§ Folded · noindex Demand Forecasting is now part of Market Intelligence — forecasting and market sizing on one surface.
Page preserved for reference · 2026-05-10
Home· Expand· Demand Forecasting
Demand Forecasting · M.12

Three numbers. One forecast.

Pipeline, history, and market signal rolled into one model. Bookings, revenue, and margin reconciled — at SKU, region, and rep level. The board sees a range, not a fake-precise point. Variance attributes back to the deal that moved the number.

POST /v1/forecast/q3 200 OK · 412ms
Q3 2026 forecast · 8 segments · 12 SKU families
$24.6M P50 bookings
P25
$22.1M
P50
$24.6M
P75
$27.2M
pipeline.weighted · enterprise
close-rate model · seg
$11.4M
renewals.expected · install_base
churn model · 8q
$8.2M
capacity.constrained · west
crew + parts
−$1.1M
What it looks like

Cone. Waterfall. Variance.

P25/P50/P75 confidence cone. Bookings → margin waterfall. Variance attribution after the quarter posts.

§ 01 · Forecast cone

Q3 2026 · $24.6M P50

Confidence band ±$2.5M
today
§ 02 · Bookings → margin

Three numbers reconciled

$49.4M pipeline → $6.2M margin
Pipeline
$49.4M
Bookings
$24.6M
Revenue
$19.4M
Margin
$6.2M
§ 03 · Variance heatmap

Q2 actual vs P50

By segment × region
E
W
C
EU
ENT
+8
−3
+4
0
MID
+2
−7
+1
+3
REN
+12
+6
+5
+9
The problem

Three forecasts. Three different numbers.

Sales has a number. Finance has a number. Ops has a number. None of them agree, and the board meeting is in 48 hours.

Without Demand Forecasting

Today's status quo

  • Sales forecast, finance forecast, ops forecast — three spreadsheets, three numbers
  • Manual roll-up from regional managers — two-week cycle, gut-check at every level
  • Pipeline coverage ratios applied as a flat percentage, regardless of segment
  • Bookings ≠ revenue ≠ margin — translation between the three is a black box
With Demand Forecasting

What changes

  • One forecast — sales, finance, ops all reading the same dashboard
  • Pipeline + history + market signal weighted by segment-specific close patterns
  • Drillable from board KPI to the deal that moved the number
  • Bookings → revenue → margin reconciled — every cut, every cohort, every quarter
Capabilities

What's inside.

Six capabilities that make this module work end-to-end. Pick any one as your starting point — they compound.

01

Multi-dimensional model

Forecast at SKU, region, segment, rep, or vertical level — and roll up consistently. No more reconciling a regional view that doesn't match the segment view.

Cube · Roll-up · Cuts
02

Pipeline reweighting

Each opp weighted by its actual close-rate signature — segment, deal size, sales motion, time-in-stage. Not a flat 30%.

Pipeline · Weight · Stage
03

Bookings → revenue → margin

All three numbers reconciled in the same model. Bookings translate to revenue via your rev-rec rules; revenue to margin via Cost Engine.

Rev-rec · Margin · Recon
04

Capacity planning

Sales capacity, fulfillment capacity, install crew capacity — modelled against the forecast. Where the bottleneck is, before it bites.

Capacity · Bottleneck
05

Scenario planning

Model the price hike, the macro hit, the competitor entry. Show the board the range, not a fake-precise point.

Scenario · Range · Risk
06

Variance attribution

When actuals come in, variance attributes back — segment underperformed, deal slipped, cohort missed. Not 'we missed by 6%.'

Variance · Attribution
The autonomous loop

Four steps. Two-hour cycle.

From inputs to a defended P25/P50/P75 in two hours, not two weeks. Conviction overrides logged. Variance closes back automatically when the quarter posts.

§ 01 · Pull

Inputs flow in

Pipeline (live CRM), 8 quarters of close patterns, market signal (Market Intelligence), capacity (ops planning).

~860ms
§ 02 · Model

Range computed

Returns P25, P50, P75 — not a single number. Assumptions and weights surfaced for review.

~412ms
§ 03 · Tune

Conviction overrides

Leadership layers known macro tailwinds or strategic ramps. Override and rationale logged.

human
§ 04 · Close

Variance attributes

Quarter posts, actuals reconcile back. Model learns. Conviction calls scored.

~24ms
Forecast policy

Weights you can read. Overrides you can audit.

Forecast policy in plain language — segment weights, capacity floors, override routing. No black-box ML. Every number on the deck shows the rule and the input that drove it.

Conviction overrides are first-class. Logged with rationale. Scored against actuals next quarter — you see whose calls were right.

policy · forecast.alm
# Demand Forecasting · Model Policy
when opp.segment = "enterprise":
  weight_by close_rate(seg, deal_size, time_in_stage)
  capacity_check  install_crew

when opp.stage = "verbal_yes":
  fast_track  p50_inclusion

when override.applied:
  require rationale
  score_against actuals(+1q)

when capacity_constraint > 85%:
  surface bottleneck  ops_lead
  cap forecast at_capacity

# Last edited: D. Yates · 4d ago
Where it lives

Forecast lives where your operators already plan.

Pipeline

CRM

Salesforce, HubSpot, Dynamics — live opportunities feed the model.

History

Data warehouse

Snowflake, BigQuery, Redshift — 8 quarters of close patterns.

Capacity

FSM + HR

Crew rosters, install capacity, leave calendar.

Output

Board + ops

Live dashboard, drill-down, export-to-deck for the QBR.

Real outcomes

"Sales said $24M. Finance said $19M. The model said $22M ±2. We landed at $22.4M."

Hannah Ko CFO · Operator A
±4%Forecast accuracy, 2-quarter horizon
2 hrsCycle time, was 2 weeks
3→1Forecasts to a single number
Operator A · case study 14 states · 8 segments · QBR-grade forecast
See it on your data

Run the forecast on your last quarter.

We import your last 8 quarters of pipeline and actuals, run the model, and walk you through a P25/P50/P75 forecast — defended against your real numbers.