast that is owned and refreshed makes the first renewal cheaper, and because the same model carries forward, it makes every subsequent renewal cheaper too, since the estate never rebuilds the shelfware that an unmanaged forecast quietly accumulates between cycles.
Done well, forecasting turns every renewal into a quantity-led negotiation rather than a discount-led one. For a managed forecast tied to your drivers, see our SaaS license optimization and software licensing advisory services, and pair it with the vendor consolidation work where the supplier set is changing.