Broken down by week, the monthly production report in a large number of plants carries a single recurring shape: a first week meaningfully below the average, a second week near it, a third week rising, and a fourth week that carries a disproportionate share of the month's total output on its own. The relationship between that curve and the actual intra-month distribution of demand is typically weak; the peak coincides instead with the date on which shipments close, invoices are cut, and the sales commission becomes payable. Overtime hours, temporary labor calls, defect density in the quality log, and the count of expedited orders placed on suppliers all track the same curve almost point for point. What gets discussed in the planning meeting, however, is rarely the curve itself but a single customer order that missed the month-end window and the question of whose account it lands in.
A second and less frequently examined pattern appears not in volume but in mix. While the weekly plan carries several product families side by side on paper, the actual sequence on the line is built by running one reference for several consecutive days before moving to the next; the commercial side, having quoted a delivery commitment for a product that is not on the line that week, extends the promise according to the expected wait for a slot rather than the product's real processing time. The lead time communicated to the customer thus becomes the sum of two distinct components: the time production actually consumes, and the frequency with which the schedule returns to that product. The second component is usually larger than the first, yet appears as a separate line in no cost report.
These two patterns share a name: heijunka, the leveling of a production schedule in both volume and mix so that the same output composition repeats at even intervals throughout the period. The logic standing against it is not irrational. Enlarging batches spreads changeover time across more units and lowers setup cost per unit, and that arithmetic holds in a world where changeover time is treated as a given. The conditions under which the shortcut remains functional are clear: where changeover duration is large relative to run time, where the product range is narrow, and where inventory at the end of the line is cheap, batch logic genuinely lowers total cost. The difficulty lies not in the shortcut but in its persistence once conditions shift — as the range widens, order sizes shrink, and customer delivery windows narrow, batch discipline in most plants remains where it was.
When conditions do shift, the dominant effect is less a loss of efficiency than a corruption of signal. An unleveled schedule becomes the origin of demand variability inside its own supply chain; the order series reaching suppliers is more volatile than the series generated by the end market, and that gap widens by some increment at each tier moving upstream. Two kinds of variability warrant separation here: the exogenous variability produced by the market, and the endogenous variability produced by the schedule itself. The first is managed, the second is designed — and being designed, it is the only category of variability that can be removed rather than absorbed. A defensible share of the safety stock a plant carries exists to protect it not from the market but from its own calendar.
The balance sheet expression of this tendency hides not in the total value of the inventory line but in the movement of that value across the month. Finished goods run low and work in process runs high at the start; the relationship reverses by month-end. Even where the period-end photograph shows an acceptable inventory turn, the working capital requirement must be financed against the mid-month peak, which means the short-term facility in use is priced against the maximum rather than the average. The cash conversion cycle lengthens from both ends at once, through expedited orders that must be paid closer to cash and through month-end collections that slip into the following period.
The cost on the capacity side is more durable. A line dimensioned to clear the peak week is, by definition, underused in the average week; the fixed asset base grows, the depreciation charge is locked in, and the next capacity request reaching the investment committee is justified by the bottleneck in the peak week rather than by average utilization. That same week is where quality cost concentrates: rework rates, scrap, and customer returns rise predictably on a line operating under schedule pressure, and part of that cost is carried forward into later periods as a warranty provision. The combined overtime premium and temporary labor line typically runs above the incremental headcount cost that a leveled schedule would have required.
The table at which this pattern looks most expensive is not the company's own cost report but the review table of a sale or financing process. Revenue concentration in the closing days of the month opens a cut-off risk heading in the quality of earnings analysis; the buyer's adviser will seek confirmation that title and risk transferred on shipments recorded near period-end, and that request alone extends the closing calendar. The net working capital reference level turns into an argument about which date should govern an inventory line that swings by nearly a factor of two within the month, and the negotiation usually knots around who absorbs the difference between the average and the period-end figure. Where an earn-out is measured quarterly, the same swing creates an additional surface for dispute at the boundary of each measurement period.
The valuation effect is quieter. The question ultimately posed in diligence is whether this plant's output is repeatable independently of the extraordinary effort visible at month-end. Where the answer can be evidenced through plan-versus-actual adherence records, operational performance is priced as a capability belonging to the company; where the answer rests only on monthly aggregate production figures, the same performance reads as an outcome dependent on individuals and effort, and that reading enters the price as a discount, a condition precedent, or an expanded representation and warranty package. A parallel effect appears on the credit side: with covenant tests locked to quarter-end, and quarter-end being structurally the point of highest inventory and highest short-term borrowing, the ratios present themselves at their weakest precisely on the test date.
Establishing leveling depends not on a more disciplined operations team but on designing four components separately. The first is an explicit definition of the smallest interval at which the schedule repeats itself — unless the frequency with which each product returns to the line is written down as a target, leveling remains an intention. The second is treating changeover duration as an engineering objective rather than a given; because batch size is a derivative of that duration, an instruction to reduce batches without first reducing the changeover produces only additional downtime on the line. The third is the separation of the commercial calendar from the production calendar: as long as commission accrual, shipment close, and reporting cut-off fall on the same day, every leveling gain on the line is surrendered at month-end. The fourth is making the location and purpose of inventory a deliberate decision, since a leveled schedule does not eliminate inventory but relocates it.
BEIREK's intervention in a structure of this kind begins not with an improvement proposal but with the establishment of three records. The first is the repeat-interval record: the frequency with which each product family enters the schedule, target interval and achieved interval held side by side on a weekly basis. From the moment that record exists, the argument over whether a delay originated in capacity or in slot frequency ends. The second is the changeover record: the measured duration of each transition, tracked with setup and adjustment components separated, because every decision about batch size is a derivative of that number. The third is the deviation record — the date on which the leveled schedule was broken, at whose request, and on what stated ground, captured at the moment the decision is taken rather than at the moment its consequence appears.
The governance layer sitting on those records answers two questions: who holds the authority to break the schedule, and above which threshold that authority escalates. The arrangement that works in practice assigns deviation authority to a single role, reviews that role's decisions in aggregate on a monthly cycle, and defines the review metric as plan-versus-actual adherence rather than total output. The same rhythm extends to the commercial side: unless delivery window commitments, liquidated damages caps, and order acceptance rules are aligned with the line's real repeat interval, the contract continues to carry a promise the operation cannot support. What is ultimately placed on the review table in the next financing or sale process is precisely this set of records, not the production figures themselves.
The maturity of a production system is measured less by how much it can produce in its busiest week than by how narrow the gap is between its busiest and its quietest week; and that gap is less a performance indicator than an indication of the degree to which the company sets its own calendar through its own decisions.