Field guide

The agency PPC field guide

A working agency owner’s guide to running paid-media work that survives contact with the calendar. Operational, not theoretical.

This is the guide I would have wanted before I started the agency. Most agency-operations content is either war stories without process, or process without lived experience. This sits in the middle: the processes that survived the calendar, with enough operational detail that you can copy them.

Section 1: Picking which clients to take

The single most consequential agency decision is which clients you accept. The work is downstream of the client selection — bad clients can’t be saved by good operations.

The agency’s current filter: clients spending $30K–$400K/month, with a founder or marketing leader who can articulate their unit economics, in either ecom DTC or B2B SaaS. We turn down accounts that are too small (the time-to-fee ratio is wrong), too large for our delivery capacity, in verticals we don’t understand, or with leadership that can’t tell us what a profitable customer looks like.

The disqualifying signals

• “We just need someone to manage the ad account.” This client wants execution, not strategy. The agency model doesn’t fit; they need an in-house operator or a managed-service product.
• “Our previous agency stole from us.” Sometimes true. More often, the client doesn’t understand what they were paying for, which means they’ll think the next agency is stealing too.
• “We need to be at break-even ROAS by month two.” Either the unit economics support this and previous agencies failed (rare), or the client is going to fire us in month two regardless of progress.
• “We don’t know our gross margin precisely.” Hard pass. Without margin, we can’t derive break-even ROAS, can’t set tROAS targets, can’t tell if we’re winning.

Section 2: Onboarding a new account in week one

The first week determines the relationship for the next year. Skipped or rushed in this phase, everything downstream is harder.

Day 1–2: The taxonomy audit

Pull every existing conversion event, audit which ones fire, which ones are duplicated, which ones are misclassified. Most accounts arriving from another agency have a conversion-event taxonomy that’s been accreted over years. The cleanup typically removes 30–60% of the events and consolidates the remainder.

Day 3–4: The margin conversation

Sit with the founder and finance lead. Derive break-even ROAS, profitable ROAS at 1.5× break-even, and the marginal-spend ROAS curve. This is the conversation that produces the operational targets for everything that follows. Without it, the team is optimizing toward an arbitrary number.

Day 5: The stack audit

What’s currently in place? Native ad platforms, third-party tools, reporting, attribution. Document everything, identify duplications, identify gaps. Don’t change anything yet — this week is mapping, not modifying.

Section 3: The weekly operating rhythm

Across a portfolio of twelve client accounts, consistency requires explicit rituals. Three meetings per week per account:

Section 4: How to evaluate a new tool for the stack

The agency runs every new tool on three client accounts simultaneously, for a fixed 90-day window. The methodology is on the methodology page. A tool passes if it produces statistically meaningful revenue-weighted ROAS lift on at least two of the three accounts.

The non-obvious part of the test protocol: the three accounts should be heterogeneous. Three apparel ecom accounts isn’t a test — it’s three replicates of the same test. Pick accounts in different verticals, different spend tiers, different team configurations. If the tool works across heterogeneous conditions, it’s ready to standardize. If it works only in specific contexts, that’s an actionable insight — standardize it for those contexts.

Section 5: The client conversations that matter most

Three recurring conversations that determine whether a client engagement renews:

Conversation 1: Why the reported number disagrees with the P&L

Every ecom client will see Google Ads report a 5× ROAS while their CFO shows the account as unprofitable. The conversation: walk through what reported ROAS includes (gross revenue divided by spend) and what it excludes (returns, COGS, processing fees). Use the True ROAS calculator or a similar tool to make the gap visible. Re-anchor the team’s targets to the true ROAS rather than the reported one.

Conversation 2: Why we’re not intervening

When Smart Bidding or a third-party model is in its exploration phase, performance looks bad. Clients want intervention; the right answer is usually patience. The conversation: explain the exploration phase, set a pre-committed measurement window, and resist intervention until the window completes.

Conversation 3: Why we’re killing a campaign that “works”

Sometimes the campaign reports positive ROAS but is structurally cannibalizing organic or brand-defense spend. The conversation: walk through the marginal-spend math, show that the campaign’s “wins” aren’t incremental, and propose redirecting the budget. This is the hardest conversation in the rotation because it requires the client to accept that a positive-looking number is misleading.

Section 6: What to do when the calendar gets tight

Every quarter there’s a week where the calendar collapses — two clients have urgent needs, a strategist is sick, an unexpected platform change requires emergency intervention. The agency’s contingency rules:

• Defer optimization work, not measurement work. Skipping a weekly check-in misses signal you’ll need to interpret a later week’s data. Skipping optimization is recoverable.
• The COO routes emergencies. Strategists don’t take their own escalation decisions — the COO sees the portfolio and decides which client’s urgent need outranks which other client’s urgent need.
• Tell the affected clients explicitly. “We’re defering your weekly check-in by 48 hours because of a portfolio-level issue” is uncomfortable but survives. Going dark doesn’t.

Section 7: What I won’t do

The list of things the agency has declined over five years is longer than the list of things we’ve done. Worth naming because it’s a structural part of the work:

• Sponsored placements in editorial content (this site or its sister sites).
• Performance-based pricing where the agency captures a share of revenue lift. Aligned-incentives in theory; tangled-attribution and arbitration in practice.
• Engagements with companies whose unit economics don’t support paid acquisition. No bidding optimization fixes negative LTV:CAC.
• Multi-channel agency expansion. The agency does paid search and paid social. Email, SEO, content marketing get referred out.

The discipline of saying no is the same discipline that makes the work survive years. Most agency failure modes start with saying yes when no was the right answer.