Giveaway Surge and a Safety Message in the Inbox

Start with intent and risk, not channel. Create four queues:

1. Safety/adverse event: keywords such as rash, reaction, burn, swelling, allergic, hospital, ingredient. Route to a named agent immediately. Do not auto-close. Capture product, batch, date, symptoms, photos and consent, then escalate to the legal or safety owner within the 24-hour requirement.
2. Order and product care: order status, delivery, returns, how-to-use. Saved replies plus order-lookup steps; human sends the final response when a case is open.
3. Giveaway entry: 'Did I win?', rules, entry confirmation. Automate an acknowledgment with the rules link, expected winner date and a tag. This is high volume and low risk, so deflection is appropriate.
4. General praise/complaint: human triage with sentiment tags.

Unify Instagram, Facebook and X in one inbox view so a safety message cannot sit unseen in a separate channel. Use keyword rules and channel routing, not one giant queue. Staffing: one agent owns safety and escalated care; the second handles giveaway and order volume, with a rotation every two hours to prevent safety fatigue. Automation should acknowledge and route; it should not assess a rash or deny a claim.

I would report a small dashboard by queue, not one blended average:

• First response time, median and 90th percentile, for safety, order care and giveaway.
• Safety escalations completed within 24 hours, with any breach named.
• Backlog and oldest unanswered message by queue.
• Auto-reply deflection rate for giveaway messages.
• Resolution time and reopen rate for care cases.
• Volume by intent and channel, plus sentiment on product-reaction messages.

Thresholds should trigger action, not just reporting. If any safety message waits more than one hour for first response, or an adverse event is not escalated within four hours, the care lead reassigns an agent immediately. If the giveaway backlog exceeds 500 unacknowledged messages, automate a broader acknowledgment or pause promotion spend. If the 90th-percentile care response exceeds four hours for two consecutive days, move the second agent to care and let giveaway automation absorb more volume.

I would use p90 rather than averages because a few very slow safety messages can disappear inside a good mean. The key threshold is not total volume; it is whether a safety case can wait.

I would decline to hide genuine safety complaints. Hiding them does not make the reports disappear; it increases the chance that affected customers do not get follow-up, and it creates a trust and record-keeping problem. The legal requirement to escalate adverse events also means the company should preserve and act on those messages, not suppress them.

I would explain the distinction to the creator:

• Moderate or hide: spam, abusive content, off-topic harassment, duplicate bot posts, or comments containing personal data that should move to a private channel.
• Do not hide: product reactions, safety questions, complaints about orders or claims. Respond publicly with a brief, caring acknowledgment and move the detail to DM. Log and escalate internally.

The creator's concern is understandable: a rash comment looks bad during a giveaway. The better response is visible care, not deletion. I would offer the creator a shared response protocol and a daily summary of safety cases so they see action being taken. If a comment includes medical details or a photo, the public reply should not repeat the symptoms; it should invite a private conversation and route to the safety queue.