Built for teams choosing a forecasting model

Every forecasting model is wrong in a way. The right one is wrong in a way you can manage.

Sales forecasting models fall into four families: weighted-stage probability (default in most CRMs), bottom-up rep commit (what AEs believe will close), scenario-based (upside/base/downside with assumptions), and AI-adjusted (second opinion from pipeline signals). Each model is wrong in a way, and the right methodology for a team combines them with explicit role inputs. Strkr runs all four natively with the methodology codified rather than left to spreadsheet convention.

Start free trial See pricing
What this audience is actually dealing with

The pains that bring buyers here.

Forecasting model choice is one of the highest-leverage decisions a revenue leader makes and one of the most under-examined. Most teams inherit the model their CRM shipped (weighted-stage) and never evaluate alternatives, which leaves specific accuracy gains on the table. The six issues below are the methodology failures that drive teams to re-examine the forecasting model rather than just tune its parameters.

Weighted-stage probability oversimplifies

Every demo-stage deal is not 40 percent.

Weighted-stage probability (stage times probability) is the default in most CRMs and gives a reasonable aggregate at scale. The model misses deal-level nuance: a late-demo deal with a strong champion is not 40 percent, a demo-stage deal with a stalled champion is not 40 percent either. The aggregate is right but the deal number is misleading, which drives mistrust from AEs and managers who know the deals.

Bottom-up commit is biased

Reps optimize for the forecast conversation.

Pure bottom-up commit (AEs name the deals they will close) is biased by the incentive structure around the forecast conversation. A rep who gets celebrated for committing high and criticized for commits that miss learns to commit conservatively early in the quarter and raise commits as deals confirm. The resulting commit shape does not reflect genuine probability. Without historical calibration, the bias is invisible.

Scenario modeling lives in spreadsheets

The "high/medium/low" view is theoretical.

Scenario-based forecasting (upside/base/downside with specific assumptions) is powerful but typically lives in a separate spreadsheet with assumptions that are never updated. The scenarios drift from the live pipeline immediately. The "high case" that the board saw in April no longer reflects the deals in September. Strkr's scenario modeling runs on live pipeline with the assumptions tracked explicitly.

AI estimates lack context

The AI number appears without reasoning.

AI-adjusted forecast estimates (second opinion from pipeline signals) are powerful when the AI output includes citation to the signals that drove the adjustment. Without the citation, the AI number is a black box that AEs mistrust. The feature that converts AI estimates from novelty to operational artifact is signal transparency: "the AI estimate dropped because deal X had no champion-response in 11 days".

Model drift is invisible

The weighted-stage probabilities are 3 years old.

Weighted-stage probability settings commonly reflect the team's pipeline from 3 years ago (the last time anyone recalibrated the model). The current market conditions, product mix, and team composition no longer match the probabilities. The forecast runs on stale coefficients and the retrospective is unclear whether the miss was model drift or execution. Strkr recomputes stage probabilities quarterly from actual close data.

No methodology documentation

Each manager forecasts their own way.

Without codified methodology, each sales manager runs their own version: one uses weighted stage, one asks reps for commit, one applies a personal haircut. The roll-up mixes incompatible numbers and the VP applies a judgment haircut on top. Strkr codifies the methodology at the program level with the inputs each role contributes documented.

The four model families

Nine forecasting approaches Strkr supports natively.

Forecasting models fall into four families (weighted-stage, bottom-up, scenario, AI-adjusted) with nine specific approaches that span the common revenue-forecasting methodology landscape. The approaches below are what Strkr supports natively, with the use case for each.

Weighted-stage probability

Stage times historical conversion rate.

The baseline weighted-stage model multiplies deal value by the historical conversion rate for the stage. Strkr computes conversion rates per segment, per product line, and per quarter so the probabilities reflect current reality rather than 3-year-old defaults. The weighted-stage output is the default forecast baseline that other models adjust against.

Weighted-stage with override

Per-deal AE adjustment where the AE has insight.

For specific deals where the AE has insight that the stage probability misses (a late-stage deal with a stalled champion, a demo-stage deal with executive sponsorship), the AE applies a per-deal override. The override replaces the default probability with the AE's specific number and retains the reasoning. The per-deal override is where weighted-stage and bottom-up models combine naturally.

Bottom-up commit

AEs name the deals that will close.

The bottom-up model asks AEs to name the deals they commit to close in-quarter. The commit list is deal rather than probability-weighted. The forecast analyst applies historical calibration (per-rep commit-to-close) at roll-up to adjust for optimism bias. The bottom-up approach is strongest when the team has 4 or more quarters of calibration history.

Calibrated bottom-up

Rep commits adjusted by historical accuracy.

Calibrated bottom-up applies per-rep commit-to-close ratios to the raw commit. A rep whose historical ratio is 85 percent has their raw commit scaled by 0.85 at roll-up. The rep calibration beats aggregate haircut because it respects the rep behavior. Over several quarters, rep accuracy improves because the calibration is visible to the rep.

Scenario-based

Upside, base, downside with explicit assumptions.

The scenario-based model produces three forecasts: upside (what if top-10 at-risk deals close at 70 percent), base (expected case given current state), and downside (what if the same deals close at 30 percent). The assumption transparency supports board communication and executive escalation decisions. The scenarios run on live pipeline so they stay current.

AI-adjusted

Strkr AI second-opinion with signal citation.

Strkr AI reads each rep's pipeline and produces an AI-adjusted estimate alongside the rep commit. The adjustment considers historical commit-to-close accuracy, deal-state signals (champion engagement, response latency, stakeholder count), timeline patterns, and comparable deals from historical data. The AI output includes citation to the signals that drove the adjustment.

Time-series extrapolation

Trailing-period baseline for predictable segments.

For specific segments with predictable deal patterns (SMB with monthly deal rhythms, recurring customer expansions, specific product-line patterns), time-series extrapolation produces a baseline forecast from trailing-period data. The extrapolation is strongest for high-volume, low-variance segments and weakest for enterprise with lumpy quarterly dynamics.

Composite forecast

Multiple models combined with specific weighting.

The composite approach combines multiple models: 40 percent weighted-stage, 30 percent calibrated bottom-up, 20 percent time-series, 10 percent AI estimate. The weights reflect the team's historical accuracy by methodology. The composite often beats any single model because different models have different error patterns that partially cancel.

Monte Carlo for board-grade

Probability distribution rather than point estimate.

For specific board-grade or investor-reporting contexts where the forecast needs to express as a probability distribution rather than a point estimate, Strkr's Monte Carlo simulation produces a distribution across deal-level probabilities. The distribution supports "the quarter will land between X and Y with 80 percent confidence" statements that point estimates cannot.

When to use which model

Six team contexts with distinct model recommendations.

The right forecasting model depends on team size, deal size, cycle length, segment mix, and quarter maturity. Six specific team contexts drive distinct model recommendations with the reasoning for each.

Early stage, under 10 reps

Bottom-up commit with lightweight calibration.

Early-stage teams typically have insufficient historical data for weighted-stage calibration. The pure bottom-up commit model with lightweight manager calibration (gut-feel per-rep haircut) is operationally practical. As the team crosses 15 to 20 reps and accumulates 4 quarters of history, migrate to calibrated bottom-up with data-driven per-rep haircuts.

Mid-market, 20 to 100 reps

Calibrated bottom-up with weighted-stage sanity check.

Mid-market teams typically run calibrated bottom-up as the primary forecast with weighted-stage as a sanity check. When the two models disagree meaningfully, the disagreement surfaces for diagnostic conversation. The dual-model approach supports forecast analyst conversations with the deal evidence from both perspectives.

High-volume SMB segment

Time-series extrapolation with weighted-stage overlay.

High-volume SMB segments with hundreds of deals per quarter benefit from time-series extrapolation as the primary forecast with weighted-stage overlay for deal visibility. The segment's predictable pattern makes time-series highly accurate; the weighted-stage overlay supports deal-level AE conversations without the time-series baseline shifting weekly.

Enterprise segment

Scenario-based with deal attribution.

Enterprise segments with 5 to 20 deals per quarter benefit from scenario-based modeling with deal and assumption transparency. Each large deal gets individual scenario treatment ("if this specific deal closes at the expected amount, the quarter lands at X; if it slips to next quarter, Y"). The scenario-based approach is defensible against board scrutiny.

Multi-segment, 100+ reps

Composite with segment-specific weights.

Large multi-segment teams typically run composite forecasting with different model weights per segment. SMB segment weights time-series heavily; enterprise segment weights scenario heavily; mid-market weights calibrated bottom-up heavily. The composite approach at scale beats any single model because different segments have fundamentally different dynamics.

Public or pre-IPO

Monte Carlo for probability-distribution reporting.

Public companies, pre-IPO companies, and teams reporting to sophisticated boards or investors benefit from Monte Carlo simulation for probability-distribution reporting. "The quarter will land between X and Y with 80 percent confidence" supports financial-planning conversations that point estimates cannot. The Monte Carlo approach layers on top of the composite forecast as a reporting view.

What makes the model defensible

Six properties of a forecasting model that passes CFO scrutiny.

Forecasting models face scrutiny from CFOs, boards, auditors, and investors. Six specific properties make a forecasting model defensible against the standard challenges that drive model abandonment.

Documented methodology

The inputs each role contributes.

The methodology document names the inputs each role contributes: AE commit, manager calibration, director segment adjustment, VP executive haircut. Each layer's input is traceable so the layer accountability holds when the forecast misses. Methodology documentation is the baseline defensibility requirement.

Transparent parameters

Stage probabilities, haircuts, and weights are visible.

All model parameters (stage probabilities, per-rep haircuts, scenario assumptions, composite weights) are visible and auditable. The transparency supports challenge conversations where a parameter choice is questioned. The parameters are not hidden in code; they are configuration that the forecast analyst and CFO can review together.

Reproducibility

Same inputs produce same forecast number.

Model computation reproduces: run the same inputs twice and the number lands identical. The property distinguishes defensible modeling from recalculated estimates that drift. Standalone tools that recompute weighted probability on evolving pipeline often cannot reproduce last week's number, which is exactly when defensibility matters.

Historical back-test

The model would have predicted past quarters within X percent.

For any proposed model, Strkr runs historical back-test: "if this model had been in effect for the trailing 8 quarters, the forecast accuracy would have been X percent." The back-test supports the methodology-choice decision with data rather than vendor claims. The back-test also surfaces period edge cases that the model does not handle well.

Signal citation for AI

The AI estimate names the signals.

The AI-adjusted estimate includes citation to the signals that drove the adjustment: "the AI estimate dropped because deal X had no champion-response in 11 days and 3 of 4 comparable deals with similar response-latency patterns did not close." The citation makes AI estimates auditable rather than black-box.

Change control

Methodology changes are deliberate with version history.

Methodology changes go through documented change-control: proposed change, approval from VP of Sales and CFO, version history retained. The year-over-year accuracy comparison respects the version in effect at each point. The change-control discipline distinguishes "we deliberately changed approach" from "the methodology drifted silently".

Head-to-head

Strkr forecasting model options vs single-model CRM default.

Most CRM forecasting ships with a single model (weighted-stage probability) with limited configuration beyond basic probability settings. The comparison below is Strkr's multi-model approach against the typical single-model CRM default for a 40-to-150-rep revenue team.

Feature Strkr Single-model CRM default
Model families supported Weighted-stage, bottom-up, scenario, AI-adjusted Weighted-stage only
Per-deal override Native with reasoning retention Varies by platform
Stage probability calibration Recomputed quarterly from close data Set once, rarely updated
Per-rep calibration Historical curves across 8 quarters Aggregate accuracy only
Scenario modeling Live pipeline with assumption transparency Separate spreadsheet or BI tool
AI-adjusted estimate Signal citation retained Varies by platform; often black-box
Composite forecasting Multiple models with configurable weights Not supported
Monte Carlo distribution Supported for board-grade reporting Not supported
Historical back-test Model would have predicted past X quarters at Y% Not supported
Methodology documentation Program artifact with version history Lives in configuration screen
How teams use Strkr

How revenue teams chose their forecasting model.

The common thread across teams reconsidering their forecasting model: the inherited CRM default was no longer serving the team's accuracy requirements. The playbooks below are motions on customer accounts.

55-rep multi-segment team

Composite model with per-segment weighting.

A 55-rep multi-segment team moved from pure weighted-stage to composite forecasting with segment-specific weights. SMB segment weighted time-series heavily (65 percent), mid-market weighted calibrated bottom-up (55 percent), enterprise weighted scenario-based (50 percent). The composite accuracy improved from roughly 76 percent to 92 percent across the first two quarters on the new methodology. The segment-specific weighting respected the fundamentally different dynamics of each segment.

Enterprise team moving to scenario-based

Deal scenarios replaced weighted probability.

A 22-rep enterprise team with 8 to 15 deals per quarter moved from weighted-stage to scenario-based modeling. The weighted-stage approach had been producing accurate aggregates with wildly inaccurate quarter numbers because the enterprise segment's small-deal-count made aggregate statistics unreliable. The scenario-based approach treated each large deal individually with assumption transparency, which produced defensible forecasts for board communication.

Pre-IPO adopting Monte Carlo

Probability-distribution reporting ahead of S-1.

A pre-IPO company facing S-1 disclosure requirements adopted Monte Carlo simulation for probability-distribution reporting. The underwriters required specific confidence-interval statements on the forecast rather than point estimates. The Monte Carlo approach layered on top of the composite forecast produced the specific "80 percent confidence between X and Y" statements the underwriters needed.

New VP of Sales re-examining methodology

Back-test informed methodology change.

A new VP of Sales joining a 40-rep team ran historical back-tests of multiple methodology options in Strkr. The back-tests showed that calibrated bottom-up would have outperformed the inherited weighted-stage approach by roughly 11 percentage points of accuracy across the trailing 8 quarters. The VP changed methodology with the back-test as the evidence base, documented the change in the methodology document, and the subsequent quarters validated the back-test prediction.

AI-adjusted estimates as second opinion

Signal citation surfaced deal risks.

A team adding AI-adjusted estimates as a second opinion to the existing rep commit found that the signal citation surfaced deal risks the rep commits were missing. Deals where the AI estimate was materially lower than the rep commit frequently had signal patterns (champion response latency, stakeholder-count drops) that the AE had not weighted. The AI second opinion drove coaching conversations and the subsequent commits became better calibrated.

Every forecasting model is wrong in a way. Pick the one you can manage.

Start a 14-day trial with the forecasting model configured for your team context: weighted-stage with override for early-stage, calibrated bottom-up for mid-market, scenario-based for enterprise, composite for multi-segment, Monte Carlo for pre-IPO. Run historical back-tests to validate the model choice before committing.

Start free trial See pricing
Common questions

What buyers in this bucket ask most.

Which forecasting model should we use?

The right model depends on team size, deal size, cycle length, segment mix, and quarter maturity. Early-stage teams under 10 reps typically run bottom-up commit with lightweight calibration. Mid-market teams run calibrated bottom-up with weighted-stage sanity check. Enterprise segments run scenario-based with deal attribution. Large multi-segment teams run composite with segment-specific weights. Public or pre-IPO teams add Monte Carlo for probability-distribution reporting. The decision depends on the team's current context and the available historical data. Strkr's back-test support helps the choice by running past-quarter simulations across methodology options.

How does the weighted-stage probability calibration work?

Strkr recomputes stage probabilities quarterly from actual close data: for each stage, what percentage of deals in that stage at the start of the quarter closed in-quarter. The recomputation runs per segment, per product line, and per team so the probabilities reflect current reality. For teams migrating to Strkr with existing stage definitions, the migration tooling imports historical conversion data and computes the initial probabilities from the historical rather than vendor defaults.

What makes the AI-adjusted estimate different from black-box forecasting?

The property that distinguishes Strkr's AI estimate from black-box tools is signal citation. The AI output names the signals that drove the adjustment: "the estimate dropped because deal X had no champion-response in 11 days and 3 of 4 comparable deals with similar response-latency patterns did not close in-quarter." The signal transparency makes the AI estimate auditable, discussable, and actionable rather than a mysterious number that reps mistrust. The forecast analyst reviewing material differences between rep commit and AI estimate can route the signal evidence to the sales manager for an informed conversation.

How does Strkr handle a methodology change mid-year?

Methodology changes go through documented change-control: proposed change, approval from VP of Sales and CFO, version history retained. The year-over-year accuracy comparison respects the version in effect at each point. The audit trail retains the pre-change and post-change state so the impact of the change is analyzable rather than disruptive. For specific changes (switching from weighted-stage to calibrated bottom-up, adding scenario modeling, changing stage probabilities), Strkr runs back-tests on the proposed change so the impact estimation is data-informed.

Can we run multiple models simultaneously?

Yes. Strkr supports simultaneous models for comparison and composite. Teams can run weighted-stage and calibrated bottom-up side-by-side with the disagreements surfaced for diagnostic conversation. When the two models disagree meaningfully on a deal, the diagnostic reveals either a stage-probability drift or a rep-commit bias that merits investigation. The composite approach combines multiple models with configurable weights; the comparison approach keeps models separate for diagnostic surface.

What historical data does the model back-test need?

Back-tests work best with at least 4 quarters of historical commit and close data per rep. For teams with less history, Strkr runs team-level back-tests across the available period with specific caveats about the small-sample conclusions. For teams migrating from Clari, spreadsheet-based programs, or other CRMs with existing historical data, the migration tooling imports the historical records so the back-test runs on the team's history rather than vendor benchmarks. The back-test output includes confidence intervals around the accuracy estimate given the available sample size.

Also built for

Related audiences.

Forecast software Forecasting tool Forecast program Forecast tracker

Try it free. Bring your team next week.

No sales call, no migration consultant, no four-month implementation. Enter your card, get 14 days of the full Pro tier, cancel any time before day 14 with zero charge. Spin up a workspace, import your CSV, and have something useful before lunch.

Start free trial See pricing