Where the numbers come from, and how to read them.
No black box. Here are the seven sources, how we measure each one, what counts as a spike, and how to use it all in practice.
Seven public sources, one view.
Each source answers a different question. Together they tell you whether the attention is real.
Google News
- What we watch
- How many news articles mention your keyword in the last 3 hours.
- How we match
- The exact phrase, in English-language news.
- How often
- Every hour on Free and Starter, every 15 minutes on Pro and Business.
- You are alerted
- Coverage is at least 3 times its usual level, at least 4 articles above it, and clearly outside its normal swings.
Hacker News
- What we watch
- Stories about your keyword posted in the last 24 hours on the tech community site.
- How we match
- The exact phrase in stories.
- How often
- Every hour on Free and Starter, every 30 minutes on Pro and Business.
- You are alerted
- A story less than 6 hours old reaches 50 points.
Wikipedia
- What we watch
- Daily views of the English Wikipedia article about your keyword, by people, not bots.
- How we match
- We find the best matching article once, then follow it. No article? The dashboard says so.
- How often
- Every 6 hours. Wikipedia publishes views per day, so the newest value is yesterday.
- You are alerted
- Views are at least 2 times the usual day, at least 150 higher, and clearly outside normal swings.
Steam
- What we watch
- Players online right now in the Steam game with that name.
- How we match
- The first game in the Steam store whose name contains your keyword. Not a game? The dashboard says so.
- How often
- Every hour on Free and Starter, every 30 minutes on Pro and Business.
- You are alerted
- Players are at least 1.3 times the usual number for that time of day, and at least 1,500 higher.
Google Trends
- What we watch
- Whether your keyword is among the trending searches on Google right now.
- How we match
- Your keyword is in the trending topic, or in at least two of its headlines. United States and United Kingdom.
- How often
- Every 15 minutes on every plan.
- You are alerted
- Your keyword appears in the list.
GitHub
- What we watch
- The most starred new repositories of the last 7 days.
- How we match
- Your keyword is in the repository name or description.
- How often
- Every 15 minutes on every plan.
- You are alerted
- One of those repositories mentions your keyword.
Product Hunt
- What we watch
- The latest launches featured on Product Hunt.
- How we match
- Your keyword is in the launch title.
- How often
- Every 15 minutes on every plan.
- You are alerted
- A featured launch mentions your keyword.
Google Trends, GitHub and Product Hunt are events, not curves: you see Trending, New repo or Launched. GitHub and Product Hunt cover only the top new items, so most keywords stay quiet there. That is expected.
The usual level, and what a spike means.
- Now
- The value from the latest check.
- Usual level
- The average of the earlier checks in the last 24 hours (9 days for Wikipedia). It appears after three checks. Until then the status is Learning.
- Change
- How far Now is from the usual level, in percent. Usual 3 and now 54 is 18 times higher, shown as +1700%.
- Status
- Learning: not enough data yet. Normal: close to usual. Rising: clearly above, for news from 1.5 times usual. Spike: above the alert line, for news from 3 times usual.
Every keyword has its own usual level in every source, so a small brand and a global one are each judged against themselves.
For alerts we go one step further where we can: news and Steam are compared with the same time of day on earlier days. A game that is always busy in the evening does not alert you every evening.
Quiet by design.
A real jump
An alert needs a big ratio and a big absolute change together. Going from 1 article to 2 is +100%, but it is not a spike.
No repeats
After an alert, the same thing stays quiet for a while. You are not pinged again about a story you already know.
One signal, every channel
The same alert goes to your dashboard log and to Telegram. Business can also send it to a webhook, and Pro and Business can ask Claude or ChatGPT about it.
Quiet periods
| Google News, Steam | At most one alert per keyword every 3 hours |
| Google Trends | Once a day per trending topic |
| Wikipedia | Once per day of data |
| Hacker News | Once per story |
| GitHub, Product Hunt | Once per repository or launch |
When alerts begin
Hacker News, Wikipedia, Google Trends, GitHub and Product Hunt can alert from the first check. News and Steam first need to learn what is normal: about 3 hours on Pro and Business, about 10 hours on Free and Starter.
You will get a message here when one of your keywords breaks out. Send /stop to pause alerts.
Usually about 3 articles, now 18.0× higher
• GTA 6: everything we know so far (IGN)
• The new GTA 6 trailer, scene by scene (Polygon)
• Why everyone is talking about GTA 6 today (Eurogamer)
Open in dashboard
Searches today: 200K+
Open in dashboard
A simple way to read the numbers.
- Start with the Overview. Count the sources that are spiking. One is a hint. Two or three together usually mean real news.
- Read the percentage with the usual level. +300% from a usual level of 1 is small. +80% from a usual level of 50,000 is big.
- Open the headlines before you act. Click the source to see what the articles actually say. We show that attention jumped, not why.
- Check that it is your topic. A short or common word can match other things. A more specific keyword gives cleaner signals.
- Keep the history. Pro and Business can export the signal log as CSV for reports.
How often we check.
| Free | Starter | Pro | Business | |
|---|---|---|---|---|
| Keywords | 3 | 15 | 75 | 300 |
| Google News | Hourly | Hourly | 15 min | 15 min |
| Hacker News, Steam | Hourly | Hourly | 30 min | 30 min |
| Google Trends, GitHub, Product Hunt | 15 min | 15 min | 15 min | 15 min |
| Wikipedia | 6 hours | 6 hours | 6 hours | 6 hours |
| Signal history | 5 latest, 7 days | 30 days | Full, about 13 months | Full, about 13 months |
| Telegram | 1 chat | 1 chat | 1 chat | 5 chats |
| Claude and ChatGPT connection, CSV export | No | No | Yes | Yes |
| Webhook, API, team seats | No | No | No | Yes, 5 seats |
Checks run in a queue, so a busy moment can add a few minutes.