Deep SensingTrusted by 120+ planners

The world’s first Deep Sensing planning system.

The planning system that reads every signalThe planning system that models every impactThe planning system that ranks your priorities

Start with Horizon Instant Book a demo

CRM updatesSales chatsCustomer emailsPOS dataWeatherMarket data

Trusted by planning teams at

What changes with Deep Sensing

The sharpest demand plans, in a fraction of the time.

18–45%

lower forecast error than traditional systems

Up to90%

of planning time back

Typically1–2%

of revenue added to EBITDA¹

¹ €2–4M a year. Worked example for an illustrative €200M company. We run the same calculation with you, on your own numbers.

Customer stories

Hear it from the planners.

Olivier Van Hoef

Supply Chain Manager, Ahrend

Siddhant Bhalinge

CEO, Ugaoo

[Name]

[Role], Inve Aquaculture

HORIZON DEEP SENSING

For decades, plans looked backward.The future stayed in the dark.

What’s about to happenis already out there.

Deep Sensing reads every signaland models every impact.

Illustrative example

Ori, your planning agent

Every signal read. Every impact modelled. Your priorities ranked.

Signals arrive from inside the company and out. HZN-1 and the Scenario engine put a number on each one. Ori hands your planners only the decisions that matter, ranked by impact.

Hover a signal to follow it through

1What Horizon read today43

CRM · 08:12

Retailer North launch moves to week 12

Read and linked

Email · yesterday

Retailer East opens 12 new stores

Read and linked

Weather · 06:00

Heatwave in Spain and Italy, weeks 27–28

Read and linked

40 more, already in the baseline.

2The engines model it· impact already modelled

HZN-175,000 unitsmove to weeks 12–13

Scenario engine11,520 unitsfirst fills, feasible

Impact map+18%ice cream, Madrid

3Ori· today’s decisions for Lena, key planner

Move the Retailer North launch

75,000 units to weeks 12–13

Plan first fills for 12 new stores

11,520 units, weeks 21–22

Add the heatwave uplift

+18% ice cream, weeks 27–28

Scenario engine

Every impact modelled, before it reaches the plan.

HZN-1 runs the baseline on autopilot. When something is sensed, like a port closure that stops deliveries reaching your customers, the Scenario engine models what it does to demand, with its drivers and assumptions, and turns each scenario into a feasible plan.

Hover a scenario to compare it

Trust

Every number, with a reason.

Start from the baseline forecast, adjust it with each live input, weighed by its track record, then anchor the result back toward the plan. Every step is visible.

43 signals read, 3 change the plan

SKU-4471

DC_01 to RETAILER_NORTH

Illustrative example

Final forecast

7.2K

Baseline forecast

6.2K

Difference

+15%

Step by step: from plan to final number

Each live input moves the forecast by what it is worth, weighed by its track record.

Baseline forecast

6.2K

Order book

+380

POS signal

+240

Store sell-out

−150

Customer update

+310

Sales update

+270

Pure signal read

7.3K

Anchor back toward plan

−90

Inside the engines

16+ proprietary models compete for every item. The best combination wins.

HZN-1, Horizon’s proprietary AI engine, runs models in four families and picks the most accurate combination per item by backtest. The optimisers turn every forecast into a feasible plan you can run.

Same data, same answer
Retrained every week (or more often)
A 5,000-SKU portfolio forecast in minutes

Inputs

HistoryOrdersPricesChannel dataProduct life cycleEvery event from the signal layer

HZN-1Horizon’s proprietary AI engine · demand side16+ models compete per item

Machine learning

Learns across items, locations and customers

Gradient boosting across itemsGradient boosting across locationsGradient boosting across customers

Deep learning

Retrained every week (or more often) on your data

Proprietary neural networks

Foundation models

For short histories and new products

Time-series model for short historiesTime-series model for new products

Causal models

Proprietary models for drivers and events

Price driversLaunchesEvents

Backtest for SKU-4471 × Retailer NorthAccuracy per model, higher is better

Gradient boosting across items85.8%

Gradient boosting across locations83.7%

Gradient boosting across customers87.2%

Proprietary neural networks86.9%

Time-series model for short histories84.4%

Time-series model for new products86.1%

Price drivers · causal85.2%

Launches · causal88.6%

+ 8 more models in the same backtest

Best combination for SKU-4471 × Retailer North

Chosen by backtest, settings tuned automatically

Launches · causal, 45%Gradient boosting across customers, 30%Proprietary neural networks, 25%

Auto-tuning · model choice and settings per item, by backtest

Across the hierarchy · total, product group, item, item × location, item × customer

Event memory and judgment model · every event kept with its measured effect

OutputThe baselineThe scenariosThe drivers behind every number

The optimisers · supply side

Inventory optimisationOr your own stock rules

Distribution planningLinear and mixed-integer

Procurement planningLinear and mixed-integer

Production planningLinear and mixed-integer

Production schedulingFinite capacity, constraint programming

A feasible plan

Integrations

Connect to every system. Plan for every horizon.

Our connectors for your ERP, APIs, or a file. AI maps items, customers, locations, units and dates into one model, and approved decisions flow straight back into the ERP.

Inner orbit: the sources Deep Sensing readsOuter orbit: the ERPs we connect to

SAP

Oracle

JD Edwards

NetSuite

Dynamics 365

Infor

IFS

Epicor

QAD

Plex

SYSPRO

Acumatica

Sage

Odoo

ERP

CRM

Data warehouse

APIs

Excel and files

Chats and emails

Explore all integrations →

Horizon Instant

Up and running in less than 24 hours.

Start with Horizon Instant, on the data you have: your ERP connected, and real demand planning on your own data. Add Deep Sensing and Ori when you’re ready.

Hour 0

Connect

Our connectors for your ERP, APIs, or a file

Map

AI maps items, customers, locations, units and dates into one model

Less than 24 hours

Plan

Planners work on the HZN-1 baseline and a first scenario

Every cycle

Keep

The mapping is kept and reused, every cycle

Start

Horizon Instant

ERP connected, the HZN-1 baseline and the full planning screens, on the data you have.

Then go as deep as you want

Deep Sensing, with Ori

CRM, sales updates, customer emails, tenders, competitor and market data, read and modelled. Ori ranks the priorities, explains them, and follows up with the people who know.

Data everywhere and never cleaned is yesterday’s problem: start with the data you have.

Three modulesDemand planning Supply planning Detailed scheduling

Start with Horizon Instant Book a demo