The problems

Five problems

The solution

One system that carries the forecast forward and tells you what to order, when, and from whom.

A forecast that already knows Q4 Your own math, run weekly, with seasonality and trend applied as visible factors you can adjust. Later, sales sentiment and launch ramps.
Months on hand, live Per product, at both levels you use: buildable and on the shelf. Plus what is on a boat and the projected stockout date.
PO planning against a dollar target Set the inventory dollars you want to hold. The system proposes what to order, when, from whom, and shows scenarios side by side.
Built in three phases. Each one ships working software you use before the next starts. Your own ritual, formalized: the engine runs beside Christian's sheet until it matches, then it gets better.

Phase 1

Months on hand, plus a forecast that sees Q4 Eight weeks. $26,000.

  • NetSuite connected. Reads sales, inventory, BOMs, and open POs. One write: reorder and build points, uploaded the same way Christian's CSV is today.
  • Forecast engine v1. Christian's math ported exactly and run weekly, then seasonality and trend on top, each shown as "we noticed this and adjusted like so."
  • Months on hand per product. Buildable, on the shelf, in transit, and a projected stockout date. Order-more and build-more shown as different states.
  • Shared components split by the forecast. If purple is 60% of a shared PCB's demand, purple gets 60% of the PCBs. No percentage table to maintain.
What it solves Black Friday visible in August. Months on hand without the monthly export. The gut layer gets named, adjustable slots with an audit trail.
How it goes Four weeks to a working version in Ryan's hands. Four more weeks of iteration with you. Parity with Christian's sheet is the acceptance test.

Phase 1 inputs and signals

Eight inputs, six already in NetSuite and two forward signals on top. Nothing new to instrument.

InputWhere it lives todayWhat it feeds
Line-level sales per SKUNetSuite, the daily report Shirley already exportsThe 12 / 6 / 3 / 1 month baseline and current velocity
Monthly sales history, 24+ monthsNetSuiteSeasonality per SKU and per group
Inventory on hand, kits and 900sNetSuiteMonths on hand at both levels
BOMs, 900 to kitsNetSuiteComponent demand, buildable months, shared-part splits
Open POs: quantity, vendor, land dateNetSuiteIn-transit stock and projected stockout dates
Current reorder and build pointsNetSuite, Christian's monthly uploadThe parity target and the write-back path
Lead-time constantsRyan's headThe reorder horizon math
Christian's generator sheetDropboxThe baseline, ported cell for cell
Signals in Phase 1: seasonality and trend. Both computed from your history, both shown and adjustable. Later phases add sales trajectory, channel mix, part availability, sentiment, and launch ramps.

Phase 2

PO planning against a dollar target $35,000. Timing set at kickoff.

  • The PO planner. Set the inventory dollars you want to hold. It proposes what to order, when, and from whom, and runs scenarios side by side.
  • Approvals with an audit trail. Proposed, reviewed, approved, issued. The PO email thread ends.
  • Availability watch. Mouser stock on the commodity parts in your BOMs, plus an alert when an open PO's land date slips.
  • Sales signals. Demand by channel, and Shirley's trajectory calls as an input the forecast can see.
What it solves The ask from the first call: set a dollar target and the system says what to order. The ten-weeks-becomes-twenty surprise shows up weeks early, while spot buying is still cheap.
What it fixes underneath Distributor lumps stop reading as trends. Vendor lead times and terms live in one place with dates on them.

Phase 3

Forward signals, cash fit, and the board view $30,000. Timing set at kickoff.

  • Forecast engine v2. Sales sentiment and launch ramps, a second labeled engine beside the first.
  • HubSpot and Shopify connected as signals into the forecast.
  • Cash fit. A PO payment calendar built from vendor terms, held against the line-of-credit comfort zone.
  • Board view. Built from the system's own numbers, not typed in.
What it solves Mantle-class launch misses. Cash-aware PO timing through the summer lull. The buffer-stock case quantified for the board instead of a number Matt types in.
Sequenced, not ignored Annual budgeting, the production scheduler, and AR collections are parked behind this, in that order.

Agent Base

Everything lives in Agent Base One managed connection for the company. Every agent, screen, and utility runs through it.

ContextWhat the business knowsCompany facts, products, vendors, terminology. Served to every agent from the first message. Maintained once.
ConnectionsWhat the business can reachNetSuite, later HubSpot and Shopify. Connected once at the company level, scoped per capability, every call on the record.
SkillsThe best way to do itRyan's hand-taught flows, like the weekly reorder review, captured once and versioned for everyone.
UtilitiesWork the company depends onThe weekly forecast run. The months-on-hand service. Run the same way every time, with a record of every run.
What a utility is. Custom-built software that plugs into Agent Base and does one job the same way every time (some people would call it an automation). It runs on a schedule or on request, keeps a record of every run, and works from any AI tool your team uses: Claude today, anything that speaks the same protocol tomorrow.

Agent Base

What Walrus gets from it The inventory system is delivered on Agent Base. That is where it runs, and where the next thing lands.

One connection instead of many NetSuite is connected once, for the company. Nobody's personal login is load-bearing. Access is granted per role, and every action is logged.
Ask it from Claude Ryan already asks Claude for months on hand. Now the answer comes from the same engine as the screen: "months on hand for the Julia" gets the governed number, with the factors behind it.
The next thing is configuration, not a project Sales signals in Phase 2, HubSpot and Shopify in Phase 3, budgeting after that. Each lands in the same place without a new integration.
$1,000 a month, starting when Phase 1 goes live. Maintenance and monitoring included. Benali runs it; your team uses it. It is part of the plan, not an add-on.

The price

Three phases and Agent Base $91,000 to build. $1,000 a month to run.

01 · 8 weeks

Months on hand + forecast

$26,000
  • NetSuite connected
  • Forecast engine v1, weekly
  • Months on hand at both levels
  • Shared components split by forecast
02 · timing set at kickoff

PO planning against a dollar target

$35,000
  • PO planner with scenarios
  • Approvals with an audit trail
  • Availability watch and slippage alerts
  • Sales signals
03 · timing set at kickoff

Forward signals, cash fit, board view

$30,000
  • Forecast engine v2
  • HubSpot and Shopify connected
  • Cash fit
  • Board view
Agent Base

Where it all runs

$1,000a month
  • Starts when Phase 1 goes live
  • Maintenance and monitoring included
  • Grows as capabilities are added
Each phase is its own project: half at kickoff, half at completion. Phases 2 and 3 are planned prices, confirmed with you before each one starts. Nothing in a later phase is owed until you say go.