# Ping Slack when Algolia zero-result searches spike

> Every weekday morning we check yesterday's Algolia search health against a 14-day baseline and only post to Slack when the zero-result rate is really off.

- Workflow type: agent
- Services: Algolia, Slack
- Categories: Product, Operations
- Published: 2026-07-17

## What it does

- Checks yesterday's Algolia search volume and zero-result queries against a trailing 14-day baseline every weekday morning.
- Posts to Slack only when the zero-result rate spikes past your threshold or a brand-new failing query breaks into yesterday's top 10 by volume.
- Names each offending query, its search count, and suggests a concrete fix such as adding a synonym, creating a rule, or flagging a content gap.
- Stays silent on healthy days so the alert channel doesn't become background noise you learn to ignore.

## What you'll need

- An Algolia account with analytics access to the index you want to monitor.
- A Slack workspace and the channel you want the alerts to land in.

## Prompt

Every weekday at 8am, check my Algolia search health and only ping Slack when something is actually off.

Pull yesterday's total search count and top zero-result searches from Algolia analytics for my index. Also pull the same figures across the trailing 14 days so you have a baseline to compare against.

Compare yesterday's zero-result rate against the trailing 14-day baseline. If yesterday's rate is materially above the baseline (expose the spike threshold as a configurable knob, defaulting to somewhere around 1.5x the baseline rate or +2 percentage points, whichever is larger), or if a brand new zero-result query cracked yesterday's top 10 by search volume, that counts as an anomaly worth flagging. Algolia's own guidance is that a no-results rate above 3-5% is not optimal and teams should target under 2%, so use that as a sanity check on whether the baseline itself is already unhealthy.

When there is an anomaly, draft a concise Slack message to the configured channel (also a knob) that names each offending query, its search count for yesterday, and a suggested fix. Pick one of: add a synonym for a likely misspelling or alternative phrasing, create a query rule to steer the result, or flag it as a content gap the catalog does not cover. Keep the message short and skimmable, do not paste raw analytics tables.

If yesterday's numbers look normal, post nothing at all. The channel should stay quiet on healthy days so people actually pay attention on the days you do speak up. This runs alongside the existing weekly Monday recap, which fires unconditionally, so anything routine belongs there, not here.

## How to customize

- Change the spike threshold (default: yesterday's zero-result rate materially above the 14-day baseline).
- Pick which Slack channel the alerts go to.
- Adjust the schedule if you want a different cadence than weekdays at 8am.

## FAQ

### What counts as a spike?

By default we compare yesterday's zero-result rate against the trailing 14-day average and only alert when it's materially higher. You control the threshold, so you can start strict and loosen it if it turns out to be too chatty.

### Will this spam our Slack channel?

No. If yesterday looks like a normal day, nothing gets posted. The whole point is to stay quiet on healthy days and only speak up when there's something worth looking at.

### What if a single query goes bad but the overall rate looks fine?

We also alert when a brand-new zero-result query breaks into yesterday's top 10 by search volume, so a spike in one specific broken query still gets flagged even when the aggregate rate stays flat.

### Does it suggest fixes or just report the problem?

For each flagged query the agent suggests a likely fix. That's usually one of adding a synonym for a misspelling or alternative phrasing, creating a rule to steer the result, or flagging it as a content gap the catalog doesn't cover yet.

### Can I run this weekly or hourly instead of daily?

Yes. The schedule is a knob. Weekly is fine if your search volume is lower; hourly is overkill for most teams but works if you want it.

Use this prompt in General Input: https://www.generalinput.com/prompts/ping-slack-when-algolia-zero-result-searches-spike