Building a creator tracking system for an agency

by CIA updated 21/09/2026 3 min read

A four-table structure for an agency creator tracking system

Also available in: Tiếng Việt

An agency tracking many creators for many clients outgrows a spreadsheet fast. The minimum structure needs four tables and one update rule. This is a design, not a line-by-line tutorial.

Four tables

Table 1 — channels. One row per creator: channel ID, platform, niche, which client cares about them. You fill this yourself; it does not come from the API.

Table 2 — snapshots. One row per measurement per channel: channel ID, measurement date, median views, engagement rate by views, follower count, number of posts in the 45 videos, date of the oldest video.

That last column matters: it tells you how long the 45 videos span, and two channels only compare when the spans are similar. See posting frequency.

Table 3 — runs. One row per API call: channel ID, tool, date, the Idempotency-Key used, credits charged. This table reconciles cost and stops duplicate calls.

Table 4 — campaigns. One row per booked post: channel, client, post date, views, interactions, unique discount code. This is the only table that connects public data to revenue.

The update rule

One snapshot per channel per month. Enough for a trend line, not too expensive. For channels in an active campaign, measure before and after.

The Idempotency-Key contains the channel ID and the period. For example a hash of "channel 123, 09/2026". Within the month the key is the same, so a retry is not charged twice; next month the key changes. See what Idempotency-Key does.

Never overwrite an old snapshot. The value of the system is its history.

Cost and how to control it

Tool Price Use when
Channel report 5 credits monthly snapshot, every channel
Audience analysis 12 credits only before confirming a booking
Comments analysis 12 credits after each booked post
Multi-platform report 20 credits only when a client asks about other platforms

For 50 channels, a monthly snapshot is 250 credits. That is a predictable number, and that is the point of scheduling only the channel report and running the more expensive tools on demand.

Set a limit in your code: maximum runs per day. Automation makes fetching easy, and easy means expensive.

What this system does not do

It does not connect views to revenue unless you have a unique discount code in table 4. It does not store creator contact details — the API returns no email addresses or phone numbers, and that is a hard boundary.

And it does not replace watching by eye: edit quality, tone, how they answer comments are things no column measures.

Frequently asked questions

Is there a database template? No. The four tables above are the structure; data types and technology are your choice.

Monthly or weekly measurement? Monthly for the general list; weekly only for channels in an active campaign.

Is the data official analytics? No. It is an estimate from public data the platform displays.

More on this topic: Data & API