The most cost-effective approach is usually:

1) Use a cheap/“lite” topic tracker for collection

Pick a tracker that can monitor:

• keywords/topics
• RSS/news sources
• social mentions if needed
• basic filtering

Goal: keep the raw monitoring layer inexpensive and stable.
If you only need daily trend alerts, you usually do not need a full enterprise listening suite.

2) Aggregate once per day, not continuously

Instead of processing every mention in real time:

• pull data on a schedule (e.g., every 24 hours)
• store counts, top terms, and source breakdowns
• compare today vs. yesterday / 7-day average

This reduces API calls, compute, and noise.

3) Build alerts from simple trend rules

The cheapest useful alerting logic is usually rule-based:

• volume spike: mentions up > X% day-over-day
• threshold crossing: topic exceeds Y mentions
• velocity change: 3-day moving average increases sharply
• new keyword emergence: a new subtopic appears in top terms

These rules are easy to implement and usually enough for daily alerts.

4) Use a lightweight delivery channel

Send alerts through inexpensive channels like:

• email
• Slack/Teams webhook
• SMS only for high-priority alerts

Avoid paid notification platforms unless you need advanced routing.

5) Keep the data model simple

Store only what you need:

• topic
• timestamp/day
• mention count
• source type
• top entities/keywords
• alert status

A simple database or spreadsheet can work at small scale; a small SQL database is usually best once volume grows.

6) Prefer automation over custom dashboards

If the requirement is just “daily trend alerts,” dashboards add cost without much value.
A scheduled job + rule engine + notification is typically the lowest-cost setup.

Cheapest practical architecture

Topic tracker → daily scheduled export/API pull → small database/spreadsheet → rule checks → Slack/email alert

This is often the best balance of:

• low subscription cost
• low engineering cost
• low maintenance
• fast time to value

When to upgrade

Spend more only if you need:

• many topics at scale
• complex sentiment/entity analysis
• deduping across sources
• real-time alerts
• multi-user workflows and approvals

If you want, I can also suggest:

1. a no-code version,
2. a low-code version, or
3. a custom stack on a small budget.