A SaaS budget built as last year's number plus a growth percentage is a guess wearing a spreadsheet, and in 2026 it's a guess with two new ways to be wrong: vendors repricing around AI features, and an AI subscription layer growing outside the plan entirely. The alternative is a bottom-up build from your actual stack: reconciled baseline, modeled renewals, headcount scaling, initiative spend, committed savings, and an honest buffer. This is that build, step by step.
There's a ritual that plays out in most organizations every budget season. Someone pulls last year's SaaS number, adds a growth percentage that feels defensible, distributes it across departments roughly the way it fell last time, and submits. Finance trims it slightly, everyone signs off, and eleven months later the actuals arrive somewhere entirely different, and nobody can quite explain why, because the budget was never really a model of anything. It was last year's guess, compounded.
That ritual limped along for a decade because SaaS spend grew smoothly enough for extrapolation to pass as planning. Two things broke it. First, vendors are repricing around embedded AI: copilot tiers, per-seat increases, usage-based AI add-ons, which means your existing stack inflates on renewal even if you buy nothing new. Second, a whole layer of spend, the AI subscription tail, now grows outside the plan entirely, expensed and card-billed and invisible to a budget built from last year's invoices.
The fix is to build the budget the direction the money actually flows: bottom-up, from the stack you actually run. This guide covers the inputs you need, the six-step build, how to allocate it across departments defensibly, and the reforecast loop that keeps the budget a living model instead of a January artifact.
Why SaaS Budgets Miss
Before the build, it's worth naming exactly where the traditional approach leaks, because each leak maps to a step in the fix:
The baseline is fiction. Budgets anchored on last year's budget (or on invoiced spend alone) start from a number that never included the expensed subscriptions, the tail, and the AI layer. You can't plan spend you can't see, which is why budget planning inherits everything from the visibility practice underneath it; that foundation is covered in our guide to SaaS spend management.
Renewals arrive unmodeled. The year's largest, most predictable spend events, contract renewals with known dates and negotiable terms, enter most budgets as flat continuations. Every renewal is actually a fork: rightsize down, renew flat, or absorb a vendor increase, and a budget that models none of the forks will be wrong at every one of them.
Headcount and spend are modeled separately. Per-seat costs are the most headcount-sensitive line in the company, yet SaaS budgets routinely get built before, or without, the workforce plan.
The drift layer is nobody's line item. Tail subscriptions and AI tools accumulate through the year in increments too small to budget individually and too numerous to ignore collectively. Unbudgeted, they surface as unexplained variance; the mechanics of that layer are covered in our guide to SaaS tail spend.
The Inputs: What You Need Before You Plan
A bottom-up budget is only as good as its inputs, and there are five:
- A reconciled spend baseline: current run-rate from actual financial transactions matched to applications, not invoices alone, not last year's plan
- License utilization data: seats paid versus seats genuinely used, per application, because this is what turns each renewal into a modeled decision
- The renewal calendar: every contract's renewal date, notice window, and current terms for the budget year
- The workforce plan: hiring, attrition, and reorganization expectations by department, from HR and finance
- The initiative roadmap: what the business intends to launch, and what tooling (in 2026, usually meaning what AI tooling) those initiatives require
If input 1 or 2 doesn't exist yet, build that first; a budget without them is the ritual again, with more steps.
The Six-Step Build
Step 1: Baseline From Reconciled Run-Rate
Start from what the organization is actually spending right now: the trailing run-rate from reconciled transactions, annualized. Not last year's budget (a guess), not last year's invoices (incomplete), but the number that includes the card-paid layer, the tail, and the AI subscriptions.
This single substitution, actuals for artifacts, is the step that fixes most of the traditional budget's error before any modeling happens. It's also, usefully, the step that surfaces the gap between what leadership thinks SaaS costs and what it costs, which is a conversation better had in planning season than in variance season.
Step 2: Model Every Renewal as a Decision
Walk the renewal calendar for the budget year, and for each material contract, budget one of three modeled outcomes rather than a flat continuation:
- Rightsize: utilization data supports a lower quantity or tier; budget the reduced figure and note the notice deadline as the execution date
- Flat: utilization supports current terms; budget flat but verify the contract carries no known escalator
- Increase: vendor repricing is expected (check for announced AI-tier changes, expiring promotional pricing, and usage growth); budget the realistic post-negotiation figure, not the list-price threat and not blind hope
This step is where the budget and the optimization program become the same motion: the utilization evidence that models the renewal is the same evidence that negotiates it.
Step 3: Scale With the Workforce Plan
Convert the hiring plan into seat economics. For each department, the per-employee SaaS cost profile (which tools a new hire in that function actually receives) multiplied by net headcount change gives the growth line. Two refinements keep it honest: use utilization-informed ratios rather than license-count ratios (if the sales team uses 70% of its current seats, growth hires may need fewer net-new licenses than headcount math suggests), and model attrition's reclamation explicitly, because every departure should release seats back into the pool rather than into waste.
Step 4: Add Initiative Spend, With Governance Attached
New tools for new initiatives get their own budget lines, and in 2026 this is substantially the AI line: platform contracts, team tooling, usage-based model costs. Budget them explicitly rather than letting them arrive through expense reports, and attach the governance at budget time: a named owner, a usage review cadence, and renewal discipline from month one. New spend budgeted without governance is simply next year's waste pool with a line item, a dynamic we cover in SaaS cost optimization.
Step 5: Subtract the Committed Savings
If an optimization program is running (and after steps 1 and 2, its targets are sitting in plain sight), the budget should carry its committed captures as a negative line: license reclamation, tier downgrades, consolidation completions, tail cleanup. Two disciplines apply. Budget committed captures, not identified potential, using the same honest separation your reporting already makes between the two. And route the recovered budget explicitly, most commonly into step 4's initiative line, which turns the budget document itself into the reallocation story: the stack visibly self-funding a share of the roadmap.
Step 6: Buffer for Drift, Deliberately
Reserve an explicit contingency for the layer that grows between plans: tail accumulation, mid-year tool needs, AI subscriptions that will appear no matter how good the intake process is. Sizing it from last year's actual unbudgeted drift (visible now, thanks to step 1's reconciled baseline) beats picking a percentage, and making it explicit beats the alternative, which is every department quietly padding its own lines to self-insure.
The six steps sum to a budget where every line traces to a data source or a decision, which is precisely what makes it defensible in the room where it gets approved.
Allocating It: Department Budgets That Departments Accept
A total budget still has to land on department P&Ls, and this is where most SaaS budgets generate their friction, because allocation by arbitrary averages invites every department lead to dispute their number.
The fix is allocation by explicit chargeback rules: each application's cost attributed by who holds licenses, who's actively using, or an agreed custom split, then aggregated per department. Rules-based allocation does two things a peanut-butter spread can't: it makes each department's number inspectable (dispute the rule, not the total), and it puts the cost of each team's tool choices on the team making them, which is the single strongest behavioral lever in the whole practice. Department leads who see their own utilization rates release their own unused seats without IT asking.
Keeping It Alive: The Quarterly Reforecast Loop
A budget touched once a year is a photograph; spend is a movie. The operating rhythm that keeps them synchronized:
- Monthly: budget variance review in the standing IT-finance meeting, at department granularity, alongside the operational spend metrics
- Quarterly: a genuine reforecast. Update the run-rate baseline, re-walk the coming quarter's renewals, true-up the headcount model against actual hiring, and re-size the drift buffer against actual drift
- Annually: score last year's forecast accuracy honestly, and feed the misses into next year's model. Improving forecast accuracy over cycles is the visible proof the practice has matured from explaining the past to predicting the future
The metrics that power this loop, budget variance, forecast accuracy, and the reconciliation checks underneath them, are covered in detail in our guide to SaaS reporting metrics, and the working relationship the loop runs on is the subject of our guide to CIO-CFO collaboration.
How Zluri Powers the Budget Build
Each of the six steps has a data dependency, and Zluri's platform supplies them from one reconciled system.
The baseline comes from discovery through eight distinct methods, matched against a SaaS library of 240,000+ applications, with every financial transaction recognized and mapped to its application, which is what makes step 1's run-rate include the card-paid and AI layers that invoice-based baselines miss. License-level, activity-based utilization powers step 2's renewal modeling and step 3's headcount ratios, with the renewal calendar's notice-window alerts turning modeled decisions into executed ones. Cost and Spend tracked as separate, reconciled figures keep the whole model checked against what's actually being billed, per-application chargeback rules generate the department allocations that survive dispute, and the automation layer (1,500+ workflow actions across 300+ integrations) executes step 5's committed captures so the savings line lands as actuals, not intentions. The mechanics behind all of it are documented in how Zluri handles SaaS spend management.
The practical effect on budget season: the six-step build stops being a six-week data archaeology project, because the inputs are standing state rather than annual excavation. Organizations typically deploy in 2 to 3 months, which is worth noting for anyone reading this in the quarter before planning begins.
Frequently Asked Questions
How do I plan a SaaS budget?
Build bottom-up in six steps: baseline from your reconciled actual run-rate (not last year's budget), model every renewal as an explicit decision using utilization data, scale per-seat costs with the workforce plan, add initiative spend with governance attached, subtract committed optimization savings, and reserve an explicit buffer for tail and AI-subscription drift. Then keep it alive with monthly variance review and a quarterly reforecast.
Why is last-year-plus-a-percentage a bad way to budget SaaS?
Because it compounds an incomplete number. Budgets anchored on prior budgets or invoiced spend miss the card-paid, expensed, and AI-subscription layers, treat renewals as flat continuations when each is actually a negotiable decision, and can't account for vendors repricing existing products around AI features. Extrapolation passed as planning while spend grew smoothly; it fails now because the growth is no longer smooth or fully visible.
How should renewals be handled in a SaaS budget?
Model each material renewal as one of three outcomes: rightsized down (when utilization supports it), flat (verified against contract escalators), or increased (when vendor repricing is realistically expected). Budget the modeled figure and note the notice-window deadline as the execution date, because a rightsizing that isn't executed before the notice window closes becomes a flat renewal regardless of what the budget said.
How do I budget for AI tools?
Two lines, treated differently. Deliberate AI initiative spend gets explicit budget lines with governance attached from day one: named owners, usage review, renewal discipline. Organic AI subscription drift, the tools that will arrive through cards and expense reports regardless, gets covered by an explicit contingency buffer sized from last year's actual unbudgeted drift, and contained by intake and visibility controls.
How should SaaS costs be allocated across departments?
By explicit per-application chargeback rules (licensed users, active users, or an agreed custom split), aggregated to department level, rather than proportional spreads. Rules-based allocation makes every department number inspectable and puts the cost of tool choices on the teams making them, which is what turns budget allocation from an annual dispute into a behavioral incentive.
How often should a SaaS budget be revisited?
Monthly for variance review at department granularity, quarterly for a genuine reforecast (updated baseline, re-walked renewals, trued-up headcount, resized buffer), and annually for an honest forecast-accuracy scoring that feeds the next cycle. A budget reviewed only at year-end isn't a plan; it's a prediction waiting to be graded.
.svg)
