Most support delays on social aren't a staffing problem. They come from people typing the same answer to the same five questions all day. Trigger words let the inbox spot those questions and either handle them or flag them, so your team's time goes to the messages that need judgment.
TLDR: Start with six trigger-word rules: order status, shipping, refunds, pricing, opening hours and urgent problems. Automate the answer for the first five, send the sixth straight to a human. Test with real phrases, then review what was caught after one week.
Find your five repeat questions first
Open your last 100 messages and comments and tally up the questions. Almost every brand finds the same pattern: a handful of questions make up most of the volume. Those are your trigger-word candidates. Questions that need context, like a complaint about a specific order, are not.
Write each one as a plain list of the words people really use. Customers say "where is my order", "tracking", "hasn't arrived", not your internal term.
Six starter rules
| Rule | Example trigger words | What happens |
|---|---|---|
| Order status | "where is my order", "tracking", "hasn't arrived" | Reply with the tracking page link |
| Shipping | "delivery time", "ship to", "international" | Reply with your shipping policy summary |
| Refunds | "refund", "return", "money back" | Reply with the steps, flag for a human to follow up |
| Pricing | "price", "how much", "cost" | Reply with the pricing page link |
| Opening hours | "open", "hours", "closed" | Reply with hours and the booking link |
| Urgent | "broken", "damaged", "urgent", "wrong item" | No auto reply, send to a person now |
Notice the last row. The most important rule is the one that doesn't answer. It catches the messages where a fast, human reply saves the customer.
Write replies that sound like a person
An automatic reply should be short, name the next step and say what happens if it didn't help. For example: "Thanks for asking. Your tracking link is here. If it doesn't show your parcel, reply here and we'll look into it today."
Don't pretend to be a person, and don't say "we value your feedback". Give the answer, then the way out.
Set it up in Mydrop
Mydrop has trigger words for comment and message automation, and a social inbox that gathers conversations from your connected accounts in one place. Build one rule per row above, give each a short reply, and keep the urgent rule as a flag. If you run several brands, create the rule set per brand, since each one has its own questions and policies. For choosing between tools, see the best comment trigger-word tools.
Test with typos, then review after a week
Before turning rules on, send yourself test messages: the exact words, a typo, a plural and a sentence that contains the word by accident. A rule on "open" can fire on "I just opened the box". Tighten the words that over-trigger. If rules fire less than you expect, this guide to missed messages walks through the usual causes.
After seven days, read every auto-handled conversation. Look for two things: customers who replied angrily after an auto reply, and questions that still arrived untouched. Edit the rules for both.
What to leave to a human
Automation is good at questions with one right answer. It's bad at anything with feeling or risk. Keep these with a person:
- Complaints about a person or a past experience. A template makes them angrier.
- Anything legal, medical or financial. One wrong auto reply is costly.
- Questions from a big customer or a partner. Those deserve a name, not a rule.
- Messages where the customer is confused after an auto reply. If they write back twice, a person takes over.
A good target isn't "automate as much as possible". It's that every customer gets either a correct instant answer or a fast human one.
Add your first three rules this afternoon
Pick the three questions with the highest volume from your last 100 messages, write down the words people used and a two-line reply for each. Add the urgent rule on top. Then start free with Mydrop and set them up in the inbox. Review on Friday.
Google review
Trustpilot review
Product Hunt comment
