Department workflows: PPC vs SEO vs social vs CRO vs web design (the agency operating model)

TL;DR

The five departments inside most digital marketing agencies — PPC, SEO, social media, CRO, and web design — have structurally different workflows. PPC runs in weekly cycles tied to ad-platform optimization; SEO runs in 8-12 week project cycles tied to content + technical pushes; social runs in 24-48 hour content cycles; CRO runs in experiment cycles of 2-6 weeks; web design runs in 4-12 week project arcs with milestone approvals. A generic PM tool that uses one set of statuses, one report template, and one approval flow for all five gets the math wrong on every department. This post: the actual workflow per department, the inputs and outputs, the tooling that's department-specific, and the cross-department coordination patterns that break first as you scale.

---

You can run a 5-person agency on Asana with one workflow template. By 15 people you can't. By 25 people you definitely can't.

The reason is that "agency work" isn't one thing — it's at least five distinct workflow patterns inside the same team. PPC and web design have basically nothing in common operationally. The senior PPC specialist's day-to-day is checking conversion data, tweaking bids, rebuilding audiences, looking at search queries, writing new ad copy. The senior web designer's day-to-day is mockups, client feedback rounds, dev handoff specs, asset libraries. Their tools are different, their cadence is different, their definition of "done" is different, their reporting is different. Forcing them through the same statuses + templates makes both teams less effective.

Below: the real workflow each department runs, what it needs from tooling, and where most agencies get the cross-department coordination wrong.

PPC: the weekly optimization loop

Cadence: Weekly, sometimes twice-weekly for high-spend accounts. Most senior PPC specialists batch their account work into one or two long sessions per week.

The actual workflow per client per week:

1. Data pull. Check yesterday's spend, conversion, CPL across all active accounts. Flag anomalies (a campaign that spent 3× its daily target, a conversion rate drop on a specific ad group).
2. Search-query review. Look at last 7 days of search queries; flag negatives, promote high-intent terms to exact-match.
3. Creative + copy iteration. Identify worst-performing ads + replace. New creative briefs to designers if needed.
4. Bid + budget rebalancing. Move budget from low-CPC-but-low-volume campaigns to high-volume-good-CPC ones. Adjust bidding strategy when a campaign matures.
5. Audience hygiene. Are remarketing audiences fresh? Are first-party customer lists still uploaded? Are exclusion lists current?
6. Landing-page check. Quick sanity check that the landing pages are loading + the conversion tracking is firing. The tracking infrastructure map makes this a 2-minute audit.
7. Weekly report. Performance vs target, what changed, what's working, what's next.

What this needs from tooling:

• Per-account dashboards that pull live from each ad platform. Looker Studio is the default; agency tools like AgencyAnalytics work too.
• Tasks scoped per account, not per project. PPC work is recurring, not one-shot. The PM tool needs to support recurring tasks per account.
• Version-tracked changes. When something works (or breaks), the PPC specialist needs to remember what they changed last week. Most teams use a Slack channel or a Notion page; both decay.
• Conversion-tracking visibility. Without a clear view of which conversion actions are wired correctly per account, you spend 30% of your time figuring out whether the data is real (see the hidden cost of broken tracking).

The gotcha most agencies miss: PPC quality scales with rep count, not effort. A senior who sees 20 accounts every week catches patterns a junior on 5 accounts won't. So the org structure should concentrate PPC across senior specialists with junior support, not spread it across many generalists. Most agencies stretch PPC across too many people and get medium results everywhere.

SEO: the 8-12 week project cycle

Cadence: Quarterly, sometimes longer. SEO projects don't fit weekly status updates because most SEO work doesn't show measurable impact for 90+ days.

The actual workflow per client per quarter:

1. Quarterly audit (week 1). Technical SEO: site speed, Core Web Vitals, indexability, schema, internal linking. Content gap analysis vs competitors. Backlink profile review. Output: a prioritized punchlist of 15-30 items.
2. Content + technical push (weeks 2-10). Execute the punchlist. Publish new pages, fix existing pages, ship technical improvements. Most clients expect 4-12 published pieces per quarter on a real engagement.
3. Verification + measurement (weeks 11-12). Indexation health (was the new content indexed?). Ranking movement (don't expect much in 12 weeks; some quick wins; most pages take 4-9 months to rank). Traffic + conversion attribution.
4. Quarterly review with the client. Show what shipped, what we expect to compound, what's queued for next quarter.

What this needs from tooling:

• Long-running task lists per client. Asana milestones work; Linear cycles work. Generic Kanban boards struggle because the work doesn't move through statuses linearly.
• Documented technical state per client. Site map, indexed page count, current schema markup, current internal-linking patterns. Without this baseline, every audit reinvents the wheel.
• Content calendars. What's drafted, what's editorial-reviewed, what's published, what's awaiting promotion. Most agencies run this in a separate spreadsheet that decays.
• Search Console + GA4 access per client. Documented in your tracking map (see the 21 things every agency must track per client).

The gotcha most agencies miss: SEO clients churn because the rate-of-change feels slow. The fix is not faster SEO; the fix is better quarterly storytelling. Most agencies under-document the work shipped + the leading indicators (impressions up before clicks; positions improving across many keywords before traffic shows it). When the client only sees "traffic flat" three months in, they cut. When they see "we shipped 8 pieces of content + technical improvements + here's the indexation data + the impression curve", they renew.

Social media: the 24-48 hour content cycle

Cadence: Daily, sometimes multiple times per day. Social moves faster than any other agency department.

The actual workflow per client per week:

1. Content calendar planning (usually weekly, sometimes biweekly). Themes, posts, hashtags, CTAs.
2. Asset production. Designers ship visuals; copywriter (or specialist) ships copy. Often same-day turnaround.
3. Approval cycle. Client reviews + approves. The cycle time on this kills more social engagements than any other failure mode — clients expect 24-hour turn but commonly take 3-5 days.
4. Scheduling. Post via Later / Sprout Social / Hootsuite / native platform schedulers. Most agencies use a mix.
5. Monitoring + community management. Reply to comments, flag inbound DMs that need attention, escalate negative sentiment.
6. Weekly performance review. Reach, engagement, follower growth, conversion (where measurable).

What this needs from tooling:

• A content calendar that shows the next 4 weeks at a glance, color-coded by client. Most agencies use a Google Sheet or Notion. The calendar UI matters because the planning conversation happens against this view weekly.
• Approval workflow tied to content state. Most agencies use Slack ("can you approve this?"); doesn't scale past 5-7 active social clients. Productized approval flows (clients approve via a portal, not via thread digging) save 20% of senior-team time.
• A scheduling tool per platform. Sprout, Later, and Hootsuite all do this; pick one + standardize.
• Moderation queue. Flag inbound that needs response. Most agencies miss this because it's not in the PM tool — it's in whichever scheduler has it. Hostile comments that go un-addressed for 12 hours are a service-quality issue.

The gotcha most agencies miss: social margins are thin because production is per-post, not per-month. A 30-post-per-month engagement that prices like a 10-post-per-month one bleeds margin invisibly. Repricing should reflect actual production volume, not "we always charged that for social."

CRO: the experiment cycle

Cadence: 2-6 week experiment cycles per concurrent test. Most agencies run 1-3 concurrent tests per client.

The actual workflow per client per cycle:

1. Hypothesis (week 0). What problem are we solving + what's the test? E.g. "the form abandonment on /signup is 78% — we hypothesize that reducing the field count from 7 to 4 will lift conversion 15-25%".
2. Test design (week 1). Variant treatments, sample size calculation, success metric, runtime estimate. Statistical power matters — running underpowered tests is the most common CRO mistake. A test that needs 12,000 users to reach significance and only gets 3,000 ran for nothing.
3. Implementation (weeks 1-2). Devs ship the variant. QA. Make sure the conversion tracking still fires correctly on both variants (this is where tracking infrastructure discipline pays back).
4. Run (weeks 2-5). Test runs. Watch the data, but don't peek-and-call early — most "wins" called in week 2 are noise.
5. Analyze + decide (week 5-6). Did the variant win at significance? Implement winner. Document the outcome.

What this needs from tooling:

• An experiment registry per client. Every test ever run, its hypothesis, its outcome. Without this, agencies repeat experiments they already lost on.
• A/B testing platform (Optimizely, VWO, Convert, Statsig). The decision is platform-cost-vs-features; most digital agencies under 25 people run on Convert or VWO.
• Conversion tracking that survives the variants. Most CRO breakage is conversion-tracking breakage during variant deployment. The variant deploys but the conversion event doesn't fire on the variant version, so the test reads as a loss when it's actually a wiring error.
• Statistical-significance calculator. Most A/B platforms include one; agencies running offline tests should standardize on Bayesian or frequentist + stick to it.

The gotcha most agencies miss: CRO results compound across clients. A test that wins on Client A's pricing page is a 60-70% bet to win on Client B's pricing page (similar industry, similar funnel shape). Most agencies don't structure their experiment registry to surface these patterns; they treat each client as a clean slate. Building the cross-client pattern library is one of the highest-leverage things a CRO-heavy agency can do.

Web design: the milestone-based project arc

Cadence: 4-12 weeks per project, with 4-6 named milestones.

The actual workflow per client per project:

1. Discovery + brief (week 1-2). Client interviews, competitor review, scope, success criteria.
2. Sitemap + wireframes (week 2-3). Information architecture + low-fidelity layouts. Client approval at this milestone is critical — changes after this point are 5× cost.
3. Visual design (weeks 3-5). High-fidelity mockups in Figma. Usually 2-3 review rounds with the client.
4. Dev handoff + build (weeks 5-9). Designs to dev. Specs, design tokens, component library handoff. Mid-build review.
5. QA + accessibility check (week 9-10). Cross-browser, mobile, accessibility (WCAG 2.1 AA at minimum).
6. Launch + handoff (week 10-12). Pre-launch checklist, launch, post-launch monitoring, knowledge transfer to the client team if applicable.

What this needs from tooling:

• Milestone-based PM, not Kanban. A 6-milestone project doesn't fit a "to do / in progress / done" board well. Asana's timeline view, Linear's roadmap, or Notion timelines work better.
• Design tool (Figma, almost universally in 2026). With component libraries, design tokens, and a documented handoff workflow.
• Asset library per client. Brand guidelines, logos, fonts, color tokens, photography. Most agencies keep this in Google Drive or Dropbox; the Figma library can also serve as the source of truth.
• Client-approval workflow with timestamps. When a client approves a milestone, that's the design baseline. Later changes are scope changes, not iterations. Without explicit approval logging, the project drifts.

The gotcha most agencies miss: web design projects almost always run over because clients underestimate how much their input is needed. The fix is to require named decision-makers up front + cap revision rounds explicitly in the SOW (e.g. "two rounds of revisions per milestone, additional rounds priced at $X/hour"). Agencies that do this halve their web-design margin leakage.

Cross-department coordination: where most agencies break

The departmental workflows above are coherent within each team. The chaos is at the boundaries.

Boundary 1: PPC + Web design. PPC needs landing-page changes; web design treats those as scope. The fix: a standing "PPC asks" lane in the web-design backlog. PPC writes a brief, web-design slots it; clear SLA (e.g. 5 business days for non-urgent landing-page tweaks).

Boundary 2: SEO + Web design. SEO needs technical fixes (schema, page speed, internal links) that touch the dev side. The fix: SEO punchlist gets reviewed quarterly with the dev/design team. Items get sized + slotted into upcoming sprints. SEO doesn't get a separate pipeline; it shares the dev capacity.

Boundary 3: Social + Web design. Social production volume eats designer capacity invisibly — 30 posts/month at 30 minutes each is 15 designer-hours, half a day per week. Either price it in (most agencies don't), give social its own production designer, or push social to in-house client teams.

Boundary 4: CRO + everyone. CRO lives at the intersection of design (variant treatments), dev (variant deployment), tracking (measurement), and PPC (test traffic source). Without a senior-CRO orchestrator, the testing program stalls. Smaller agencies don't need a dedicated CRO lead, but they do need a designated CRO orchestrator on the senior team.

Boundary 5: All departments + tracking infrastructure. Every department touches tracking, none owns it. Result: pixels break and nobody notices for weeks. The fix is exactly what we've written about elsewhere: a tracking infrastructure map per client with a named owner + a quarterly verification cadence. Without that, the tracking is everyone's-job-and-no-one's-job.

What good cross-department workflow looks like

The agencies that scale past 25 people while protecting margin (see scale a marketing agency without losing margin for the operating math) almost always have these in place:

1. Per-department workflow templates in their PM tool — one for PPC weekly cycles, one for SEO quarterly punchlists, one for social content calendars, one for CRO experiment registries, one for web-design milestone projects.
2. Department-shared rituals — daily standup is per-department; weekly all-hands is whole-agency; quarterly OKR review is whole-agency.
3. Explicit SLA boundaries between departments (especially Web → PPC and Design → Social).
4. A unified per-client view that shows status across departments, so account managers don't have to ask each department individually. This is the part most agencies don't have. We built Phloz around it: every client has tasks across departments, the tracking infrastructure map cuts across departments too, and the per-client dashboard shows the full picture without department-specific context-switching.

The Phloz angle

Phloz models work as Task records with a department field — PPC, SEO, social, CRO, web-design, or other. Each department gets its own filter, its own report, its own templates. The same task table holds everything; the department field is what lets each team see only their work without losing the cross-department picture for account managers + leadership.

Same principle for the tracking infrastructure map: nodes are typed (GA4 property, GTM container, ad pixel, audience, conversion action, etc.), and each node-type "belongs" to the department that primarily owns it — but every department can see the full graph for any client they touch.

This is the architectural answer to "departments need their own workflows but the agency needs the unified view": typed objects with department metadata, plus filters per role. Generic PM tools can be retrofitted to look like this; agency-shaped tools start there.

Closing

The mistake most agencies make at the 15-person band is generalizing — one workflow, one report template, one set of statuses. The mistake at the 25-person band is over-specializing — five completely separate ops, no cross-team visibility, departmental fiefdoms.

The right answer is in between: per-department workflows with shared infrastructure, explicit SLAs at department boundaries, and a unified per-client view above it all.

If your agency is stuck on either end of that spectrum, the lever to pull is whichever one feels less natural. Most generalist agencies need to stop forcing departments into one shape; most fiefdom agencies need to make the cross-department view as good as the per-department views.

Try the free tier (2 active clients, 2 seats) to see what cross-department per-client workflow looks like, or see pricing for the full breakdown.