Business Growth

Margin by Service Line: What Activity-Based Costing Reveals About Agency Profit

Ask a founder which service line is most profitable and you usually get an answer based on revenue, not margin - because most service firms have never measured cost at the service-line level. The research says that blind spot is expensive. Robert Kaplan and Steven Anderson's work on time-driven activity-based costing, published in Harvard Business Review in 2004, showed firms relying on blended costs systematically misjudge which products, services, and customers make money, and offered a simplified method requiring only two estimates: the cost per unit of capacity time, and the time each activity consumes (Kaplan and Anderson, 2004). For agencies and consultancies running multiple offers through shared teams, that method is the difference between managing a portfolio and guessing at one. This article applies the evidence to service-line economics.

Joshua Agonya Pi'Rwot

By Joshua Agonya Pi'Rwot

Founder, Business Growth Accelerator

Executive summary

Most agencies price every service line off one blended cost assumption, so winners quietly subsidize losers. Kaplan and Anderson's time-driven activity-based costing research shows how to see true margin by service line.

Section 1

The five challenges at a glance

Service-line margin blindness has a consistent anatomy across agencies, consultancies, and studios. Revenue is tracked by service while costs are tracked by category, so the two never meet; senior time - the scarcest resource - is allocated by urgency rather than economics; pricing inherits historical rates instead of measured cost; unprofitable clients hide inside profitable service lines and vice versa; and traditional costing projects collapse under their own survey weight before producing answers. The table summarizes the evidence for each. The methodological anchor is peer-reviewed-adjacent practitioner research from Harvard Business School (Kaplan and Anderson, 2004); the financial-pressure context comes from SPI Research's vendor-published benchmark of 403 professional services firms, flagged accordingly (SPI Research, 2025). The rows interact: blended-cost pricing creates the cross-subsidy, senior-time misallocation deepens it, and margin-blind growth scales it - which is why firms that fix only their rate card without fixing measurement relapse within a year. The encouraging finding from the implementation literature is that the measurement fix is far lighter than founders assume: the time-driven method was designed specifically because its survey-based predecessor demanded more analytical capacity than most organizations could sustain (Kaplan and Anderson, 2004), and a quarterly spreadsheet snapshot is sufficient at boutique scale.

Section 2

Challenge one: the blended-rate illusion

The standard small-firm costing model is one blended hourly cost - total payroll plus overhead divided by total hours - applied to every service line. The Kaplan and Anderson research line explains why that average is not neutral: it systematically overcosts simple, standardized work and undercosts complex, exception-heavy work, because complexity consumes disproportionate amounts of expensive senior time, coordination, and rework that the average smears across everything (Kaplan and Anderson, 2004). In agency terms: the productized SEO retainer delivered by trained mid-level staff probably earns more than your books show, while the bespoke strategy engagement that routinely pulls founders into unbilled crisis calls probably earns less - possibly nothing. The distortion compounds at the pricing layer. Rates inherited from history get sanity-checked against the blended cost, pass the check, and perpetuate the cross-subsidy; the firm then markets hardest the line that feels successful, which may be the one quietly destroying margin. The macro context removes the slack that once forgave this: SPI Research's 403-firm benchmark - vendor-published, flagged as such - shows professional services EBITDA at 9.8%, the lowest in over a decade (SPI Research, 2025). At single-digit firm-level margins, a single service line running at negative-15% margin can consume the entire profit of two healthy lines, and no amount of utilization discipline or sales effort will fix what is fundamentally a measurement failure.

Section 3

Challenge two: why traditional costing failed and what replaced it

Activity-based costing was supposed to solve this decades ago, and in large firms it often collapsed under its own machinery: armies of employee surveys asking people to allocate their time across dozens of activities, data that was stale on arrival, and - critically - self-reported allocations in which, as the research observed, people rarely report idle time, so reported percentages conveniently summed to fully utilized days (Kaplan and Anderson, 2004). Small service firms that attempted scaled-down versions usually abandoned them within two quarters. Kaplan and Anderson's time-driven revision is the method that actually fits a 10-50 person firm because it inverts the data demand. Instead of surveying everyone about everything, managers estimate two numbers: the cost per time unit of supplying capacity - roughly, a department's fully loaded cost divided by its practical available minutes - and the time each defined activity consumes, captured as simple time equations (Kaplan and Anderson, 2004). Multiply them and every project, service line, and client acquires a defensible cost. Two properties matter most for founders. First, the model prices unused capacity separately rather than smearing it into product costs, which connects directly to utilization economics: bench time becomes a visible line item instead of an invisible tax on every job. Second, time equations accommodate complexity drivers - a rush job, a new client, an extra revision round - so the cost of difficult clients finally appears in the numbers rather than in delivery-team burnout.

Section 4

Challenge three: client-level profit hides inside service-level averages

Even firms that measure service-line margin usually stop one level short: the client. The TDABC literature's most repeated practical finding is that profitability is wildly skewed at the customer level - implementations such as the Kemps dairy case documented in Kaplan and Anderson's work showed companies discovering that a meaningful share of customers were unprofitable once order patterns, service demands, and exceptions were costed honestly, and that modest behavior changes or repricing converted many of them (Kaplan and Anderson, 2007; HBS Working Knowledge, 2007). Service firms exhibit the same skew with sharper teeth, because the expensive resource - senior attention - is the one demanding clients consume most. The familiar profile: a large legacy account on discounted rates, heavy meeting cadence, scope drift tolerated for relationship reasons, and first call on the founder's calendar. On blended numbers it looks like an anchor client; on time-driven numbers it is often the firm's largest unpriced cost center. Goldratt's constraint lens explains why the damage exceeds the account itself: every senior hour the demanding client absorbs is an hour unavailable to the highest-margin work, so the true cost is the margin of the work displaced (Goldratt, 1984). The managerial implication is a two-axis view - margin by service line, then margin by client within line - because the intervention differs: weak lines get repriced, repackaged, or retired; weak clients get behavior changes, rate corrections, or graceful exits.

Section 5

Innovative solutions

Modern service firms have adapted TDABC into lightweight operating practices. Time equations as delivery templates: for each service line, define the standard activity sequence and expected minutes per step, with explicit modifiers for known complexity drivers - new client, rush timeline, extra stakeholders. This doubles as a scoping tool: the same equation that costs the work prices the proposal. Quarterly margin snapshots instead of perpetual systems: rather than maintaining a live costing engine, small firms run the TDABC math quarterly from time-tracking exports - cost per available hour by role, multiplied through actual logged hours per line and client. One analyst-day per quarter typically suffices for a sub-50-person firm. Complexity surcharges made explicit: once complexity drivers are measured, they migrate into the rate card - rush fees, revision-round limits, stakeholder-count pricing - converting previously absorbed costs into either revenue or deterred bad-fit demand. Unused-capacity reporting: following the TDABC principle of pricing idle capacity separately (Kaplan and Anderson, 2004), bench cost is reported as its own line, keeping pricing honest and pointing capacity decisions at the real problem. And kill-or-fix reviews: each quarter, the lowest-margin line and lowest-margin clients get an explicit decision - reprice, redesign delivery, or retire - because measurement without a forcing decision cadence reverts to wallpaper. Together these practices make portfolio economics a routine management input rather than a one-off consulting project.

Section 6

Solution framework: a founder-scale TDABC implementation

A 10-50 person firm can implement credible service-line costing in roughly six weeks without new software. Step one: define the capacity cost rates. For each role band - senior, mid, production - compute fully loaded monthly cost (salary, benefits, allocated overhead) and divide by practical capacity, typically 80-85% of nominal hours after meetings and administration, following Kaplan and Anderson's practical-capacity guidance (Kaplan and Anderson, 2004). The output is a cost per hour by role that already accounts for realistic availability. Step two: map each service line to a time equation - the activity steps, the role that performs each, expected hours, and the two or three complexity modifiers that actually move effort. Build these from time-tracking history where it exists and structured team estimates where it does not. Step three: run the trailing twelve months through the model - hours logged per line and client, multiplied by role rates - and produce the two-axis margin view. Step four: separate unused capacity as its own cost line rather than loading it into services. Step five: act on the three standard findings. Repricing candidates: lines whose true cost exceeds assumed cost. Redesign candidates: lines whose margin problem is delivery mix - too much senior time on steps a trained mid-level could own. Retirement candidates: lines that are structurally unprofitable at any price the market will pay. Step six: institutionalize the quarterly snapshot and tie the rate card to it.

Section 7

Evidence-based action plan

Week one: list every distinct service line and pull twelve months of revenue per line; if revenue cannot be split by line, that itself is finding number one - fix invoice line-item discipline immediately. Weeks two and three: build role-band capacity cost rates at practical capacity, per the TDABC method (Kaplan and Anderson, 2004), and draft time equations for your top three lines by revenue. Week four: run the first margin snapshot and rank lines and top-ten clients by true margin. Expect the documented pattern - a skew in which a minority of work funds the rest (Kaplan and Anderson, 2007). Month two: take the three standard actions. Reprice the worst mispriced line on new proposals; redesign one delivery process to shift hours from senior to mid-level execution; open a respectful repricing or restructuring conversation with the single most unprofitable client. Month three: publish the internal rate card with explicit complexity surcharges, and schedule the quarterly snapshot as a standing finance ritual. Measure success on blended numbers twelve months out: gross margin by line converging upward, senior-hours share on low-margin work falling, and firm EBITDA pulling away from the 9.8% benchmark average (SPI Research, 2025, vendor benchmark). The strategic prize is bigger than repair: a firm that knows its true margins can scale the right line deliberately - which is the actual meaning of profitable growth. For adjacent evidence in this pillar, see [The Growth-Investment Decision: Evidence on When to Reinvest vs Distribute](/blog/growth-reinvest-vs-distribute-decision) and [Profitable Growth Has Replaced Growth at All Costs: The Evidence and the Operator Playbook](/blog/growth-profitable-growth-replaces-growth-at-all-costs).

FAQ

Direct answers for operators.

What is time-driven activity-based costing in plain terms?

It is a costing method needing only two estimates: what an hour of each role truly costs - fully loaded cost divided by practical available hours - and how many hours each activity in a service consumes, adjusted for complexity drivers (Kaplan and Anderson, 2004). Multiplying them gives every service line and client a defensible cost without the employee surveys that sank traditional ABC.

How often should a small firm recalculate service-line margins?

Quarterly is the practical sweet spot. A live costing system is overkill below roughly 100 staff; instead, run the model from time-tracking exports once per quarter - typically one analyst-day of work - and pair it with a kill-or-fix review of the weakest line and clients. Update capacity cost rates annually or whenever compensation and headcount shift materially.

What do firms usually discover the first time they measure margin by service line?

Three patterns recur: standardized, mid-level-delivered services earn more than the books implied; bespoke, senior-heavy services earn less, sometimes negative; and client profitability is sharply skewed, with a few demanding accounts consuming unpriced senior time. The TDABC literature documents the same skew in implementations, where honest costing revealed unprofitable customers behind healthy averages (Kaplan and Anderson, 2007).

Do we need special software to do activity-based costing?

No. A sub-50-person firm needs disciplined time tracking against service lines, a spreadsheet holding role-based capacity cost rates, and time equations for each offer. The method was explicitly designed to be lighter than survey-based ABC (Kaplan and Anderson, 2004). Software earns its place later, when client count and service variety make quarterly manual snapshots genuinely burdensome.

Joshua Agonya Pi'Rwot

Written by

Joshua Agonya Pi'Rwot

Founder, Business Growth Accelerator · Country Director, AVODA Group Uganda · EMBA

Joshua helps service-business operators turn scattered marketing into a clear path from first attention to booked call. He is Founder of Business Growth Accelerator and Country Director of AVODA Group Uganda.