Per-Post Analytics that
actually retrains the AI

When you publish through Content Drifter, every post gets a unique identifier we follow into the wild. Every 30 minutes, our analytics worker pulls the latest engagement numbers (likes, comments, shares, views, impressions, reach, and engagement rate) for every account that post landed on.

The result is a per-post performance history, not a one-time snapshot. You see how a post grew over its first 24 hours, where it plateaued, and which accounts amplified it. That history lives forever in your dashboard and, more importantly, in your AI's context.

1. You publish. A single click sends your post to all platforms via Outstand's publishing API. Each platform replies with a post URL we tag.
2. We measure. Adaptive cadence: hourly for the first six hours, then every six hours for two days, daily for the first week, then weekly. We don't spam the APIs and you don't miss the spike.
3. AI adapts. Your top three posts and bottom two from the last 30 days are distilled into a short brief that gets injected into the system prompt every time you generate new content. Patterns that work get amplified. Patterns that don't get retired.

This is what we mean by “closed loop”: the output of one batch becomes the input of the next. Every other AI writer treats generation as a one-shot transaction. We treat it as a feedback system.

A refresh tick has a budget: it can only call so many platform APIs before it has to stop. Ours used to spend that budget on the newest posts, which sounds sensible and is not. Rank by publish date only ever falls as you publish more, so a post that dropped below the cut-off could never come back into the candidate set, no matter how many ticks ran.

The number that would have shown it was not on the record. Every run reported 50 posts considered, because 50 was the budget, and it read exactly 50 on all 106 recorded runs while 188 published posts sat waiting and were never captured once. A capped number cannot tell a drained queue from a starving one.

Two things changed. Selection is now ordered by how stale a post's capture is rather than by how recently it published, so the oldest uncaptured post is always first in line. And every tick now records the uncapped count of what was still due, so a backlog that never falls is visible on the run record instead of hiding behind the budget.

We are telling you this because the alternative is a page that says "we track every post" and hopes you never check. If your account publishes faster than the budget drains, that backlog number is the one that will say so.

Open Dashboard → Analytics and the first thing you see is the “What's Working” panel. Two columns:

• Top performing: your highest-scoring posts, ranked by a weighted formula: comments x 3, shares x 5, plus an engagement-rate multiplier so small accounts can still compete with viral ones.
• Underperformers: the posts that didn't land. We surface them on purpose. Knowing what failed is half the value.

You can switch the window (7 / 30 / 90 days), refresh on demand, and click into any post to see the full per-account breakdown.

Generic AI writers have one job: produce a post that sounds plausible. They have no signal whether the post worked. So next week's output is just as random as this week's.

A closed-loop system compounds. The longer you use it, the more it looks like a strategist who knows your audience. After 30 days of publishing, the AI has rewritten its own “style guide” for your account based on what your readers actually rewarded.

• Lower cost per win. Less time spent generating posts that go nowhere.
• It gets less generic as it learns. The first batch is written from researched per-network data. Once your own posts have been measured, later batches are shaped by what your audience actually rewarded. We do not publish a multiplier for that, because we do not hold a cross-customer benchmark that would make one honest.
• Real evidence for the boss. Every dashboard view exports as a CSV. Take it to the meeting.

Engagement data is yours. We never train any model that's shared across customers on your post content or numbers. The signals only shape your account's prompt context, and they're dropped after 90 days unless you keep the post archived.

Outstand is the publishing API behind the scenes. They have OAuth permission scopes; we only request what's needed for analytics and publishing.

The full closed-loop system (analytics fetching, the What's-Working panel, and prompt injection) is included from Starter up. Pro and Agency add per-platform diagnostics and your monthly Month in Review on top of the same loop.