Sales Forecasting Pack

The Arithmetic Theater of Static Probabilities

Most sales forecasting tools are just glorified spreadsheets with a "probability" column. You take a deal worth $50k, slap a 20% close rate on it because it's in "Negotiation", and sum it up. That's not forecasting; that's arithmetic theater. The industry calls this "weighted pipeline forecasting" ^[1], but in practice, it's often a static, uncalibrated guess that ignores deal velocity, historical win rates, and the actual intent of the decision-maker ^[5]. When your CRM is a black box of stale data and your probability weights haven't been updated since Q1, your forecast is a lie. We built the Sales Forecasting Pack because we saw teams drowning in crm-setup-pack configs that were never connected to a rigorous scoring engine, leaving revenue ops to manually reconcile reality with the pipeline numbers ^[2].

Engineers know that a system without feedback loops is just a leaky bucket. Sales forecasting is no different. You export a CSV from Salesforce or HubSpot, open it in Excel, and start playing with conditional formatting. You create a "Forecast" tab, copy-paste values, and pray. This process is fragile. One misplaced decimal point, one unmerged cell, and your entire quarter's projection collapses. Worse, it's opaque. When the VP of Sales asks why a $200k deal is weighted at 10%, you have to dig through email chains or ask the rep. There's no audit trail. There's no version control. There's no schema validation. If you're relying on gut feel to predict next quarter's revenue, you're already behind. The gap between a good CRM and a good forecast is the math you apply to the data. Without a structured approach to probability mapping and scenario modeling, you're just reporting history, not predicting the future.

The Cost of a Broken Feedback Loop

What happens when the forecast breaks? It's not just an embarrassing board meeting. It's a cascade of operational failures. You hire based on a $2M forecast that turns out to be $1.4M, and suddenly you're burning cash on headcount you don't need. Or you under-hire, miss revenue targets, and lose market share. The cost of inaccuracy compounds. In B2B environments, a 30-day Weighted Mean Absolute Percentage Error (WMAPE) over 15% is considered weak, meaning your forecast is off by more than 1 in 6 dollars ^[8].

When your forecast drifts that far, executive trust evaporates. You start getting asked for "scenario planning" every Friday, which usually means someone is trying to find a number that makes the monthly report look green. This is the "sandbagging" problem. Reps learn that if they over-promise, they get squeezed; if they under-promise, they get a bonus. The pipeline becomes a game of chicken, not a reflection of business reality. Without a structured approach to accuracy metrics and bias correction, you're flying blind. You might have a sales-pipeline-pack to track the deals, but if the math behind the prediction is broken, the tracking is useless. You end up with "shadow forecasts" in private spreadsheets that contradict the official CRM numbers. This creates a culture of distrust. Engineering teams can handle a system that fails fast; sales teams often hide failures until it's too late. The result is a reactive organization that spends more time explaining why the forecast was wrong than fixing the underlying issues.

The financial impact is real. High-performing forecasters are 2.5 times more likely to exceed their revenue goals than low-performers ^[4]. The difference isn't better reps; it's better data hygiene and rigorous statistical methods. When you ignore the weighted pipeline methodology, you're essentially gambling with your company's runway. Every dollar of variance is a dollar you didn't plan for. In a tight market, that variance can be the difference between growth and stagnation. You need a system that moves from intuition to data, where every probability is backed by a measurable attribute and a historical win-rate.

A SaaS Team's Forecasting Wake-Up Call

Imagine a mid-market SaaS company with 150 reps and a 12-month sales cycle. Their CRM has three stages: "Qualified", "Proposal", and "Negotiation". The VP of Sales assigns static probabilities: 20%, 50%, and 80%. On paper, the pipeline looks healthy. But in reality, deals are getting stuck in "Proposal" for six months, and the "Negotiation" stage is full of low-intent deals that never close. This is a classic failure of stage-weighted forecasting ^[7].

The team ignores the fact that historical data shows only 10% of deals in "Proposal" actually move to close, regardless of the 50% weight. They also fail to account for the decision-maker's budget cycle, which is tied to the fiscal year-end, not the deal age. Without a mechanism to apply Multi-Criteria Decision Analysis (MCDA) weights for attributes like budget, authority, and competitor presence, the forecast remains biased. The team ends up reporting $5M in weighted revenue to the board, but only $3.2M books. The gap isn't a "variance"; it's a structural flaw in the scoring model.

They needed a system that could ingest the raw pipeline, apply dynamic probabilities based on historical win-rates, and run scenario models to show what happens if the economy shifts ^[4]. Using a weighted pipeline approach ^[6], they could have multiplied each deal's value by a probability derived from its stage, historical win rates, and the decision-maker's intent ^[5]. This tool allows you to measure the expected revenue in your sales pipeline ^[3]. But instead, they relied on a static matrix that didn't account for the nuances of each deal. The turning point came when they implemented a rigorous scoring framework. They started tracking "stage progression" and "bias correction" ^[2]. They realized that deals with a "Champion" in the buyer's organization were 3x more likely to close than those without. They also found that deals involving legal review had a 20% longer sales cycle. By incorporating these factors into their forecast, they reduced their WMAPE from 25% to 12% in two quarters. The forecast became a tool for decision-making, not just a number to report. They could now answer questions like "What if we lose our top 3 reps?" or "What if the market slows down by 10%?" with data-backed scenarios.

Engineering Rigor for Revenue Operations

Once you install the Sales Forecasting Pack, the guessing game ends. The skill orchestrates an end-to-end workflow that treats your sales pipeline as a data science problem, not a spreadsheet exercise. The agent uses pipeline-probability-matrix.yaml to apply MCDA attribute weights, ensuring that a $100k deal in "Discovery" with a weak champion scores lower than a $50k deal in "Negotiation" with a signed LOI. The matrix includes stage proximity multipliers, budget/decision-maker/competitor weights, and fallback probabilities, so every deal is scored against a consistent, auditable standard.

The forecast-calculator.py script ingests your pipeline CSV, computes base, optimistic, and pessimistic scenarios, and outputs a validated JSON report that passes strict schema checks. You get accuracy-metrics.md baked into the workflow, so you're tracking WMAPE, MAE, and bias, not just "weighted revenue". The executive-report-schema.json ensures your reports are machine-readable and consistent, ready for downstream tools like demand-forecasting-ml-pack or sales-enablement-pack to consume. The validator validate-forecast.sh checks CSV headers, runs the calculator, validates JSON output against the schema, verifies numeric bounds, and exits non-zero on any structural or logical failure. This means bad data never makes it to your executive dashboard.

You stop arguing about probabilities and start analyzing drift rates and stage progression. The result is a forecast you can actually audit, with every number traceable back to a deal attribute and a statistical method. The pack provides the infrastructure for rigorous forecasting, including time-series methods like Holt-Winters and SARIMA for long-term trends, and MCDA for deal-level scoring. You can integrate this with cold-outreach-pack to see how outreach volume impacts pipeline velocity, or with proposal-writing-pack to analyze how proposal quality affects close rates. The key is that everything is automated, validated, and reproducible. You can run the forecast daily, weekly, or monthly, and always get a consistent, auditable result. The config/scenario-params.yaml allows you to tweak economic factor adjustments and stage-specific drift rates without touching the core logic, giving you the flexibility to model different strategic plans.

What's in the Sales Forecasting Pack

• skill.md — Orchestrates the end-to-end forecasting workflow; references all templates, references, scripts, tests, examples, and config files to guide the agent through pipeline analysis, scoring, scenario modeling, and executive reporting.
• templates/pipeline-probability-matrix.yaml — Production-grade CRM stage-to-probability mapping with MCDA attribute weights for deal scoring; defines stage proximity multipliers, budget/decision-maker/competitor weights, and fallback probabilities.
• templates/executive-report-schema.json — JSON Schema validating forecast report structure; enforces required fields (period, scenarios, weighted_pipeline, accuracy_metrics), type constraints, and numeric bounds for executive consumption.
• references/statistical-methods.md — Canonical reference on time-series forecasting methods: Holt-Winters, SARIMA, exponential smoothing, and linear regression; includes formulas, assumptions, seasonal decomposition steps, and selection criteria.
• references/pipeline-scoring-framework.md — Canonical reference on weighted deal scoring and probability mapping; covers MCDA methodology, attribute normalization, bias correction, stage progression tracking, and historical win-rate calibration.
• references/accuracy-metrics.md — Canonical reference on forecast evaluation: MAPE, WMAPE, MAE, bias, tracking signals, and calibration techniques; includes formulas, interpretation thresholds, and reporting best practices.
• scripts/forecast-calculator.py — Executable Python workflow that ingests pipeline CSV, applies stage probabilities + MCDA weights, computes base/optimistic/pessimistic scenarios, calculates accuracy metrics, and outputs validated JSON report.
• scripts/validate-forecast.sh — Programmatic validator that checks CSV headers, runs the calculator, validates JSON output against the schema, verifies numeric bounds, and exits non-zero on any structural or logical failure.
• tests/test-forecast.sh — Test runner that executes the validator against sample data, asserts exit codes, checks output schema compliance, and verifies scenario multiplier logic; fails fast on regression.
• examples/pipeline-data.csv — Realistic sample pipeline dataset with deal IDs, CRM stages, values, close dates, and scoring attributes for end-to-end testing of the calculator and validator.
• config/scenario-params.yaml — Configurable scenario parameters including base/optimistic/pessimistic multipliers, economic factor adjustments, and stage-specific drift rates for strategic planning.
• examples/expected-report.json — Reference output demonstrating a valid forecast report structure, populated with realistic weighted pipeline values, scenario breakdowns, and accuracy metrics for validation comparison.

Install and Ship

Stop guessing your revenue. Start calculating it. Upgrade to Pro to install the Sales Forecasting Pack and bring engineering rigor to your sales operations.