How we actually work

Q1SOP structure

Every recurring task we run for a client lives in a documented SOP. New SOPs get created the first time a task happens for that client, refined the second time, and become reference-grade by the third.

The SOP library is structured as a Notion workspace per client with a fixed folder pattern:

Each SOP follows the same internal structure: trigger (when this happens), steps (numbered list with screenshots), escalation (who to ping if blocked), QA check (what "done" looks like), last reviewed date.

SOPs are versioned. When a process changes, the previous version stays accessible with a date stamp so we can track why a process evolved and roll back if a change causes problems.

Q2Onboarding a new Shopify brand

Onboarding is 14 calendar days from discovery call to full handover. Compressed timeline because most clients are already losing money to the operational gap we're stepping into. The schedule:

1. Day 0: Discovery call (30 min) Map current operational layer, identify the highest-value gaps, agree scope. Sign engagement letter same day or next.
2. Days 1-3: Audit week Full read of the Shopify admin, app stack, support inbox, supplier directory, current SOPs (if any). Output: a documented assessment with the top 10 operational risks and recommended fixes.
3. Days 4-7: SOP write We write the initial SOP set covering every recurring task identified in the audit. Founder reviews and edits, we revise. By end of day 7 the brand has a working operations manual that did not exist before.
4. Days 8-12: Shadow week The assigned operator handles tasks under review from the lead. Founder sees daily progress in Slack. The lead spot-checks every output against the SOP and flags any drift.
5. Day 13: Handover review Lead, operator, and founder do a 45-minute call walking through what is now running, what stayed manual, what is queued for week 3+. Any final SOP adjustments documented.
6. Day 14: Full ownership Operator owns the account day-to-day. Lead steps back to weekly review. Daily 5-line briefing email starts going out each morning.

After day 14 the engagement is in steady-state mode. We do a 30-day review, 60-day review, and a structured quarterly review with the founder.

Q3Automations we build for clients

Our default automation stack is n8n (self-hosted, no per-task fees), with Make as a fallback for simple workflows and direct Shopify webhooks where speed matters. We avoid Zapier on retainer clients because per-task pricing scales badly at our usage.

The most common automations we build in the first 90 days of a new engagement:

Every automation we build comes with monitoring: if it fails three times in a row we get a Slack alert and the failure is logged. The operator owns triage; the lead owns the fix.

Q4How we QA VA work

QA runs at three layers, each with a different cadence:

Layer 1: First-hour check-back (continuous)

New VAs and new task types: the lead reviews every output in the first hour of work and again at the end of day one. Cuts onboarding errors by roughly half compared to letting the operator run unsupervised in the first 24 hours.

Layer 2: Daily spot checks (random)

The lead pulls 3-5 outputs at random each day across the active engagements and checks them against the SOP. Any drift gets logged in the operator's review document and addressed within 24 hours.

Layer 3: Weekly QA report (structured)

Every Friday the lead writes a 1-page QA summary per active engagement: what went right, what drifted, what got fixed, what is still open. The founder receives this with the weekly performance report.

QA outputs that flag a systemic issue (not a one-off) trigger an SOP update within the same week. The change is logged with a version stamp so we can trace why a process evolved.

The escalation matrix

For any operator-level decision:

1. If the SOP covers it: operator executes.
2. If the SOP doesn't cover it but the precedent is clear: operator executes and logs for SOP addition.
3. If the impact is over $250 or affects a customer relationship: operator pauses and pings the lead.
4. If the lead is unsure or the impact is over $1,500: lead pings the founder before action.

Q5KPIs we track weekly

Every retainer engagement reports against a standard KPI set, customized per category. The default Shopify operations report includes:

Operational KPIs

• Orders processed (count, by day-of-week)
• Exception orders (count, resolution time, root cause categories)
• Stockout incidents (count, days-of-stock-remaining at risk SKUs)
• Supplier SLA compliance (on-time delivery rate per supplier)
• App stack failures (count, categorized by app)

Support KPIs (when in scope)

• First response time (median + 90th percentile)
• Resolution time (median + 90th percentile)
• CSAT (1-5 rating, with comment review)
• Ticket volume (by category)
• Reopen rate (tickets that came back after resolution)

Marketing KPIs (when in scope)

• Email and SMS revenue share (by flow vs campaign)
• Klaviyo flow performance (welcome, abandon, post-purchase, win-back)
• Paid channel ROAS (per platform, with 14-day attribution window)
• Conversion rate (storewide + top 5 products)

Reports go out every Monday by 10am client local time, covering the previous week's data. The founder reads it; the lead is on standby for the day to discuss any anomaly.

Q6AI-assisted vs human-only tasks

The split is deliberate. AI is the worker for repeatable, low-ambiguity, easily reviewable work. Humans own anything that requires judgment, brand-critical creative, or irreversible decisions.

We do not let AI auto-publish to customer-facing surfaces. Every AI output passes through a human gate before it reaches a customer, a supplier, or anyone outside the operations layer. The gate exists because AI mistakes compound when they go live, and the cost of one bad auto-published reply is higher than the time savings across a hundred good ones.

Q7What happens if your VA disappears

A single dedicated VA is one set of hands. When that person gets sick, takes vacation, or quits, the engagement should not pause. Our continuity model:

Backup operator

Every active engagement has a primary operator and a designated backup. The backup is briefed on the account during onboarding and stays current via weekly SOP digests. They can step in within one hour if the primary becomes unavailable.

SOP-driven coverage

Because every recurring task is documented, the backup operator is not learning the account from scratch. They are executing against a written playbook. The first day of backup coverage might be 70% as fast as the primary, but the work gets done correctly because the SOP is the source of truth.

Lead operator escalation

If both primary and backup are unavailable in the same window (rare but possible), the lead operator takes over personally until coverage is restored. Lead capacity is reserved specifically for these situations.

Founder-visible continuity

The founder sees the change immediately in the daily briefing: "Today's briefing is from [backup operator] while [primary] is on planned leave through Friday." No surprise gaps, no missed touches, no need for the founder to wonder why Slack went quiet.

Q8Example workflows, end to end

Three real workflow patterns we run for clients, walked through step by step. Names of specific clients omitted to protect engagement details, but the operational shape is exact.

Workflow A: Daily order exception handling

1. 06:00 client local time, n8n workflow pulls last 24h of orders from Shopify Admin API.
2. Each order is scored on 4 dimensions: fraud signal (Shopify's own + Signifyd if installed), shipping address validation, gift-order pattern (note text contains "gift," "wrap," etc.), and B2B classification (email domain matches commercial list).
3. Flagged orders get tagged in Shopify with the appropriate label (review-fraud, review-address, review-gift, review-b2b).
4. Operator opens the exception queue at 08:00, works through the flags. Most resolve in under 5 minutes each.
5. Anything operator can't resolve gets escalated to the lead by 10:00.
6. Daily briefing email at 08:30 includes count of exceptions handled and any escalations pending.

Workflow B: New product launch (SKU rollout)

1. Founder shares the supplier sheet (CSV or Notion page) with 10-30 new products.
2. Operator opens an SOP-defined launch checklist for each product.
3. n8n workflow runs each row through Claude to generate a draft product description, bullet list, meta title, meta description, and primary collection assignment.
4. Operator reviews each draft, edits for brand voice, approves.
5. Operator publishes to Shopify (draft status), uploads images, sets variants and pricing.
6. Lead does a final review against the launch checklist before products go live.
7. Day-of launch: operator monitors for ad-account approval, payment processor flags, and customer questions in the first 4 hours.

Time per product without the AI draft: 35-45 minutes. With: 8-12 minutes. The AI gain compounds across catalogs of 50+ SKUs.

Workflow C: Weekly support performance review

1. Sunday night: n8n pulls last 7 days of Gorgias data into a Notion database.
2. Claude reads all closed tickets, produces a 1-page synthesis: top 3 issue categories, any pattern that suggests a product or policy problem, any agent reply that drifted from brand voice.
3. Monday 09:00: lead reviews the synthesis, drafts the founder's weekly support summary, includes specific recommendations.
4. Monday 10:00: founder receives the summary alongside the broader weekly operations report.
5. Any action items (policy change, product page fix, SOP update) are logged in the engagement's open-items list with an owner and target date.