The Activity-Based Costing
Adoption Playbook.

Score each yes = 1, no = 0. Your total drives where to start — answer all ten to see your tier.

• 8–10: Strong readiness. You can start the 30-day quick-start in Module 04 immediately, in parallel with planning the full rollout.
• 5–7: Mixed readiness. Spend the first two to four weeks closing your biggest gaps (usually #1 sponsor or #3 routings) before starting the pilot.
• 0–4: Pre-readiness phase. Don't start activity-based costing yet — start with a data-foundation project (build routings, instrument one cell, establish work-order traceability) and revisit in 90 days. Trying to cost activities without the foundation produces numbers the team won't trust, which kills the program before it produces value.

Every level in this ladder produces an event that the costing engine can consume. The engine doesn't know or care which level the event came from — a “unit completed” event from a tablet tap and one from a PLC look identical downstream.

Most plants will live across multiple levels: Level 1 at the assembly bench, Level 3 on the legacy mill, Level 4 on the modern 5-axis cell, Level 0 (paper) on the inspection desk nobody has touched yet. That's fine. Start where you are. Lift the lowest-value, highest-effort gaps later.

Don't try to instrument everything at once. The order of value:

1. Work order start / stop (Level 1 minimum) — without this you can't attribute anything to a specific work order. A tablet tap is sufficient.
2. Good / scrap / rework counts (Level 1 minimum) — without these you can't compute yield, scrap cost, or first-pass quality.
3. Machine hours per work order (Level 3+ ideal, Level 1 acceptable) — the single largest cost driver in most discrete manufacturers.
4. Setup events (Level 1 or 3) — a high-value non-value-added activity that standard costing hides. Cheap to capture; high analytical payoff.
5. Downtime events with reason code (Level 1 acceptable, Level 3+ better) — feeds the cost-of-downtime story and OEE simultaneously.
6. Energy consumption per machine (Level 3 ideal) — only matters meaningfully on energy-intensive operations such as coating, heat treat, and large machining.
7. Quality inspection time (Level 1 or 4) — important for cost-of-poor-quality analysis; often manual.
8. Material consumption deltas vs BOM (Level 1 acceptable, Level 4 ideal) — feeds material-variance analysis.

Older or simpler equipment often has no PLC, no Ethernet port, and no signals to read. You do not have to retrofit a full PLC to get costing-quality data — add a small remote I/O module and wire off-the-shelf sensors to it. This is where most “we couldn't do activity-based costing because our floor isn't modern enough” objections collapse.

Typical hardware

Total per machine: $500–$2,000 depending on signal count and complexity. That is one or two orders of magnitude cheaper than a full PLC retrofit, and it typically pays back in three to six months on a single high-cost SKU.

What to wire first

1. Machine-on signal. A current clamp on the motor lead. Tells you when the machine is actually running, independent of whether the operator remembered to log start and stop. Single biggest improvement in machine-hour accuracy.
2. Cycle-complete signal. A proximity switch on the spindle, ram, or door, or a photoelectric beam break on the discharge chute. Gives you part count without operator effort.
3. Energy consumption. A kWh meter on the disconnect. Required for accurate energy-cost attribution and ESG reporting.
4. Vibration or temperature (optional). Doesn't feed costing directly, but feeds the predictive-maintenance and equipment-ROI stories.

When this approach is right

• The machine has five or more years of useful life remaining.
• A full PLC retrofit cannot be justified in the next 12 to 18 months.
• You need costing-quality data from this machine within the next 90 days.
• You can run two-conductor signal wiring, or wireless I/O, to the remote I/O location.

When this approach is wrong

• The machine is being replaced within 12 to 18 months — instrument the new one.
• Operator interaction is already frequent (one tap per unit) — Level 1 tablet capture is cleaner, and it already exercises the operator interface.
• The wire run is impractical and your floor doesn't permit wireless I/O — fall back to Level 1 capture and accept lower precision.

A real plant looks like this six months into adoption:

Cell A (assembly bench):       Level 1 — operator tablet
Cell B (legacy CNC mill):      Level 3 — Opto 22 RIO with 4 signals
Cell C (modern 5-axis):        Level 4 — PLC OPC UA bridge
Cell D (CMM inspection):       Level 1 — operator tablet
Cell E (powder coat booth):    Level 3 — energy meter + cycle switch
Cell F (manual deburr):        Level 0 — daily paper summary, OK for now

This is the goal. Don't aim for uniform instrumentation — aim for enough instrumentation per cell to trust the activity volumes that matter for costing.

Cell selected (rule: the one with the most painful unexplained variance). One operator champion identified. One finance lead assigned. Product family of three to five SKUs identified.

Use the starter dictionary in Module 06. Map the activities to existing cost pools in the GL. The activity list is signed off by operations and finance. No more than six activities.

Begin Level-1 event capture. By the end of the week, the operator successfully taps work-order start and stop, good and scrap counts, and at least one downtime event during a full shift.

Compare to standard costing. One page showing cost per unit by activity for each SKU running on that cell, side by side with standard cost. At least one “surprise” identified.

Goal: team formed, sponsor committed, current state documented, quick-start cell selected and launched.

• Define the “why” in writing. Name specific problems: margin erosion, variance blind spots, quoting accuracy, customer-profitability questions. Run the worked example in Module 01 against your own SKU data — find your SKU-B.
• Assemble the team. Sponsor: VP Finance or COO. Core: plant controller, production manager, IT/data lead, one shop-floor supervisor. Optional: an outside costing advisor for the first cycle.
• Audit current state. Map GL → cost centers → product lines. Inventory data sources: ERP, MES, PLCs, time tracking, maintenance, energy. Score the readiness diagnostic in Module 02 and address the obvious gaps.
• Launch the quick-start track described in Module 04.

Goal: the activity dictionary and cost-allocation model are defined and signed off by operations and finance.

• Build the activity dictionary (Module 06). Start with the twelve-activity starter template and adapt — most plants land at 10 to 18 activities.
• For each activity, define the cost pool, the cost driver, the capture method, and the driver volume estimate.
• Design the cost-allocation model: per-activity rate = pool ÷ driver volume. Run it on last quarter's actuals as a paper exercise.
• Confirm the GL reconciliation approach (Module 07).

Goal: the costing model is configured, instrumentation is deployed, and a multi-cell pilot is running in parallel with legacy costing.

• Configure activities, cost pools, and drivers in the costing model.
• Deploy data capture: Level 1 terminals, Level 3 remote I/O modules, and Level 4 PLC bridges.
• Verify the event stream is receiving signals from each cell.
• Expand the pilot from one cell to three to five cells covering machining, assembly, and finishing.
• Run activity-based and standard costing in parallel for six to eight weeks, and investigate every variance over 10%.

Goal: activity-based costing is live across the full product portfolio, the team is operating it day to day, and governance is established.

• Validate accuracy by tracing three to five actual products from raw material to shipment. Finance reconciles the activity totals to the GL within 2%.
• Roll out by product family or plant area. Train cost analysts on interpreting the reports.
• Establish governance: quarterly reviews, process-change updates, and monthly non-value-added cost reviews.
• Close the reporting loop by making sure insights reach pricing, quoting, and capital decisions.

1. Every activity must have a moment of capture. If no one taps a button, scans a barcode, or fires a signal, the activity doesn't exist — it's a guess.
2. Drivers must be honest. “Allocated by direct labor hours” is often the same misallocation standard costing already makes.
3. Start with 10 to 15 activities. Resist the urge to add more. Start coarse and refine once you see what the data tells you.

Use this as a starting point for a machining and assembly contract manufacturer. Adapt it — yours will probably land at 8 to 16 activities.

Twelve activities. Eight of them are capturable at Level 1 today. Three need remote I/O or PLC integration to be accurate. One (5-axis machining) presumes you already have a modern controller.

A driver answers one question: what increases this cost pool's consumption by one unit?

Review it quarterly. Things that trigger a dictionary change:

• A new product line with materially different process characteristics.
• A new cell or capability — adding waterjet or laser, for example.
• Driver volumes off plan by more than ±15%.
• An operator or supervisor reporting that an activity definition “doesn't match what we actually do.”
• Cost pool composition shifts — depreciation dropping off a major asset, for example.

Resist mid-quarter changes — they break period comparability. Save them for the quarterly cycle unless the issue is blocking a major business decision.

Both have a place.

Avoid the temptation to “fine-tune” rates more often than quarterly during the first twelve months — it makes period-over-period comparisons impossible and undermines trust in the numbers.

At period close, total allocated activity costs should reconcile to the manufacturing-overhead pools in your GL within an agreed tolerance.

Tolerance band — typical mid-market expectation

• First 90 days of operation: ±5%
• After 90 days: ±2%
• Mature (12+ months): ±1%

The reconciliation step

1. Sum total allocated activity costs for the period.
2. Sum the source overhead GL accounts.
3. Compare. If within tolerance, post; if outside, investigate.

Common causes of over-tolerance variance

• A new cost in the GL that wasn't mapped to an activity.
• Driver volume captured differently than budgeted.
• Period boundary issues.
• Genuine new variance — a process change, an equipment failure.

The reconciliation is itself a diagnostic: investigating variances often exposes operational issues that wouldn't have been visible otherwise.

Two pragmatic patterns:

Every quarter, formally:

• Review driver volumes against forecast — flag anything off by more than 15%.
• Review cost pool composition — has anything material moved in or out?
• Review the activity dictionary — does it still describe how the floor works?
• Recalculate standard rates from the prior quarter's actuals.
• Sign off the next quarter's rates with the controller and the operations manager.

This meeting should take 60 to 90 minutes. If it routinely takes longer, the dictionary or the pool definitions are probably over-decomposed.

Activity-based costing produces value only when its outputs drive operational and commercial decisions. The four loops to close:

If the reports are produced but none of these decisions reference them within 60 days of the first reliable monthly close, something is wrong. Usually the reports aren't reaching the decision-maker in a format they can act on — fix that, don't ship more reports.

You don't add activities because they exist on the floor. You add them when all three of these are true:

• A meaningful cost pool isn't being absorbed by any existing activity, and
• the activity has a defensible driver and a capture method, and
• the cost being captured is material — 1% to 2% or more of total manufacturing cost.

If any of those fail, leave the cost in a broader pool. Splitting hairs is the fastest way to bury the activity dictionary in detail nobody uses.

Start with 10 to 15 activities. Resist additions for the first two quarters.

Activity-based costing is better than standard costing, not perfect. Target ±5% in the first 90 days.

Involve operators and supervisors in defining the activities. The people who fire the capture events have to believe in them.

Reconcile to the GL from the first period. Reconciliation is not optional.

The quarterly driver-rate review is mandatory.

Every activity needs a moment of capture — a tap, a scan, or a signal.

Run the 30-day quick-start in Module 04 on one cell at Level 1.

Drivers drift. Recalculate quarterly.

OEE optimizes machine performance. Activity-based costing optimizes business performance. A plant can be excellent at one and broken at the other.

Maintain executive sponsorship by running the SKU-A vs SKU-B exercise from Module 01 against your own SKUs.

Week 1

• Executive sponsor named in writing
• Pilot cell selected
• Product family identified (3–5 SKUs)
• Operator champion identified
• Finance lead assigned

Week 2

• Activity list defined (4–6 activities)
• Cost pools mapped to GL accounts
• Drivers defined per activity
• Sign-off from operations and finance

Week 3

• Operator terminal deployed
• Operator trained and using the terminal
• At least one downtime event logged
• Production events confirmed in the event stream

Week 4

• First weekly costing report generated
• Standard cost comparison run
• Variance over 10% identified
• Report reviewed with the sponsor
• Go / no-go decision made

For installers and IT/OT teams evaluating instrumentation options.