[ AI demand forecasting ]

Plan Production, Staffing and Stock From One Forecast Everyone Trusts

We build forecasting systems that predict demand from total company level down to product and region, account for promotions and launches, and plug into your S&OP process.

Totals and details that always add up
Hierarchical
Totals and details that always add up
Uplift and cannibalization modeled
Promotions
Uplift and cannibalization modeled
Feeds planning meetings and ERP
S&OP ready
Feeds planning meetings and ERP

Units forecast, all regions: 16,400

Workers in blue uniforms at an assembly line in a large factoryNorthCore range4,185
Seasonal1,732
New launch1,299
SouthCore range3,254
Seasonal1,994
WestCore range2,165
Promotions1,771

Example forecast. Every tile adds up to its region, every region to the total.

[ The problem ]

Why Demand Plans Drift From Reality

Forecasts built in separate spreadsheets rarely agree, and nobody trusts the final number.

  • Sales says

    55,000

    Pipeline and targets

  • Finance says

    46,500

    Budget and run rate

  • Operations says

    49,800

    Last year plus growth

  • Marketing says

    52,600

    Campaign calendar

Four example forecasts for the same quarter, built in four spreadsheets.

  1. 01

    Plans that do not add up

    Sales, finance and operations each keep their own forecast, and the planning meeting is spent arguing about whose number is right.

  2. 02

    Promotions distort history

    Past campaigns inflate the baseline, so next year's plan expects a lift that only happened because of a discount.

  3. 03

    New products guessed

    Launches are planned from instinct and a hopeful sales target, then end up overstocked or sold out in week two.

  4. 04

    Late exceptions

    Big misses surface at month end, after production runs, shifts and purchase orders are already locked in.

[ How it reconciles ]

Four Forecasts In, One Number Out

Every team keeps its view, but the plan runs on one reconciled forecast that adds up from product to company total. Example figures for one quarter.

  1. 1

    Each team's number is read in as an input, not thrown away.

  2. 2

    The model forecasts every product and region from orders, prices and promotions.

  3. 3

    Lower levels are reconciled so they add up to the company total.

  4. 4

    Planners see one consensus number, with every override tracked.

Next quarter, units (example)

  • Sales55,000
  • Finance46,500
  • Operations49,800
  • Marketing52,600
Consensus forecast50,300

Reconciled from product and region forecasts, so every level adds up to this total.

[ Promotions ]

What a Promotion Really Adds

The model separates the real gain from the noise around it, so a two-week discount does not become next year's baseline. One example campaign:

1,000
+420
-90
-60
+270

Baseline demand

Promotion uplift

Taken from a sister product

Dip after the promotion

Net gain

[ How we build it ]

How We Build Your Forecasting System

We backtest against your current plan, then embed the forecast in your planning cycle.

Start With a Backtest
  1. 01

    Planning review

    We map your hierarchy, planning horizons, S&OP calendar and current forecast method.

    Output: Forecast design

  2. 02

    Data and drivers

    Sales, orders, promotions, prices and outside drivers are cleaned and joined.

    Output: Driver dataset

  3. 03

    Model backtest

    Hierarchical models are compared with your current forecast across past cycles.

    Output: Accuracy by level and horizon

  4. 04

    Planning integration

    Forecasts, exceptions and scenarios feed dashboards, S&OP meetings and your ERP.

    Output: Live forecasting cycle

[ Capabilities ]

What the System Does

One forecast for every team, explained and easy to adjust.

  1. 01 / 06

    Hierarchical forecasts

    Company, category, product, region and channel levels reconciled.

    Workers at an assembly line in a large factory
  2. 02 / 06

    Promotion modeling

    Uplift, halo and cannibalization from planned campaigns.

    Shopper reaching for a box on a supermarket shelf
  3. 03 / 06

    New product forecasts

    Launch curves from similar past products and attributes.

    Supermarket shelf of new cleaning products
  4. 04 / 06

    S&OP integration

    Consensus numbers, overrides and versions for planning meetings.

    Planning team meeting in a glass-walled room
  5. 05 / 06

    Exception management

    Items drifting from forecast surfaced early, ranked by impact.

    Person reviewing charts on a laptop
  6. 06 / 06

    Scenarios

    What-if plans for price changes, supply limits or new markets.

    Delivery trucks parked outside a distribution warehouse
[ Your planning month ]

Built Into Your Monthly Planning Cycle

The forecast arrives when each meeting needs it, with the exceptions already ranked.

  1. Week 1
    01

    Data refresh

    Last month's orders, shipments and promotions are loaded and the forecast reruns overnight at every level.

  2. Week 2
    02

    Demand review

    Planners work through the exceptions list, biggest impact first, and record any overrides with a reason.

  3. Week 3
    03

    Supply review

    Operations checks the consensus forecast against capacity, materials and staffing, and runs scenarios where they fall short.

  4. Week 4
    04

    Executive S&OP

    Leadership signs off one plan, with the forecast, the gaps and the financial view on the same page.

[ Drivers ]

Data the Models Use

We test each driver in the backtest and keep only the ones that improve accuracy at the level you plan on.

  • 01Orders and shipments
  • 02Promotions and pricing
  • 03Product hierarchy and attributes
  • 04Customer and channel data
  • 05Economic and market indicators
  • 06Weather and calendar events

Models

  • Prophet
  • XGBoost
  • LSTM networks
  • Hierarchical reconciliation

Data

  • Snowflake
  • dbt
  • Apache Airflow
  • PostgreSQL

BI

  • Power BI
  • Tableau
  • Looker

Planning

  • SAP IBP
  • Anaplan
  • Oracle
[ Targets ]

What We Measure With You

Targets are set during the backtest; results depend on your data and planning process.

  • 01

    Forecast error

    Error by level and horizon versus your current plan

  • 02

    Forecast bias

    Consistent over- or under-forecasting

  • 03

    Inventory cost

    Holding and obsolescence cost

  • 04

    Expedites

    Emergency orders and overtime

Where It Fits

  • ManufacturingProduction plans and raw material purchasing.
  • CPG and FMCGRetailer demand with heavy promotion calendars.
  • RetailStore and online demand by category and region.
  • Wholesale distributionCustomer orders across many warehouses.
  • PharmaceuticalsDemand for critical products with long lead times.
Delivery truck parked outside a large distribution warehouse

[ One forecast ]

From the Plant Floor to the Loading Dock, One Number

Production, purchasing, staffing and finance plan from the same reconciled forecast, so the month-end surprise becomes a ranked exception in week two.

Book a Planning Call
[ Client voices ]

What Planning Teams Say

  • 5.0Clutch
  • 5.0GoodFirms
  • 5.0Google
  • 4.9Upwork
I had the pleasure of working with Sajal Tech on a travel website, and it was an incredible experience from start to finish. They demonstrated professionalism, clear understanding of requirements, and excellent communication throughout.
DKDaniel KwekaCEO, Darius Digital
Knowledgeable, efficient, and ahead of schedule. Will hire again.
TMTom McGrathFounder, Tipperary | eIrish.com
They completed our project even when the scope changed slightly. They were cooperative when we wanted to add or modify features.
AUArda UygurFounder, Hallolur
[ Cost ]

Demand Forecasting Cost

Indicative starting prices. Cloud usage is billed to your account.

Option 01

Forecast backtest

Scope
Data review and a backtest against your current plan
Timeline
2 to 3 weeks

from $2,500

Option 02

Production forecasting

Scope
Hierarchical forecasts, dashboards and exceptions
Timeline
6 to 10 weeks

from $12,000

Option 03

S&OP platform

Scope
Scenarios, consensus workflow and ERP integration
Timeline
3 to 5 months

from $35,000

What You Own at Handover

  • Hierarchical forecasting models and their training code
  • Backtest report by level and horizon against your current plan
  • Planner dashboard with exceptions, overrides and versions
  • Write-back into your ERP or planning tool
  • Monitoring for accuracy, bias and data freshness
  • Runbooks and a handover workshop for your planners
[ FAQ ]

Demand Forecasting Questions

Still deciding? A 30-minute call usually answers the rest.

Ask Us Directly
  • Demand forecasting plans what customers will want across the business; inventory forecasting turns that into stock and reorder decisions. Many clients use both.

  • Yes. Overrides are part of the workflow and are tracked, so you can see where human input improved or hurt accuracy.

  • From days to 18 months or more, with accuracy reported separately for each horizon.

  • Yes. Forecasts can be written to SAP IBP, Anaplan, Oracle or your ERP, or shown in Power BI and Tableau.

  • That is common. The first weeks focus on cleaning history, removing promotion effects and handling gaps.

  • Launch curves are built from similar past products, matched on category, price point, channel and season, then updated as the first weeks of real sales come in.

  • Yes. Scenarios can cap demand at available capacity or materials, so the plan shows both what customers want and what you can actually ship.

  • In your cloud account or warehouse. We follow ISO 27001 practices.

[ Contact ]

Let's build your next product together

Book a free strategy call and leave with a clear plan and estimate. No commitment.

  1. 01Pick a time that suits you
  2. 0230 minutes on scope, stack, timeline and budget
  3. 03Fixed-price proposal, NDA on request
Prefer email? contact@sajaltech.com