Actions
Actions keeps two facts separate:
- Did the work ship? This is the work axis: Suggested, This week, Queued, In progress, then shipped.
- Did it earn anything? This is the outcome axis: Measuring, Earned, No movement, Couldn't measure, Reverted, or Coincided with a drop.
They never merge into one status. Done and No movement is a real, honest, visible result. The work happened, but the chosen measure did not clear its fixed bar. Conflating those facts is how software takes credit for activity.
The Actions page gives both axes one home. You decide what needs attention, choose the next work, follow what shipped through measurement, and read the result without switching between separate trackers.
One list, one status column
Every row represents one piece of work. The status column tells you where it stands. The rest of the row names the work type, page, owner, timing, and evidence.
You will not see a second suggestion list on a page record or a different set of names in Optimize. A problem should read as one problem wherever you find it. When it becomes a Suggested action, the same item moves through the work axis and then into Results.
This also explains why shipped work can leave the active plan without disappearing. Once the fixed measurement window starts, the row belongs in Results. The work is done; its outcome is not known yet.
The four tabs
The tab names and subtitles tell you what each view is for.
| Tab | Product subtitle | Use it for |
|---|---|---|
| Needs you | “Decisions waiting on you. Nothing here applies itself.” | Approvals, questions, and changes that need a choice before anything happens |
| This week | “What to do next, and what the last round earned.” | The current plan, recent momentum, and the work already under way |
| Suggested | “Everything your Agent has found, ready to start.” | The full set of Suggested actions, with the evidence behind each one |
| Results | “Work stays here from its frozen window to its judged result.” | Measuring work and every judged outcome |
This week is the default and keeps a clean /actions URL. Direct links can open another tab.
Old bookmarks keep working. ?view=proof and ?view=measuring both open Results, so links sent before the tabs changed do not break. New links use ?view=results; the old query words are compatibility aliases only.
Needs you
Nothing in this tab applies itself. A row may ask you to approve a prepared change, answer a question that changes the plan, or decide what to do after a measured drop. Open the row to see the exact choice, its evidence, and what each button will do.
If a change can be safely reversed, the decision can include a revert. If the platform cannot restore the original, Trakkr says so instead of showing a button that cannot keep its promise.
This week
This is the short plan, not the whole backlog. It puts the next work beside a quiet summary of what the last set earned. Use it to answer two questions in one read: what should we do now, and did the previous work move enough to count?
The plan can include work for you or the Agent. Who does it depends on the permission saved for that work type, not on the wording of the suggestion.
Suggested
Suggested holds everything the Agent has found that is ready for a decision. Each row states the work, the evidence, the page or source it concerns, and what measurement will follow if you start it.
Starting does not mean the work has already happened. It accepts the plan, freezes measurement, and moves the item onto the work axis.
Results
Results contains both open measurement windows and judged outcomes. A Measuring row shows the current day and full window. A judged row uses the shared result words and keeps the before and after evidence attached.
Read Results for the earned bar, connection gaps, rollback rules, earn rate, and quarter recap.
The nine work types
The work type says what kind of change this is. It does not replace the status.
| Work type | What it means | What Trakkr measures |
|---|---|---|
| Technical fix | Repair access, rendering, metadata, or another technical fault on a page | Usually AI bot fetches and machine-read recovery over 14 days |
| Page improvement | Improve an existing page so more of the people who see it choose it | Usually Search Console clicks, with impressions and AI referral clicks as support |
| Content refresh | Update an existing page for a question, citation gap, or stale claim | Usually new citations of that page over 42 days |
| New content | Publish a page that does not exist yet | Usually new citations of the published page over 42 days |
| Outreach draft | Prepare a focused pitch to a source that AI already uses | Citations from that source and whether it starts mentioning your brand |
| Reply draft | Prepare a useful reply for a discussion thread | Whether AI cites the thread and whether the thread mentions your brand |
| Campaign | Coordinate several linked changes over a longer period | A combined view of citations and crawler fetches over 70 days |
| Setup | Connect a source or enable a delivery path | Not measured as an earned outcome because it unlocks other work |
| Watch | Keep checking a page, source, or signal for a meaningful change | No earned window; it surfaces a new decision when the watched fact changes |
The exact frozen plan is shown before you start. That saved plan has the final word for the page, measure, and window.
What happens when you start
Starting is a commitment to a clear plan, not a claim that the change is complete.
At confirmation, Trakkr stamps the measurement plan and freezes it. Nobody can swap the measure later because a different number looks better. Setup work skips measurement because it enables a source or delivery path rather than earning a page-level outcome.
Once the work ships, it moves to Results as Measuring. It stays there until the window closes and the data pipeline can call a result. The Agent does not mark its own work Earned.
How much the Agent can do
Each work type has its own permission. You can keep sensitive work at review while letting lower-risk work move further.
| Permission | What happens |
|---|---|
| Suggest only | The Agent brings you findings. Nothing changes without you. |
| Prepare drafts | The Agent prepares the work. You approve before anything ships. |
| Do it, then tell me | The Agent makes a reversible change, tells you right away, and you can revert. |
| Handle it | The Agent makes allowed changes up to the weekly limit and reports each one. |
Permission alone is not enough for a live change. The target must belong to your site, the platform must support the change, an original value must be saved, the weekly limit must have room, and the safety switch must be clear. If a check fails, the work falls back to Needs you.
See Agent permissions for the full rules and limits.
Where Suggested actions come from
Trakkr creates Suggested actions from evidence already connected to the brand. Common sources include:
- prompt results and lost positions
- citation gaps and lost citations
- Optimize findings
- AI crawler access and fetch patterns
- competitor movement
- Search Console page gaps
- Reddit and other discussion threads
- a diagnosis you chose to act on
- a manual action added by your team
Some suggestions come from fixed rules. A blocked page can create a matching Technical fix as soon as the evidence lands. Others combine several measured facts so the Agent can explain why the work matters now.
The suggestion still has to name concrete work and evidence. It should not ask you to “improve visibility” or track more data as a substitute for changing the site, content, outreach, or reply.
Duplicate signals update the same Suggested action instead of filling the list with copies. When the source fact is no longer true, stale suggested work can retire without erasing work you already started.
Repeated work and Automations
If the same work should happen whenever a known condition appears, Automations is the place to set that rule.
Common questions
How is this different from a to-do list?
A to-do list contains what someone remembered to type. Actions is tied to measured evidence, a page or source, a work type, and a frozen result plan. It can also retire a stale suggestion when the source fact changes.
Will Trakkr really push changes to my live site?
Only when your permission, platform support, saved original, target scope, weekly limit, and safety checks all allow it. Suggest only and Prepare drafts never publish a live change without your approval. If a platform cannot reverse a change, the Agent does not apply it on its own.
What is the difference between done and earned?
Done means the work shipped. Earned means its fixed before-and-after measure cleared the published bar when the full window closed. Done with No movement is possible and stays visible.
Why did my action move to Results?
The work shipped, so it left the active plan and entered its measurement window. It will show Measuring until the window closes, then a judged result.
What happens if I dismiss something?
It leaves the Suggested list. The same fact does not immediately create the same item again. If the underlying evidence changes in a meaningful way later, Trakkr can surface a fresh Suggested action.
Can I change the measurement after starting?
No. The plan is frozen when you start. Dismiss the incorrect item and start a new one with the right page or measure.
Why is something still Measuring after the number went up?
Trakkr waits for the full window. Calling a result early would apply a different rule to work that looks good halfway through and work that does not.
What happens when Trakkr cannot measure it?
The row becomes Couldn't measure, shows the missing evidence, and stays out of the earn-rate denominator. It does not trigger a rollback.