What Timberline is#

Timberline Harvest Co. makes and distributes woodworking products (Timber & Tools, about 60% of revenue) and agricultural and homestead equipment (Farm & Field, about 40%). FY2026 net revenue is about $820M, across 3 regions and 13 markets, 4 sales channels, and a 237-person sales organization. There are actuals for 2022 through 2026 and a 2027 Plan and Forecast.

None of it is real. All of it reconciles.

Four layers#

Timberline is not one artifact. It is four, stacked:

1. The dataset. A deterministic Python pipeline generates every number in the company from one configuration file with a fixed random seed. Rerun it and you get byte-identical output, which is why a screenshot taken last month still matches the numbers on screen today. It also verifies itself: 12 assertions check that the sales journal sums to the sales totals, that the commission journal traces to individual invoices, that the balance sheet balances at every month end, and that cash flow reconciles to the cash movement.

2. The planning model. The generated numbers load into a planning model: 7 cubes (sales, income statement, balance sheet, cash flow, sales ops, compensation, and a control cube) over 15 dimensions.

3. The content. Workbooks and decks that read that model: the board pack, the market review, the financial board book, the sales information workbook, the rep scorecard, the commission statement, and two contribution form templates, plus a set of parameterized templates built for AI to answer questions from.

4. The application. A complete Reportworq v6 workspace: 8 jobs, 2 contribution campaigns, 10 user accounts in 8 groups, 2 data connections, a schedule, message templates, run history, and an inbox full of mail that previous demo runs actually sent.

Layer 4 is what people mean when they say "open Timberline". Layers 1 to 3 are what make it defensible.

What is actually in the workspace#

Everything lives in one workspace (Default), in one folder tree:

Timberline
└─ Financial Reports          ← the 8 jobs and 2 campaigns live here
   └─ Source Files            ← the 9 workbooks and decks live here

Eight jobs:

• Board Packet · Board Packet with Division Slides · Divisional Packet, three ways to build a board deck, differing in how they replicate
• Financial Statements · Financial Statements - SharePoint: the same statements, delivered one way and four ways
• Timberline Sales Information: the deep sales analysis, and the AI showcase
• Sales Rep Scorecard: manager-and-team fan-out, with generated commentary in the email
• Commission Statements: bursting at scale, rolled up into one email per manager

A contribution campaign:

• Segment & Channel Budget Input: the full planning cycle: forms, approvals, write-back to two cubes

See the report catalog and the campaign catalog for what each one is for.

Why it is built this way#

Three design decisions explain most of what you will see.

Everything ties. The board deck, the sales report, the commission statement and the budget form all read the same model. A number quoted on one screen is the same number on the next. That is what lets a demo move around without contradicting itself, and it is the reason the pipeline exists.

The model owns the metadata, not Reportworq. Recipient email addresses, manager reporting lines, region-to-VP mappings and the current reporting year all live in the planning model as dimension attributes and control values. The workbooks look them up. So the distribution list is never stale, and rolling the year forward is one change in one place.

Delivery is part of the report. The jobs build real email dashboards: a greeting, a headline number, a picture of the report, a live HTML table, and on two jobs a generated commentary paragraph, with the full detail attached for whoever wants it. Judging Timberline by its attachments misses the point.

The safety property#

Email is configured in demo-inbox mode. Delivered messages land in an inbox inside the application instead of going to a mail server, so a demo run cannot mail a real person. Everything else follows from that: you can run any job, any number of times, and read exactly what each recipient would have received.

Check that setting before running anything on an environment you did not set up yourself.

Related#

• Signing in, and who to be
• Where the files live