The second purchase order is usually late for a boring reason: the first order was measured at the wrong resolution. A weekly sales total can look healthy while one SKU is weeks away from an empty shelf. Sell-through fixes that, but only when it is read by SKU, over a fixed window, and counted backward from real replenishment time.
Sell-through is a timing tool, not a report card
Take units sold in a period and divide by units received in the same period. That is the whole formula. The work is in the edges: set the window before the container lands, net out returns, and log markdown units separately. A rate propped up by clearance tells you how well a discount worked, not how much demand the product has.
Turnover and GMROI are useful after the fact. They describe how hard your capital worked last quarter. Sell-through tells you what is happening while you can still change the order. That difference matters most in a small assortment, where one dead SKU can absorb the open-to-buy you meant to spend on the next winner.
Avoid a universal pass mark. Set a reorder point for each SKU based on its velocity, the cover you want, and the forecast error you can tolerate. That rule can be a share of the initial order, a weeks-of-cover figure, or both. What matters is that it is written down before the launch, when nobody is defending a result yet.
Build the reorder date from lead time backward
Start with the date you want stock on the shelf. Subtract receiving and allocation. Subtract ocean freight, commonly 25-40 days depending on lane and season. Subtract production: 15-25 days for OEM on an existing mold at 1,000 units, or 30-45 days for ODM at 3,000 units with a new mold. For an OEM program, the factory and freight block alone can run 40-65 days before your warehouse touches a carton.
That block is the minimum head start. If the reorder point is crossed when cover is lower than that, you are not reordering; you are booking air freight or accepting a gap. Add buffer for the variables you cannot control: a missed vessel, port congestion, a factory holiday, or a retail reset that pulls stock forward. The buffer is a judgment call, not a number borrowed from another business.
Run the arithmetic per SKU. A slow fountain and a fast grooming kit do not share a clock, even when they arrived on the same purchase order. One can carry months of cover while the other runs dry in weeks. A single blended reorder date will be wrong for both.
Read weekly, by SKU, with one owner
Put three figures on one page: units on hand plus units in transit, units sold at full price, and open return reasons. Add weeks of cover, calculated as on-hand units divided by average weekly unit sales. Do it on the same day each week, from the same source, with one person responsible for the page. The goal is not precision; it is a consistent signal.
A blended rate hides the only problem worth catching. The category can look fine while the 6L camera feeder is sold out and the slowest all-in-one feeder and fountain is stacked to the ceiling. Those failures cost differently. The stockout loses sales you already paid to create. The overstock ties up cash and shelf space.
Segment before you summarize. Read the grooming vacuum kit as its own line, not as part of a generic pet-tech average. If you cannot explain in one sentence why a SKU is above or below its reorder point, you are reading the category, not the SKU.
Seasonality changes the date, not the method
Pet products have at least two demand bends: gifting and adoption. A smart feeder bought as a gift is often bought late, at a specific price, and returned for reasons that have little to do with daily use. A fountain bought after a new pet arrives is bought on a need date, often at full price. Both can pull the same SKU in different directions.
The calendar should move the order earlier, not the review rule later. If you expect a seasonal lift, you need production capacity and a vessel slot before the lift, which means the purchase order starts while cover still looks comfortable. Waiting for the sell-through line to bend upward is usually too late to catch the season.
Use last year's SKU data if you have it, and label the forecast as a forecast. Without a prior year, keep the first buy small and plan a fast follow. A seasonal bet made on a blended number is a guess with a deadline.
What to renegotiate when the numbers earn it
The first order proves the product can sell. The second is where you negotiate from a fact base. Bring four items: price steps tied to committed volume, payment terms that reflect a proven sell-through record, color or firmware configurations your competitors cannot stock, and written confirmation of warranty and certificate files per SKU. The standard warranty is 12 months; CE, FCC and RoHS files should match the exact model you are buying, not a catalogue page.
Do not ask for a lower unit price without offering something in return. A larger committed quantity, a narrower color range, a longer forecast window, or a deposit can all reduce the factory's risk. Good reorder terms are traded, not extracted. If a supplier cuts the price without changing any of those variables, ask what changed in the specification.
Renegotiate the boring parts too: carton markings, mixed-carton rules, spare parts, and the claim process. Those details cost less to fix before the PO than after a container lands.
When not to reorder
A rate that depends on markdowns should hold the order. If the units moved because the shelf tag was red, you have learned about price sensitivity, not repeat demand. Returns that cluster around one failure mode are another warning; more units will produce more of the same problem until the fix is confirmed in writing. A missing consumable path is the third: a fountain without filters, or a feeder without access to the parts that wear, becomes a support cost after the first sale.
A fourth signal is softer: the SKU only moves because it holds the best spot on the shelf. Move it one bay and watch for a defined period before doubling down. Placement can create a false positive that looks exactly like demand.
From a second PO to a private-label program
The sequence is simple even when the calendar is not. Start with stock units from 50 pieces. Move to OEM on an existing mold at 1,000 units and 15-25 days of production once the SKU has earned a repeat. Move to ODM at 3,000 units and 30-45 days when you need a new mold and your own industrial design. Each step buys a different kind of risk: stock buys assortment, OEM buys branding, ODM buys differentiation.
Do not carry the old SKU's record over to the new one. A new mold, color, or price point is a new read. Give it its own reorder point and review date. The program can share the shelf; it cannot share the evidence.
Send the SKU, target quantity, and the date you need stock on shelf, and we will map the production and freight block backward with you. Request a quote.
