Live Chat Operations Checklist

Open Gorgias settings and verify the Shopify, Recharge, and Loop Returns connectors show 'Connected'. A broken Shopify connector means agents can't see order history or trigger refunds inline — chats stall while agents tab between tools, and AHT balloons.

Stale promo codes in macros are a common gotcha — last quarter's BFCM30 still embedded in the 'shipping delay' apology macro will fire after expiry and frustrate the customer. Filter macros by 'discount' tag and disable any code past its end date.

Pull the current promo schedule from the marketing calendar and the OOS / low-stock report from Shopify Inventory. Agents need a real answer when a customer asks 'when will the navy XL be back' — vague answers drive cart abandonment and cancellations.

Mobile checkout breakage is the highest-impact widget failure — most DTC traffic is mobile, and a hidden chat widget at the cart step kills last-minute support. Open the storefront on a real iOS device, add a SKU to cart, and confirm the widget renders without overlapping the checkout button or shipping selector.

Gorgias and Zendesk both expose first-response time as a CSAT driver. Watch the live queue gauge — anything above 90 seconds at peak flags an under-staffed shift and should trigger a Slack ping to the CX lead.

Before responding to a 'where is my order' chat, expand the order panel to confirm SKU, fulfillment status, tracking events, and prior tickets. Answering from the customer's claim alone is the fastest way to be wrong twice and burn a CSAT.

Use the standard tag set: shipping, returns, sizing, subscription, defect, pre-sale. Untagged conversations don't show up in the weekly trending-issues review and don't drive macro or product improvements.

Send the customer the Loop self-service link rather than processing manually in Shopify — this keeps return reasons coded for the analytics dashboard. Manual refunds bypass Loop's reason tracking and skew per-SKU returns rate, which the merchandising team relies on.

Every open ticket gets a 'next action' note before end of shift — 'awaiting USPS scan', 'escalated to ops, ETA Tuesday', 'pending vendor RMA'. The next agent shouldn't have to read the full thread to know what's next.

Use the built-in Gorgias / Zendesk transcript export. Confirm the customer's email matches the order before sending — never paste a transcript to an unverified address. A misrouted transcript with PII is a privacy incident.

Verify the post-resolution CSAT rule fired on closed tickets. Survey response feeds the weekly performance review — a missed survey trigger means the entire shift is invisible on the dashboard.

Capture the shift's headline numbers from Gorgias: chat volume, average CSAT, AHT, and a brief note on anything unusual (carrier outage, viral TikTok, surprise OOS). The daily log is the rolling baseline for staffing decisions.

Export the past 7 days of CSAT score, first-contact resolution rate, and average handle time. Typical DTC benchmarks are CSAT ≥ 90%, FCR ≥ 70%, AHT 4–8 minutes — below that, the team is either undertrained or understaffed for chat volume.

Sort the past week's tickets by tag count. A spike in 'sizing' or 'defect' tags on a single SKU is a product-level signal — don't just answer the tickets, surface the trend to merchandising before more units ship.

Pull 10 random closed transcripts and score against your tone guide — empathy in the opening, brand voice in the close, no apology-overdrive. Use the Gorgias QA scorecard or a Google Sheet rubric, and share scores 1:1 with each agent.

Every new SKU or promo from the merchandising calendar needs its own macro within 48 hours of launch. Stale macros are how agents end up improvising answers to launch-day questions — and improvised answers contradict each other across the team.

Check the Gorgias / Zendesk widget changelog and apply updates in a staging theme first. Push to production after a 24-hour staging soak — widget JS injects on every page, so it's high-blast-radius for layout regressions on PDP and checkout.

Confirm the chat widget is gated by your CMP (OneTrust, Cookiebot, Termly) for EU and California visitors and only loads after consent. A widget that fingerprints visitors before consent is a GDPR violation — audit the network tab on a fresh-session EU IP test and verify GPC opt-out is honored.

If the audit found issues, file a privacy ticket with screen capture and the offending pre-consent network requests. Loop in the data privacy officer or external counsel if customer data was loaded before consent — this drives breach-notification analysis under GDPR.

Smoke-test on iOS Safari, Android Chrome, desktop Chrome, and desktop Safari across PDP, cart, checkout, and account pages. Layout regressions usually surface on PDP-with-bundle-builder or cart-with-promo-banner combinations that staging didn't cover.