The income statement tells you that margin moved. It doesn't tell you that one SKU is quietly dying while another is ramping, that a warehouse on one coast is out of an item the other coast has ninety days of, or that an assembly costs 20% more to build than its standard says. Those answers are in NetSuite, in the item records, the inventory balances, the work orders, and the fulfillments. They're just spread across more tables than anyone wants to join by hand.

The Sonar edition of the prompt library went live this month, and thirteen of the new prompts are for exactly this layer. This post covers them in five groups: which items are worth keeping, where the stock is and isn't, what it costs to make, vendors and shipping, and the item master itself.

If you're new to this, Sonar AI is an AI agent that runs inside NetSuite. Every prompt in the library is a playbook that I engineered and tested against live NetSuite data, and you run it inside your own account, against your own records.

What They Share

All thirteen are read-only, and all of them carry the NetSuite quirks that cost me time to learn, so the agent doesn't have to learn them again. The item type field is called one thing and not another. The sales price on an item record isn't exposed to a query, so unit price has to come from the transaction line. Sales lines on an invoice carry negative quantities, and a positive line isn't a sale. Every prompt states its currency basis and refuses to sum across currencies. And every one of them uses the item's average cost as its cost basis and says so, because that's the number the account actually has.

Which Items Are Worth Keeping

Item Velocity Trend gives every SKU a lifecycle stage. It builds twelve months of unit sales per item and compares the last three months to the first three. Up more than 20% is rising. Down more than 20% is declining. In between is stable, unless the month-to-month swings are so wide that the item is better described as volatile, which is its own class and needs a demand buffer rather than a forecast.

Item Velocity vs. Margin Quadrant plots every item on two axes, how fast it sells and how much margin it makes, and splits the field at the medians. Four quadrants, four actions: promote, protect, reprice, or discontinue. Margin is computed from revenue and average cost, and the prompt states that method in the output so nobody mistakes it for a standard-cost margin.

Item Pricing Drift Forensics looks at the same items from the revenue side. It finds items sold at inconsistent prices across invoices, using the per-unit rate on the transaction line, and it quantifies what the variance is costing. This is where uncontrolled discounting shows up, one item at a time, and the prompt is careful not to invent an approved price list it can't see. It reports the spread and lets a person decide which end of it was authorized.

Return Rate by Item measures returns as a share of sales for the same item and period, using the three return transaction types NetSuite actually has, and it links each return back to its original sale. The output leads with the dashboard, total returns, return value, the overall rate, and the count of items that cross the critical line, then the per-item detail with whatever reason codes the account recorded.

Where the Stock Is and Isn't

Inventory Stockout Risk Score scores every active SKU from 0 to 100. The arithmetic is plain: daily demand from the last thirty days, days of supply from on-hand and demand, and a comparison against the vendor's average lead time, derived from the account's own purchase orders and receipts. If days of supply is less than the lead time, the item will stock out before replenishment can arrive, and it scores critical. Items with no lead-time data are flagged as incomplete rather than scored, because a score without lead time is a guess.

Multi-Location Inventory Imbalance answers the two-coasts question. For each item it compares on-hand and demand across locations and quantifies the rebalancing opportunity. The rule that gets a hard flag is one location at zero while another holds more than ninety days of supply. That's a transfer order waiting to be written, and the prompt names the item, the locations, and the quantity.

Backorder Aging and Priority Ranker takes every open sales order and gives it a priority score from three parts: age, order value, and whether it's already partially fulfilled, which earns a bonus because that customer is already waiting on a partial. The tiers are stated, and each tier is quantified in dollars and order count, so the fulfillment team gets a ranked list rather than a long one.

What It Costs to Make

Bill of Materials Cost Analyzer explodes the BOM structure, computes the standard build cost from component costs, identifies the components that drive it, and flags the gaps: any component with no cost or a zero cost, because those create a roll-up that's wrong by exactly the amount nobody entered. It compares the BOM-derived cost to the average cost on the assembly record and discloses the difference.

Assembly Build Profitability Analyzer compares the standard cost of what was built to the actual cost that hit the general ledger, and then to the revenue those assemblies earned. Standard and actual are kept separate throughout, and the deliverable is a dollar variance per assembly, not a count of builds.

Work Order Cost Variance Analyzer goes a level deeper, to the work order, and it exists partly because of a trap: the amount field on a work order line is empty, so the actual cost has to come from the accounting lines. The prompt knows that, distinguishes work orders from the builds that complete them, computes standard against actual by assembly and period, and points at the root cause of the overrun, whether that's component cost, quantity, or something in the routing.

Vendors and Shipping

Vendor Lead Time Reliability Tracker measures lead time from order date to receipt date for every vendor, and then does the part that a single average hides. It scores variability, because a vendor who averages ten days with a range of two to thirty is worse than one who averages fourteen and always delivers in fourteen. It scores the trend, comparing the last three months against six. And it correlates lead time with fill rate. The three scores sum to a reliability score out of 100, with the components shown.

Shipping Cost Optimizer works from item fulfillments and shipments to analyze volume, order-to-ship cycle time, and carrier patterns. Several of the shipping fields on a transaction aren't exposed to a query, and the prompt says which ones and where it goes instead, including NetSuite's own fulfillment report when that's the authority. It won't invent a shipping cost it can't find.

The Item Master Itself

Item Master Health Score is the prompt I'd run first, because every other analysis in this post depends on the item records being complete. It scores each item from 0 to 8 on eight attributes, one point each, and tiers the results. Two of the attributes are marked critical: a missing cost blocks valuation, costing, and receiving, and a missing description blocks customer-facing transactions. The output isn't a count of items with problems. It's the list of which ones, and which field.

Wrapping Up

There's a sequence hiding in this list. Fix the item master first. Then find out which items matter, using velocity and margin. Then make sure those items are in the right place, in the right quantity, with a vendor you can rely on. Then find out what they really cost to make. Each prompt answers one question, but the questions have an order, and the order is the operations review that most companies mean to do every quarter and rarely get to.

All thirteen are in the paid tier of the Sonar AI Prompt Library, under Supply Chain and Inventory.