Par Level Management: Your Pars Are Fine, Your Counts Are Stale

Every par level management conversation I have sat in on ends up being about the par.

What should the number be. Should it flex by day of week. Should the ED run a different par than the floor above it. Somebody pulls up twelve months of issue history, somebody else points out that the history is distorted by the last shortage, and the meeting ends with a decision to raise 40 line items by 15% and revisit next quarter.

The number is almost never the problem.

A par level is a target. A count is a measurement. You can tune a target as precisely as you like, and if the measurement feeding it is a week old and taken by whoever happened to walk into the supply room with a clipboard, the precision does nothing. You have built a careful model on top of an uncertain input and reported the model's confidence as if it were the input's.

The two numbers that tell you how much faith your system deserves

Start with what the regulator asks for. Under 21 CFR 1304.11, a DEA registrant has to take a new inventory of all controlled substances on hand at least every two years, and dispensers have to make an exact count of the contents of opened Schedule I or II containers. For Schedule III through V, an estimated count is acceptable unless the container holds more than 1,000 units.

Read that as what it is: a floor. Once every two years, with exact counts reserved for opened Schedule II containers and estimates accepted on most of the rest. That is the regulatory minimum, and a surprising number of supply operations have quietly adopted it as the operating standard for everything else in the building. The controlled substances get counted properly because somebody will ask. The gauze, the trays, the catheters and the single-use instruments get counted when there is time.

Now the second number. As of the second quarter of 2026 there were 227 active drug shortages in the United States, the third consecutive quarterly increase, per the shortage data maintained by ASHP and the University of Utah Drug Information Service.

That is the number that breaks par level management, and not for the reason people assume. The problem is not that you run out. The problem is substitution churn. A product goes short, you qualify an alternate, the alternate arrives in a different pack size, and the par written on the bin label is now a par for something you no longer stock. Multiply that across a few hundred line items over a few years and a meaningful share of your pars are archaeology.

Par level optimization is measurement-limited, not model-limited

Here is the turn, and it is the whole argument.

A rough par measured continuously beats an optimized par measured monthly. Not marginally. Structurally.

A par level only does work at the moment someone compares it to reality. Between those moments, the par is inert and your actual position is unknown. If the comparison happens every 30 days, you are running 30 days of unknown position against a target tuned to two decimal places. If the comparison happens every time product moves, the par barely needs to be right. Any reasonable target plus a current measurement will outperform a perfect target plus a stale one, because the reorder decision is made on the measurement, not on the target.

Which means the lever everybody pulls, the par tuning exercise, is the second-order lever. The first-order lever is time-to-discovery: how long between a supply position changing and the system knowing it changed. Almost nobody reports that number. Most materials managers I talk to cannot say what it is for their own building, which is a strange thing to be missing when it governs the performance of everything built on top of it.

What to measure instead of re-tuning the pars

Four changes, none of which require a budget cycle.

• Report time-to-discovery per supply location. Not per facility. The clean room and the third-floor supply closet are different operations with different traffic and different honesty about restock logging. One average hides both.
• Split your stockouts by cause before you raise anything. Two buckets: the par was genuinely too low, and the par was fine but the recorded count was wrong. These look identical on a stockout report and they have opposite fixes. Raising a par to compensate for a bad count is how supply rooms end up overstocked and still running out.
• Audit the bins nobody touches. Fast-moving items self-correct because somebody notices an empty shelf. Slow-moving stock drifts quietly, and in a supply room the drift has an expiry date attached to it. A weighing pass over the long tail is cheap: on a Cloudbox scale station, a shelf of slow-moving containers becomes a measured count in a single trip down the aisle, with a recorded quantity you can reconcile against instead of a tally sheet.
• Stop tuning pars on data you would not defend in a survey. If you would hesitate to show a surveyor the count history behind a par, that par is not ready to be optimized. Fix the measurement first and the tuning exercise gets much shorter.

Where weight-based automation changes the input

Cloudbox is scale-based inventory automation using smart container technology, built for operations where somebody is currently counting by hand and writing it down. The container is set on the Cloudbox scale station. The station weighs it, and because the per-unit weight for that item is on file, the weight converts to a unit count and the measured quantity is recorded. No tally, no transcription, no decision about whether to round.

One station serves a whole room of containers and line items. It moves through the supply room to the product, so a par check becomes a weighing round instead of a counting round, and the recorded position updates as it goes.

For a healthcare supply operation the practical gain is that time-to-discovery collapses from weeks to the length of one pass, and the count history behind every line item is measured instead of remembered. Your pars get tuned on real data. Your expiry write-offs shrink because the slow-moving stock is visible. And when someone asks how you know what is in that room, the answer is a measurement with a timestamp.

The pars were never the hard part. Let us show you what measured looks like.