TDABC time equations: the definitive reference

What exactly is a time equation?

A time equation estimates how long one execution of a process takes, as a function of the characteristics of the transaction:

Where T is the time consumed, b₀ is the base time for the standard case, each X is a driver (a characteristic of the transaction) and each b is the extra time that driver adds. Drivers can be quantities (number of lines), yes/no conditions (needs refrigeration: 0 or 1), or categories expressed as conditions.

In words: "this task takes so many minutes, plus so many more for each unit of this, plus extra if that applies."

The quiet power is additive structure. Classic ABC needed a separate activity, with its own surveyed percentage, for every variation of work. A time equation absorbs variation as terms, so one equation covers the simple case, the complex case, and everything between.

What does a worked example look like?

Take order picking in a distribution warehouse. All numbers below are illustrative, chosen for clarity, not benchmarks.

Picking time (minutes) = 3 + 0.8 × (order lines) + 6 × (refrigerated flag) + 12 × (export documentation flag)

The base pick costs 3 minutes. Each order line adds 0.8 minutes. Refrigerated items add a 6-minute detour to the cold room. Export orders add 12 minutes of documentation.

Now two real-feeling orders:

Order A, a domestic order with 5 ambient lines: 3 + 0.8 × 5 = 7 minutes.
Order B, an export order with 30 lines including refrigerated goods: 3 + 0.8 × 30 + 6 + 12 = 45 minutes.

Same "picking activity" in a classic ABC model; more than six times the work in reality. Averaging them, which is what survey-based allocation does, misprices both, and every customer who leans one way or the other.

To turn minutes into money you need one more number, the capacity cost rate.

How is the capacity cost rate calculated?

Capacity cost rate = cost of capacity supplied / practical capacity of resources supplied

Cost of capacity supplied is everything it costs to have the resource available: salaries and benefits, supervision, space, equipment, systems for that team or machine group.

Practical capacity is the time genuinely available for work, not the theoretical maximum. People deliver roughly 80% to 85% of paid hours once breaks, meetings and training are netted out; machines net out maintenance and setup. Kaplan and Anderson's practical shortcut of about 80% of theoretical capacity for people remains a sensible default.

Illustrative example: a picking team costs EUR 25,000 per month fully loaded and supplies 31,250 practical minutes. The rate is EUR 0.80 per minute. Order A above costs 7 × 0.80 = EUR 5.60 to pick; Order B costs 45 × 0.80 = EUR 36.00.

And because the model prices only the minutes consumed, the minutes nobody used remain visible: unused capacity, priced at the same rate, reported separately instead of smeared into product costs. That single feature changes more decisions than any other output.

How many time equations does a model need?

Fewer than newcomers expect. One equation per process, not per variation: picking is one equation, however many terms it carries.

A focused cost-to-serve model runs on a handful: order entry, picking and packing, delivery, invoicing, returns. A full mid-market profitability model typically lands between 10 and 40 equations. Our largest production model, processing 525,000 transaction rows for a logistics operator, does not need hundreds; it needs well-chosen terms in a modest set of equations.

The discipline that keeps models healthy: add a term when a driver visibly moves time and the data can carry it; add an equation only when a genuinely distinct process appears. Resist decorative precision.

What are the common mistakes?

Using theoretical instead of practical capacity. Dividing by 100% of paid hours understates the rate and silently buries unused capacity, the very thing TDABC exists to expose.

Surveying instead of observing. Time equations are estimated from observation, timestamps and interviews with the people who do the work, then validated. Importing classic ABC's percentage surveys rebuilds classic ABC's problems.

Too many terms. A term earns its place by moving minutes materially and existing in your transaction data. Five drivers nobody can populate model nothing.

Mixing units. Minutes in one equation, hours in another, per-order here and per-line there. Pick minutes, per transaction, everywhere, and audit for it.

Never revisiting the equations. Businesses drift. The virtue of TDABC is that updating means editing a coefficient, not re-surveying the company; a light annual pass keeps the model honest.

Forgetting the reconciliation. Total modelled cost should tie back to the GL for the resources in scope. If minutes times rates drifts far from the ledger, either capacity or coefficients need attention.

Where do the numbers for equations come from?

Three sources, in order of preference: system timestamps (WMS pick logs, ticket systems, call records), direct observation of a sample of transactions, and structured estimates from the people who do the work, cross-checked against payroll hours.

You do not need stopwatch precision. Kaplan and Anderson were explicit that estimates within reasonable tolerance beat surveyed percentages by a wide margin, because errors are visible and correctable at the equation level. Start rough, reconcile against capacity, refine the terms that matter.

This is also why a first TDABC model takes weeks, not quarters; the fuller build sequence lives in how to build a TDABC model, and the method comparison in TDABC vs ABC.

Fair questions.

What is a time equation in TDABC?

A short formula estimating the minutes one execution of a process consumes: a base time plus increments for the drivers that lengthen it, such as order lines or special handling. Multiplied by the capacity cost rate, it prices every transaction according to its real complexity.

What is the capacity cost rate formula?

Cost of capacity supplied divided by practical capacity of resources supplied. Practical capacity is genuinely available working time, typically around 80% to 85% of paid time for people, netting out breaks, meetings and training.

How many time equations does a TDABC model need?

One per distinct process, with variation handled by terms inside the equation. Focused cost-to-serve models run on roughly 5 to 10 equations; full mid-market profitability models typically use 10 to 40.

How accurate do the time estimates need to be?

Reasonable, not perfect. Because each estimate is explicit, errors are visible and correctable, unlike surveyed percentages. Validate by reconciling total modelled minutes against practical capacity and total cost against the GL.

Can time equations handle service businesses?

Yes, naturally. Onboarding a client, processing a claim or resolving a ticket all decompose into base time plus complexity drivers. Our peer-reviewed dialysis TDABC study applied exactly this structure in a clinical setting.