← All work

Analytics SaaS · 2026

Channel Spark

A YouTube analytics workspace for agencies that manage many client channels. It imports each channel's full history, surfaces ready-made insights and writes an AI strategy report backed by the numbers.

Demo with fictional data.

The situation

Account managers pulled numbers from YouTube Studio one channel at a time, every month, and still struggled to say what each client should do next.

What changed

Every client channel is connected once and kept up to date with its full history. Insights are ready on arrival, and an AI strategist turns the data into a prioritised plan backed by the numbers.

Built for agencies and studios that run YouTube channels on behalf of several clients.

“We used to spend the first week of every month pulling numbers channel by channel. Now the data is already there, and the conversation with each client starts from what to do next.”
Jonny MacIntoshCEO at Listener Kids

Goals

  • Bring private YouTube analytics and revenue data for many channels into one workspace.
  • Keep the full history, not just the last 28 days.
  • Turn raw numbers into prioritised, evidence-backed recommendations that an account manager can act on.

What it does

  • Client management with a connect link per client, so channel owners grant read-only access in a couple of clicks.
  • Daily, 90-day and full-history imports, with live progress and cancellation.
  • An overview with KPIs, trends, top videos, a video library and a detail page for every video, covering views, traffic sources, countries, devices and revenue.
  • Ready-made insights, including hidden gems, underperforming uploads, repackaging candidates, best topics, optimal length, publishing consistency and top earners.
  • Revenue reporting, sync logs and internal notes per client.
  • A one-click disconnect that revokes access and deletes all of that channel's data.

AI and automation

  • An AI strategist reads a compact summary of the channel and writes a structured report. It covers an executive summary, headline metrics with trends, quick wins, growth bets, risks and long-term strategy, each with its rationale, evidence and priority.
  • The model is limited to the data provided and must cite numbers. Its output comes through a strict schema instead of free text, so it renders as a clean, consistent report.
  • Reports can cover a whole channel or a single video over any date range, and they are saved for later comparison.
  • Long imports run as self-continuing background jobs that work around serverless time limits, with progress streamed to the browser.

Engineering notes

  • Imports are idempotent, and recent days are overwritten as YouTube finalises its data over about 72 hours.
  • The importer works within the API's real limits, querying per video where channel-wide breakdowns are not supported and spending quota only where it matters.
  • Only the read-only Google scopes are requested, with public privacy, terms and data-deletion pages for Google's OAuth verification.