Crisis or Noise? Triaging Five Mention Spikes Before They Become a War Room
Objective: Given five synthetic mention-spike specimens, correctly separate real crisis signals from ordinary volume noise using the velocity, spread, and amplification framework, without over-reacting to raw volume alone.
You're the comms lead at The Trade Desk, the programmatic advertising DSP, monitoring mentions during a platform incident. Five spikes hit your dashboard within an hour and you have to triage before deciding which one gets escalated.
Score each spike against velocity, spread, and amplification, not against volume alone, and decide which ones actually need a response.
Before you start
What you'll need
Free path (everything below is enough to finish)
Free, and sufficient for a small team to spot volume changes before investing in a dedicated platform
Free, and a simple log is enough to apply the velocity/spread/amplification framework by hand
Paid upgrades (optional, faster/deeper)
Automates the volume-over-time and reach tracking that this exercise does manually in a spreadsheet
No access? Google Alerts plus a manually logged spreadsheet covers the same triage workflow at lower volume
The process
Specimens to review
Is this a crisis? Score it against velocity, spread, and amplification and decide whether to escalate.
Complaint volume: 40 mentions -> 380 mentions in 24 hours. Started on X, picked up on a Reddit programmatic-advertising subreddit within 6 hours, then covered by an ad-tech trade publication by hour 20. An ad-tech journalist with 45K followers posted about it at hour 18.
Specimen: synthetic, realistic
Is this a crisis? Score it against velocity, spread, and amplification and decide whether to escalate.
Complaint volume: steady 60-70 mentions per day about pricing complexity, unchanged for 3 weeks. All mentions are confined to a single subreddit. No account over 500 followers has posted about it.
Specimen: synthetic, realistic
Is this a crisis? Score it against velocity, spread, and amplification and decide whether to escalate.
A meme account with 2M followers posts an unrelated joke that happens to mention the brand's name. 900 mentions in 6 hours, spread across X, Instagram comments, and TikTok. Sentiment is neutral to positive; zero product complaints in the set.
Specimen: synthetic, realistic
Is this a crisis? Score it against velocity, spread, and amplification and decide whether to escalate.
One enterprise customer's public post about a billing-data outage climbs from 25 to 300 mentions in 24 hours. It moves from LinkedIn to a tech newsletter with 80,000 subscribers by hour 30.
Specimen: synthetic, realistic
Is this a crisis? Score it against velocity, spread, and amplification and decide whether to escalate.
200 mentions appear in 2 hours, all from accounts created in the last week with zero followers, reposting identical scraped text with no real engagement (likes, replies) on any post.
Specimen: synthetic, realistic
Final deliverable
A triage memo classifying each of the 5 mention spikes as escalate/monitor/ignore, with the velocity-spread-amplification reasoning for each call.
See a reference example
Twilio, Incident Mention Triage (excerpt) Spike A (API status-page complaint): ESCALATE. 15->210 mentions in 18h, spread from X to Hacker News, amplified by a developer with 30K followers. Spike B (recurring pricing gripe, flat 3 weeks, one subreddit): MONITOR, no action this week. Spike C (bot repost cluster, zero-follower accounts): IGNORE, filtered as spam.
Success criteria
You're done when you can:
- Correctly identifies both real-crisis specimens (spike-1, spike-4) using all three framework signals
- Correctly rules out both noise specimens (spike-2, spike-5) and the viral-but-harmless distractor (spike-3) without escalating on volume alone