Inventory Ordering System: Reorder Points and the PO Process

What Is an Inventory Ordering System, Mechanically?

An inventory ordering system is the combination of a reorder point per SKU, a safety stock buffer built into that reorder point, and the purchase order process that fires once stock crosses it. It’s a mechanism, not a piece of software by itself: you can run one on a spreadsheet with a formula and a weekly stock count, or on a dedicated tool that checks stock levels automatically, but the underlying logic is the same either way.

For a jewellery brand doing $3M to $30M in revenue on Shopify Plus or a comparable subscription platform, this mechanism has to work across a catalogue that’s unusually mixed. A best-selling vermeil chain restocked from a domestic supplier behaves like a normal retail SKU, ordered often, on a short, fairly predictable lead time. A solitaire engagement ring built from a loose stone sourced from a gem dealer, cast at a separate house and finished on the bench behaves nothing like it: ordered rarely, on a long lead time with several independent points of delay, any one of which can slip.

Running both under one reorder point formula is where most of the breakage in this article traces back to. These mechanics apply to any SKU in the catalogue; a jewellery-specific wrinkle that changes the maths gets called out directly rather than folded into a general rule that doesn’t hold for both ends of the catalogue.

How Does a Reorder Point Actually Get Calculated?

The standard formula is: reorder point = (average daily demand × average lead time) + safety stock. The first term covers demand during the time it takes a new order to arrive; safety stock covers the gap between that average case and a worse one.

Consider an illustrative example: a 14k gold vermeil chain SKU sells an average of 3 units a day, and the domestic supplier’s lead time averages 10 days but has run as long as 16 days over the last several reorders. Using the simple max-minus-average method, safety stock = (maximum daily demand × maximum lead time) − (average daily demand × average lead time). If the SKU’s daily demand also peaks at 5 units on a strong day, that’s (5 × 16) − (3 × 10) = 80 − 30 = 50 units of safety stock. The reorder point becomes (3 × 10) + 50 = 80 units: stock on hand plus stock already on order falls to 80 units, and a new purchase order should go out that day.

That figure is illustrative only, built to show how the two inputs combine, not a number to copy into a live system. The two inputs that actually decide the outcome are the demand figures and the lead-time figures, and both need to come from your own sales and purchasing history, not from memory or a rule of thumb carried over from a previous employer.

Two things break this formula quietly. First, average lead time hides the tail: a supplier that runs 10 days on average but occasionally 16 has a materially different risk profile from one that runs 12 days every time, even though both might report the same average. Second, average daily demand hides seasonality: a chain SKU that sells 3 units a day in July and 9 units a day in the two weeks before Valentine’s Day needs a different reorder point at each point in the calendar, not one fixed number carried all year.

Why Does a Fixed Reorder Point Fail as Lead Time Varies?

The obvious approach, set a reorder point once from a supplier’s quoted lead time and revisit it occasionally, fails for a specific and predictable reason: a quoted lead time is a target, not a guarantee, and the gap between the two widens under exactly the conditions where you most need the order to arrive on time.

A gemstone supplier quoting a 30-day lead time is usually quoting the time to source and ship a stone that’s readily available. During a period of high demand across the trade, around major holiday buying seasons or after a run on a popular cut or carat weight, that same supplier’s actual lead time can stretch well past the quoted figure, because you’re no longer first in their queue. A reorder point built from the quoted 30 days, with no safety stock buffer for that stretch, triggers an order that still arrives late.

Casting adds a second point of failure that’s easy to miss because it sits between two suppliers rather than inside either one. A brand that sources a stone from one supplier and sends it to a separate casting house for setting has two lead times in sequence, and a delay at either stage delays the finished piece. A reorder point calculated from the stone supplier’s lead time alone, ignoring the casting step, will be short by however long casting and finishing actually take.

The practical fix isn’t a single, larger safety stock margin applied everywhere, that just ties up cash across SKUs that don’t need it. It’s tracking lead time as a range per supplier and per production step, updating the reorder point’s safety stock component whenever that range shifts, and treating a new gemstone supplier or a new casting relationship as an event that requires a reorder point review, not something to fold into the existing number unchanged.

How Should Safety Stock Account for Lead-Time Variability?

Safety stock exists specifically to absorb the gap between a supplier’s average performance and its worst recorded performance, and sizing it from the average lead time instead of the range is the single most common reorder-point mistake at growing brands.

Take a second illustrative example, this time for a slower-moving fine-jewellery SKU. Say a solitaire ring style sells an average of 2 units a month, the gem-sourcing and casting lead time together averages 40 days, and the longest recorded lead time over the last year was 65 days, driven by a stone that had to be resourced after the first one didn’t clear quality check. Converting the monthly demand to a daily rate for the formula, roughly 0.07 units a day, safety stock = (maximum daily demand × maximum lead time) − (average daily demand × average lead time). Using a conservative maximum daily demand of 0.13 units (roughly double the average, to allow for a pair of pieces ordered close together), that’s (0.13 × 65) − (0.07 × 40) = 8.45 − 2.8, or about 5.65 units, rounded up to 6 for a physical stock count. The reorder point becomes (0.07 × 40) + 6, or about 8.8, rounded to 9 units.

That result is illustrative and specific to the assumptions stated, not a benchmark. What it illustrates generally holds: a slow-moving, long-lead-time SKU can carry a reorder point that looks disproportionately high relative to its monthly sales rate, and that’s correct, not a sign the formula is broken. The alternative, sizing safety stock from average lead time only, would have produced a reorder point closer to 3 units, which looks tidier on a stock report and would have triggered a reorder too late during the exact month the supplier ran long.

Two adjustments matter beyond the base formula. Where a supplier’s lead time is trending in one direction, getting consistently slower as their own order book fills, or faster as a new relationship beds in, use the trend rather than a flat historical average, because a backward-looking average always lags a real shift. And where a SKU sells in genuine spikes rather than a smooth distribution, a proposal season, a specific holiday, a single influencer mention that moves a style, a safety stock formula built on daily averages understates the risk; that SKU is better served by a manually set buffer ahead of the known spike than by a formula tuned for steady-state demand.

What Does the Purchase Order Process Look Like Around a Reorder Point?

A reorder point is only useful once it triggers a purchase order that actually gets sent, approved and received correctly, and this is where the process around the number matters as much as the number itself.

The sequence, in order: stock crosses the reorder point, either checked automatically against a live stock count or reviewed on a scheduled cadence; a draft purchase order is generated at the reorder quantity, which should already be rounded to the supplier’s minimum order quantity or standard casting batch size rather than left at a raw calculated figure; the draft is checked against open budget and any recent lead-time or price change the supplier has flagged; it gets sign-off, which for a jewellery brand often means checking that a gemstone spec or metal purity is correctly stated before it goes out, not just the quantity; the order is sent and its expected arrival date is logged against the reorder point’s lead-time assumption; and on receipt, the delivered quantity and any quality rejections are recorded, because a rejected stone or a casting flaw that sends a piece back for rework effectively extends that order’s real lead time beyond what was planned.

Where this breaks down in practice is the sign-off step. A purchase order that requires manual approval for every SKU, regardless of value, either slows the whole system down until buyers start rubber-stamping without reading, or gets skipped entirely under time pressure, which defeats the reason a check was built in. A workable middle ground is a value threshold: orders below a set amount auto-send once triggered, orders above it route to a named approver, and that threshold gets revisited as order values change rather than left at whatever number felt right at launch.

Receiving deserves more attention than it usually gets in a description of this process. A purchase order that’s marked “received” the moment a box arrives, before anyone checks the contents against the order and confirms metal purity, stone count and any hallmarking requirement, hides a supplier quality problem until it surfaces as a customer complaint weeks later. Logging a rejection or a partial shipment against the original purchase order, rather than opening an untracked side conversation with the supplier to sort it out, keeps that supplier’s real lead-time and reliability record accurate for the next reorder point calculation.

What Breaks in a Jewellery Brand’s Reorder Point When a Gemstone Supplier Slips?

Picture a bench jeweller checking a work order against a stock report and finding the melee diamonds needed for a pavé setting are three weeks out instead of the ten days the reorder point assumed. That gap didn’t appear overnight: it built up over several reorder cycles where the supplier’s actual lead time crept past its quoted figure and nobody updated the safety stock to match, because the report used to track lead time only ever showed the average, not the trend.

The immediate consequence is a choice between three bad options: delay the finished piece and disappoint a customer with a fixed delivery expectation, most acutely a proposal-timed order, substitute a slightly different stone size or cut and hope the customer doesn’t notice or mind, or pull stones from a different SKU’s allocation and create a second, hidden shortage that shows up as a surprise a few weeks later.

The structural fix isn’t a bigger safety stock buffer applied blanket-wide, which just ties up more capital in loose stones sitting in a safe. It’s tracking each gemstone supplier’s lead time as a distribution rather than a single number, flagging the ones whose maximum has been creeping upward over the last several orders, and treating that trend as the trigger to widen safety stock for the specific SKUs that depend on that supplier, before the stockout happens rather than after.

How Do You Handle Reorder Points for Slow-Moving, High-Value SKUs?

A signature bridal piece that sells a handful of units a month doesn’t fit the daily-demand model built for a fast-moving chain SKU, and forcing it into that model produces a reorder point that’s either meaninglessly small or built on so little data that the average and the maximum are nearly the same number, understating real variability.

For these SKUs, two adjustments hold up better than the standard formula. Calculating demand on a monthly rather than daily basis is the first: a daily figure for a SKU selling two units a month is a fraction that hides more than it reveals. Second, treat the reorder point less as an automatic daily trigger and more as a production-slot decision: rather than reordering loose stock, you’re reserving the next available slot with your gem supplier or casting house, and the relevant question shifts from “how many units are on the shelf” to “how many production slots are committed versus how many months of expected demand those slots cover.”

A brand running a mixed catalogue of fast-moving fashion jewellery and slow-moving fine jewellery needs both models running side by side, tagged clearly by SKU, rather than one formula stretched to cover both. The mistake to watch for is applying the fast-moving SKU’s shorter review cadence to the slow-moving one too: a weekly reorder point check makes sense for a chain selling daily, but for a SKU selling two units a month, a weekly check mostly just re-confirms nothing has changed, and the review cadence is better set to match how often the underlying number could plausibly move.

What Does an Inventory Ordering System Cost to Run?

There’s no universal figure to quote here, because the real cost isn’t a subscription line, it’s the capital sitting in safety stock plus the labour of maintaining accurate lead-time data, and both scale with catalogue size and category risk rather than following a fixed price. What’s worth budgeting for, specifically, is the ongoing maintenance: someone needs to own supplier lead-time tracking, update reorder points on a review cycle, and audit safety stock against actual stockout and overstock incidents at least quarterly, because a reorder point set correctly at launch and never revisited degrades as supplier relationships, demand patterns and SKU count all shift under it.

Getting the balance wrong in either direction carries its own opportunity cost: safety stock set too wide across a fine-jewellery range ties up cash in a category where that stock is also expensive to insure and secure, while safety stock set too narrow produces stockouts that, for a bridal or gifting purchase, often convert into a lost sale rather than a delayed one, because the customer had a fixed date and couldn’t wait. Neither failure shows up as a line item on a P&L labelled “reorder point error,” which is exactly why it tends to persist uncorrected for longer than a cost with its own budget line would.

How Do You Verify a Reorder Point Is Set Correctly?

Check two outcomes against the reorder point’s own assumptions, not against a generic benchmark. First, how often did stock actually reach zero before the replacement order arrived over the last several cycles; a reorder point working as intended should produce close to zero stockouts, and if stockouts are happening on a specific SKU or supplier, that’s the signal the safety stock component is undersized for that line specifically, not evidence the whole system needs a wider margin everywhere.

Second, check how much stock is left on hand at the moment each new order arrives. If a SKU consistently still has a large amount of stock left when the replacement order lands, the reorder point is set too high for that SKU, and the excess is capital that could be freed up. The healthy target is stock running low, close to the safety stock floor, right as the new order arrives, not stock running out before it, and not stock sitting comfortably high when it does.

Run this check per SKU or per SKU group, not as one aggregate number for the whole catalogue, for the same reason forecast accuracy needs to be checked per SKU: an aggregate can look fine while individual lines are badly wrong in opposite directions. A jewellery brand’s fast-moving fashion line and slow-moving bridal line will very rarely need the same correction at the same time, and averaging the two hides which one actually needs attention.

A jewellery brand scaling past the $3M floor, where a spreadsheet-and-memory approach to reordering has started producing stockouts on fast movers and overstock on slow movers, is the specific stage this article is written for; see Pointerflow’s guidance for scaling brands for the broader operational picture this sits inside. This piece has deliberately stayed on reorder-point mechanics rather than how the underlying demand forecast gets built; that’s covered separately in our article on what an inventory planner does and where its forecasts go wrong.

Getting reorder points and the purchase order process right is fundamentally an operations problem, not a one-time calculation: it needs supplier lead-time data kept current, a review cadence that actually happens, and a purchase order process that doesn’t collapse into rubber-stamping under time pressure. That’s the coordination work Pointerflow’s ops automation service is built to take on for a growing jewellery brand.