Windows desktop · fully offline · one installer

Spend your time planning,
not running software.

Planning methods proven by real planners. Forecasting, disaggregation, purchasing, scenario comparison - and every step can be run through an AI agent to automate your workflow. No lengthy integration project. Nothing leaves your desktop.

Data stays local Forecast models built in CLI + agent ready

Scenarios Every saved plan, priced and side by side

The Home screen: saved plans as scenario cards, each with its projected units and revenue by year and by sale event - Spring Event, Summer Kickoff, Prime Day and three Black Friday windows. FY27 Base is marked active with FY27 Upside beneath it, and the scenario tree in the sidebar carries the backdated re-forecasts alongside them.
0forecasting model families
100%offline after install
0bytes of telemetry, ever
1installer, upgrades in place

What's inside

The whole planning workflow, one app

Everything ships in a single Windows installer: the forecasting stack, the BI layer, the grids. Nothing to configure, nothing to host.

Forecasts you can trust

Classical statistics, gradient boosting and the Chronos-2 foundation model, blended into one ensemble and scored only on weeks the models never saw, with your real supplier lead time built into the test. Forecast at whatever level your data supports, and sale spikes are modelled as themselves instead of being smeared across the month.

Disaggregation that holds up

Push an approved forecast down through channel, category, style, size and store. Pin a share, type over a cell, and everything else re-bases around it automatically. Advanced logic, ordinary controls.

Purchasing that knows your constraints

Lead times, MOQs and target weeks of supply are real inputs, not spreadsheet math bolted on afterwards. It is warehouse-aware, so it suggests moving stock you already own before it suggests buying more.

From plan to financial impact

Set revenue targets in dollars and the plan solves back to them. Plan future discounts and read margin rather than units alone, so the buy becomes a financial decision instead of a units exercise.

Built by planners, not consultants

Starts from the Excel workbook you already have, and exports back to it. No integration project, no APIs to wire up. Every input stays visible and auditable, and smoothing your history is something you watch happen rather than a setting you have to trust.

Automate it, then report on it

A first-party CLI and a local HTTP API drive the entire workflow headlessly, so an AI agent can run your weekly cycle end to end. The built-in BI pivot compares scenarios side by side, tracks forecast accuracy over time, and exports straight to Excel.

How it works

Six steps from history to purchase orders

Each step hands the next one a reviewed number. Nothing leaves the app until you've signed off on it.

  1. Step 01

    Data input

    Load your sales history workbook

  2. Step 02

    Smoothing

    See and shape the history you forecast on

  3. Step 03

    Run forecast

    Configure models, run, compare

  4. Step 04

    Forecast results

    Review and adjust the topline

  5. Step 05

    Disaggregate

    Edit the mix down the hierarchy

  6. Step 06

    Purchase

    Store splits, transfers and buy plan

Inside the app

Four screens most forecasting tools don't have

Forecast accuracy scored the way you actually buy, a free-form pivot that ships in the box, a share split that keeps every sale in its own column, and a purchase plan that colours every cell needing a decision.

Two accuracy leaderboards for the Ecom channel, short horizon above and long horizon below, each ranking seven models on score, SMAPE, MAPE, WAPE, bias in percent and in units per month, and RMSE. Chronos2 leads the short horizon at 15.4% error against a seasonal naive's 20.8%; over the year Prophet leads and the seasonal naive is last at 68.5%. The ensemble blend behind each horizon is named above its table.

Every candidate is scored on weeks it never saw, and ranked where you can read it. Each origin cuts history, refits every model, holds a freeze window the length of your supplier lead time - the stretch you could not react inside anyway - and only then starts scoring. You get error and bias, in percent and in units a month, so a model that is accurate on average while running consistently high is not mistaken for a good one. The same channel, two horizons: Chronos2 forecasts the next quarter to 15.4% error against a seasonal naive's 20.8%, and across the year that naive is last at 68.5% while Prophet leads. The blend reweights with it - half on Chronos2 near in, none of it far out. No model is simply the best one, which is the argument for scoring every node and every horizon rather than picking a favourite and living with it.

Backtest validation: actual demand as a dark slate reference line, with two model traces stitched across every rolling origin from 2019 to 2026. The other models stay in the legend, greyed. Dashed vertical lines mark where each window's history was cut.

The same runs drawn against the actuals, seven years of them. Where a model could not score a window its line breaks rather than bridging the gap. The dark reference line is actual demand, promotions and all - the models are fitted on a smoothed series, so a promotional week is a spike they were never shown, and the gap at one is the honest size of it.

The built-in BI dashboard: an Excel-style free-form pivot with a field list and drop zones for Filters, Rows, Columns and Values. Two saved plans, FY27 Base and FY27 Upside, sit one per row with their forecast months side by side, and the chart above shows the pair diverging.

Drag Scenario into Rows and two saved plans sit one per line, recomputed server-side from the tidy fact table. Here an upside plan carrying a +12% revenue goal is set against its base, month by month - the kind of thing you want to see before you commit the buy. It runs in-process on your own machine, and it exports straight to Excel.

The shares editor: menswear categories as rows, with each month's share of the division across history and forecast. Ribbed Crew Sweater holds around half the division and climbs during Black Friday, while Relaxed Bodysuit gives share back over the same weeks. Sale periods sit in their own columns beside the months.

A topline forecast is only useful once it is split the way you actually buy. Every level of your hierarchy gets its own share of its parent, seeded from history and editable in place. Sale periods are their own columns, not smeared across the month that contains them - so Black Friday lifting one category from 49% to 52% while another gives share back is a number you can see and change, not an assumption buried in a weekly average.

The purchasing inventory projection: a week-by-week walk-through of opening stock, inbound, buys, transfers and demand down to unmet demand, ending units and weeks of supply, shown for the total and for a single style beneath it. Buys land on their supplier's lead time rather than all at once, weeks below the cover target are highlighted in amber, and any demand the buy could not cover in time is called out in its own row.

The last mile most tools leave to a spreadsheet. Approved forecasts become buy quantities per vendor and week, walked forward period by period - opening stock, inbound, buys, transfers and demand down to ending units. Every week that needs a human says so in colour: amber where cover falls under target, red where the stock runs out and the buy could not close the gap. A clean plan shows neither. Cover runs as a sawtooth around your target rather than sitting on it - an order has to carry the weeks until the next one lands as well as the target itself, so the week stock arrives reads high and draws down from there.

Pricing

Priced for a planning team.

One subscription, three seats

Three planners are included in the base price, so a team can start without counting heads. Add more whenever you need them. Checkout is handled by Stripe and your download link is issued the moment payment clears.

Base subscription
$4,800 a year - three planner seats included
Each extra planner
+$300 a year, or $25 a month - added at checkout
Prefer to start small?
$490 a month, cancel any time - annual works out cheaper
Ten seats and up
Priced with you directly - the seat picker will hand you over
Updates
Every release while you subscribe, installed in place

PlanCore License

seats

$4,800per year

3 seats included

$4,800 a year, three seats included · +$300 per extra seat · 10+ seats priced with you directly
Have a promo code? Enter it at checkout.

  • Full Windows desktop app - every feature, no tiers
  • Multi-planner ready - edits are attributed by user
  • Code-signed offline installer (~1 GB, forecasting models included)
  • Every update while you subscribe, installed in place
  • Personal onboarding & setup call included
  • Direct email support from the developer

Secure checkout via Stripe · instant download link · questions first? Get in touch

Digitally code-signed and verified by Windows before it runs, published with a SHA-256 checksum, and fully offline after download. Upgrades install in place and never touch your data.

Support & setup

You'll be talking to the developer

No ticket queues. Email lands directly with me, and every license includes a hands-on setup session.

FAQ

Common questions

What do I get after purchasing?

Immediately after checkout you land on a page with your personal download link for the latest PlanCore-Setup.exe, plus install instructions. You'll also get a Stripe receipt by email, and I'll follow up to schedule your setup call.

Is the installer safe? Windows warned me about downloads before.

The installer is digitally code-signed, so Windows can verify the publisher before running it, and each release ships with a SHA-256 checksum you can verify yourself. It's a large download (~1 GB) because the entire forecasting stack is bundled - nothing is fetched at runtime.

Does my data go to the cloud?

No. PlanCore is fully offline after installation. Forecasting runs on your machine, your workbooks stay in your local app data folder, and the app never phones home - there is no telemetry of any kind.

How do updates work?

When a new version ships you'll get an email with the same permanent download link. Run the new installer and it upgrades in place - your scenarios, workbooks, and settings are never touched.

What are the system requirements?

Windows 10/11, 64-bit. 8 GB RAM minimum (16 GB recommended for AI forecasting on larger datasets). No graphics card required: the forecasting stack ships CPU-only and runs on an ordinary work laptop.

If the machine does have a recent NVIDIA card, the installer offers to enable GPU acceleration for forecasting - it's a tick box during setup, and it downloads about 2.4 GB the one time. It's only offered for cards new enough to actually be faster, and if the card turns out not to work the installer puts the CPU version back on its own.

Can I try it before buying?

Get in touch - I'm happy to do a live demo on a screen share with your kind of data, and walk through whether it fits your workflow before you spend anything.

Ready to plan with confidence?

One installer. Every model. Your data stays yours.

Buy PlanCore