Per-client profitability: the math most agencies avoid

TL;DR

Most agencies know their overall margin and almost nothing about where it comes from. The fix is a quarterly, two-number exercise per client: loaded delivery cost (hours actually spent × loaded hourly rate of the people spending them) against retainer revenue, expressed as a coverage ratio. Healthy retainers run at a ratio of roughly 2.0 or better (revenue at least double delivery cost); between 1.3 and 2.0 you have a "fix" conversation; below 1.3 the client is consuming other clients' margin and you raise, restructure, or exit. You don't need timesheets to get close — a per-client activity signal (tasks completed, messages handled, meetings held) calibrated against a one-week sampling sprint gets you to decision-grade numbers with a tenth of the overhead. This post walks the full method, the three classic margin leaks it exposes, and how to run the fire/fix/raise meeting without drama.

---

Agency owners can usually quote their blended margin to the decimal. Ask which three clients produce most of that margin — or which two quietly consume it — and the room goes quiet. That silence is expensive: the most common agency growth mistake is scaling the number of clients while the mix of clients gets worse.

The math below is deliberately rough. Decision-grade beats audit-grade, because the goal is a quarterly decision, not a financial statement.

Step 1 — Loaded hourly cost (one number per person, once a year)

For each delivery person: annual salary plus benefits plus their share of overhead (rent, software, admin salaries), divided by realistic billable-capacity hours — not 2,080; closer to 1,400–1,600 once meetings, PTO, and internal work are honest.

A $75k account manager with 30% benefits/overhead loading at 1,500 deliverable hours costs about $65/hour. A $120k senior PPC lead lands near $105/hour. Write these down once a year; precision past the nearest $5 changes no decisions.

Step 2 — Hours per client, without the timesheet war

Timesheets are accurate and universally hated; most agencies' data quality collapses within a quarter of mandating them. The alternative that holds up:

1. Run one honest sampling week per quarter. Everyone tags their week's hours by client. One week, not fifty-two.
2. Calibrate an activity signal. Count the trailing-quarter per-client activity you already have in your system of record — tasks completed, client messages handled, meetings on the calendar. If your work and client communication live in one workspace, this is a report, not a project; it's exactly the cross-client view we built cross-client reporting around, and the same per-client activity signal Phloz already tracks for its own per-active-client pricing.
3. Scale the sample by the signal. If Client A generated 18% of the quarter's activity and the sampling week says your team delivered 410 hours/week total, Client A consumes ≈ 74 hours/week × the people-mix cost. Rough? Yes. Within 15% of timesheet truth in every agency we've seen test both? Also yes.

Step 3 — The coverage ratio

Per client, per quarter:

Coverage = retainer revenue ÷ (estimated hours × loaded cost of the actual people-mix)

Then sort the client list by it. The distribution is always the same shape and always surprising in its specifics:

• 2.0 and above — protect. These clients fund the agency. Their account teams should be your most stable; their renewals should never be a surprise to you.
• 1.3–2.0 — fix. Usually one of the three leaks below, and usually fixable within a quarter.
• Below 1.3 — decide. At 1.0 you're a charity with invoicing. Raise, restructure, or exit — this quarter, not "after things calm down."

The three leaks the ratio exposes

Leak 1 — Scope creep with a smile. The client who "just needs one quick thing" weekly. Individually reasonable; in aggregate, an unbilled workstream. The tell: their activity share grows quarter over quarter while the retainer doesn't. The fix is rarely confrontation — it's a renewal conversation backed by the activity record: "here's everything we shipped this quarter; the retainer that covers it is X." (Walking into that conversation with receipts is the whole game — which is an argument for client work living in a system that remembers, not in DMs.)

Leak 2 — The seniority mismatch. A founder or senior lead still personally servicing a legacy client at junior-retainer pricing. The hours look fine; the people-mix cost is the leak. Fix by re-staffing the account deliberately, with a transition the client hears about as an upgrade ("dedicated day-to-day lead, senior oversight continues").

Leak 3 — The firefighting client. Broken foundations — usually tracking — turn every campaign question into an investigation. We wrote up the hidden cost of broken tracking separately: these clients burn senior debugging hours that never feel like "delivery" and never make it into scope. Fix the foundation once (billed as a project), or accept the ratio is structural.

The quarterly meeting: fire / fix / raise

Run it ninety minutes, once a quarter, owner plus department leads, the sorted list on screen:

1. Bottom three clients: for each, pick exactly one — raise (renewal with receipts), restructure (scope or staffing), or exit (with a referral and grace; agencies that fire clients kindly get referrals from them later — genuinely).
2. Top three clients: name the risk to each (single point of contact? key AM flight risk? contract end date?) and one action to de-risk.
3. One systemic fix: whichever leak pattern showed up most. The compounding move is making the next quarter's math automatic — which is the case for keeping tasks, communication, and client records in one place, so the activity signal is a query instead of an archaeology project. That's the operating-leverage argument we made in scaling without losing margin, and it's the difference between this being a 90-minute meeting and a lost weekend.

What good looks like after a year

Agencies that run this loop for four quarters report the same arc: the first quarter is uncomfortable (two clients exit or get raised), the second is corrective (staffing mixes change), and by the fourth the new-business filter has changed — they price prospective clients against the coverage ratio before signing, which is where the real money was hiding all along. Margin problems are mostly mix problems, and mix problems are mostly visibility problems. Fix the visibility and the rest follows.