“Almost right isn’t good enough” for a mission-critical process; and a standard cost released on the wrong volume is precisely almost right.
— after Christian Klein, SAP, 2026

The bill nobody watches changeA planned cost is a stack of moving drivers

A standard cost looks like a single number on a material master. It is really a bill of moving parts, each ticking at its own speed; and the one that moves without anyone touching the factory is the one that hurts.

For a discrete-manufacturing SKU at the consumer-goods group we have followed through the Hormuz shock and the close, the planned unit cost is assembled from a familiar stack: direct material, from the bill of materials multiplied by planned prices; labour and machine time, from the routing multiplied by activity rates; energy; yield and scrap losses; and overhead, absorbed onto the unit through a planned activity base. Each driver has its own clock. Commodity and freight prices, still elevated from the Cape-of-Good-Hope reroute, and currency move daily. Energy moves hourly. Labour and machine rates are set quarterly. Yield drifts continuously as tooling wears. And overhead absorption is treated, by nearly every cost model, as fixed; right up until a volume miss makes it move violently.

This is the crux. When demand planning feeds a lower planned production volume for a SKU, the fixed pool of plant overhead is spread across fewer units, and the cost per unit rises — even though no rate, no material price and no routing has changed. The factory did nothing; the plan did everything. In a conventional model this surfaces months later as an under-absorption variance, discovered in CO-PA after the quarter it ruined. In an autonomic model it is a signal the moment the demand plan moves; and the cost is re-planned before the variance is ever born.

The signalsWhat continuous scanning reads in a cost model

The cost model’s signals arrive from three directions, and the discipline is to know which cadence each belongs to. The first is the input stream: purchase-price and info-record movements on BOM components, freight and duty on landed material, energy tariffs, FX on imported inputs, and activity-rate changes as labour and machine economics shift. The second is the factory stream: actual yield and scrap drifting from routing assumptions, machine downtime and cycle-time creep, activity confirmations diverging from plan. The third, the spine of this piece, is the volume stream flowing straight from demand planning: the planned production quantity that sets the absorption base. A change here moves no rate and no price, yet re-prices every unit through the denominator. Most cost systems are blind to it because it does not look like a cost event at all.

An autonomic cost model treats all three as live signals against the SKU’s current-planned (simulated) cost, and it is careful about what it does with each. Input and factory signals update the live estimate continuously and, where a threshold is crossed, raise a re-cost recommendation. Volume signals trigger the absorption reflex: recompute fixed cost per unit on the new base, flag the SKUs whose margin now breaches guard-rails, and propose the levers; re-rate the activity plan, shift production between lines or plants, re-sequence the campaign, or escalate a pricing or mix decision to a human. The standard itself does not move. The shadow does; loudly enough that the standard’s next release is a decision, not a surprise.

The spineLower the volume, and the cost rises by itself

Fixed overhead does not care how many units you plan; it only cares how few you divide it by. Spread a plant’s fixed pool across a smaller planned volume and the absorbed cost per unit climbs on a curve, steeply, once volume falls far enough. Exhibit 1 is the reflex at the heart of this article: as demand planning cuts a SKU’s planned production, its variable cost per unit holds flat while its absorbed fixed cost per unit rises, pushing total unit cost through the margin guard-rail; and the live estimate reacts the moment the volume moves, while the released standard, still built on the old volume, says nothing.

The choreographyFrom demand feed to standard release, continuously

Cost planning has a predecessor it cannot control, the demand plan that sets its volume, and a successor it must not corrupt: the released standard that variance accounting depends on. An autonomic cost model lives between them, scanning continuously and re-costing a live estimate, while treating the standard’s release as a governed event. Exhibit 2 lays out the arc: the demand feed on the left, the continuous live-cost plan and its signals in the middle, the decision levers, and the governed standard release on the right; with a governance strip for the human gates and a bottom loop in which the estimate corrects continuously and the standard is re-released only on cadence.

Two speedsThe shadow that moves and the standard that waits

The discipline that keeps this honest is the separation of two clocks. The live estimate, a continuously simulated cost, tracks every signal in near real time; it is what the fabric reasons on and what triggers decisions. The released standard is deliberately slow: re-issued on a governed cadence, annually and on material-change triggers, through a controlled costing run with marking and release. The gap between them is not an error to be eliminated; it is the planned variance being tracked in advance — visibility a conventional model only gets after the fact. Exhibit 3 shows the two clocks side by side.

The spineHow signals flow from BOM and demand to the cost of record

The cost-planning architecture joins two streams that finance rarely sees together: the production model that determines what a unit consumes, and the signal set that determines what those consumptions cost: including the planned volume that sets the absorption base. Both meet at the Knowledge Graph, which resolves any signal to the material, the BOM component, the work centre and activity, the cost centre, the SKU and the margin line it touches. Reasoning agents recompute the live estimate, re-base absorption on the new volume, and simulate the resulting variance; the Agent Hub bounds them and holds the one gate that must stay human: the marking and release of the standard. Write-backs land in the planning and costing tables; the released standard, the cost-component split and the profitability view record the result. Exhibit 4 names the tables at each hop (SAP News Center, 2026)(Constellation Research, 2026).

The correction mechanismTwo loops: sharpen the estimate, re-release under control

The cost model corrects on two loops. The estimate-correction loop compares the live estimate with what actually happened; realised absorption in the cost-centre actuals, production variances in the Universal Journal, yield against routing; and pulls the shadow cost toward reality continuously, so the next decision is made on a truer number. The assumption-recalibration loop goes deeper: it feeds actual yields, rates and, above all, actual production volumes back into the driver assumptions the estimate is built from, with a drift monitor that flags when a routing time or a scrap rate has quietly stopped matching the floor. Neither loop touches the released standard. That changes only when a human runs the governed costing sequence; and by then the release is an informed decision, because the two loops have been narrating the gap all quarter. The estimate learns every day; the standard is re-released on purpose.

The boundaryWhat the reflex may model but never release

The bright line here is the release of the standard, and it belongs to a person. Re-marking a SKU’s standard cost, changing activity rates, re-basing an overhead pool, a make-versus-buy or line-shift decision with capital or headcount consequences, any move that re-prices the balance sheet’s inventory; these route through the release gate on an auditable trail, exactly as the parent essay’s decision-authority matrix demands. Three honest constraints keep the ambition grounded. Cost planning is only as good as its master data: a wrong routing time or an out-of-date BOM mis-costs every unit silently and forever, so the fabric’s first job is often to surface stale master data, not to re-cost on top of it. The absorption signal is only as good as the demand plan feeding the volume, which is why this piece sits deliberately downstream of demand planning in the series. And re-basing absorption must never become a way to launder under-utilisation; spreading a shrinking volume’s fixed cost is a signal to act on capacity, not a number to quietly re-absorb. The measure is the series’ measure: not how much the model automates, but how little it must escalate; with the release of the cost of record held, deliberately, on the human side of the line.

A question for cost and plant controllers: when demand planning halves a SKU’s planned volume, does your system tell you the unit cost has moved, today, in the plan, or does it wait to surprise you as an under-absorption variance next quarter?

#ProductCosting   #ManufacturingFinance   #CostAccounting   #SAP   #SAPCO   #AgenticAI   #FPandA   #StandardCost

SourcesReferences

Methodological note. A functional-and-technical companion to The Autonomic Enterprise, continuing the consumer-goods group across its discrete-manufacturing plants. Figures in Exhibits 1 and 3 are directional models, not measured series. Table names (PBIM/PBED, PLAF, MARC, MAST/STKO/STPO, MAPL/PLKO/PLPO, CRHD, KP26, MBEW/STPRS, KEKO/KEPH/CKIS, COSS/COSP, ACDOCA, and the CK-series costing transactions) are representative of the standard SAP S/4HANA Production Planning and Product Costing model; SAP Analytics Cloud and SAP Enterprise Planning are shown at representative hops, and real implementations differ. Nothing here proposes autonomous release of standard costs: marking and releasing the standard is, by design, a governed human act, so that variance analysis remains valid. This is analysis, not accounting, audit or investment advice.