Growth Actions

Summary#

Growth Actions reads the AI-visibility tools you have already run and turns them into one ranked to-do list. Each action carries a category, an impact score, an effort estimate in days, a win probability, the reasoning behind it, and two to five concrete steps. It costs 3 credits, and it does not charge when it has nothing to read.

Purpose#

Research produces findings; findings do not produce order. After a visibility scan, a competitor report and a question map, the honest question is not what did we learn but what do we do on Monday, and in what order.

Growth Actions exists to answer exactly that. It converts measurement into a sequenced queue, so the decision it supports is the only one that matters after research: which single thing to start.

Overview#

The queue is built from your own saved runs — nothing else. For each of six source tools, Metric Vault takes your newest saved result, preferring one whose target matches the domain you entered:

Those payloads are trimmed and passed to a language model with one instruction that governs everything: ground every action in the provided data, name the specific competitor, question, prompt or gap it addresses, and never invent a metric. It returns 6 to 15 actions, each scored.

Metric Vault then re-ranks them itself. Priority rises with impact and win probability and falls with effort, so a big win that takes a month can sit below a smaller win you can ship this week. The top 20 are numbered 1 to N.

If you have not run any of the six source tools, nothing is generated and nothing is charged. The screen tells you what is missing instead.

Benefits#

• One ordered list instead of six reports to reconcile.
• Every action names the evidence it came from, so nothing is generic advice.
• Impact, effort and win odds are on the card, so you can pick by capacity.
• Done and Dismiss are remembered per domain, so the queue behaves like a real backlog.
• No charge at all when there is no data to work from.

Use Cases#

Deciding the sprint. You have the research and need the top five things to do in the next two weeks.

Handing work to someone else. Each card copies as a brief with steps and target keywords, ready to paste into a ticket.

Justifying the plan. Every action carries a Why this line traced to a real measurement, which is what a stakeholder actually asks for.

Re-planning after a change. Run more source tools, regenerate, and the ordering updates against the newer evidence.

Choosing between two options. When two plays look equal, the effort and win-odds figures break the tie.

Requirements#

• A signed-in account. Signed out, the queue shows Sign in to generate Growth Actions.
• A paid plan. At 3 credits, this tool is metered against your monthly quota and is blocked on Free.
• At least one saved run from one of the six source tools listed above. The Generate button stays disabled and reads Run a tool below to unlock until that exists.
• 3 credits of monthly allowance available.

Permissions#

Cost#

3 credits per generated queue. The cost chip on the button reads 3 credits.

At 3 credits this tool sits at the premium threshold, so it is metered against your monthly quota rather than only the hourly fair-use limit. See How credits work and Quotas and rate limits.

Two things are free: a run that finds no source data is not charged, and reopening a previously generated queue for the same domain costs nothing — it loads from your saved work. Regenerating charges again.

Navigation Path#

Dashboard → Get Started → Growth Actions

The view header carries a green LIVE chip.

Inputs#

The field is disabled until at least one source tool has been run. Pressing Enter in it generates the queue.

Step-by-Step Guide#

1. Open Dashboard → Get Started → Growth Actions.
2. Look at STEP 1 · FEED THE QUEUE — RUN ANY ONE TO UNLOCK and the progress chip reading FED 0 / 4.
3. Click Run now on any one of the four funnel cards — Visibility Overview, Competitor Research, Questions or Prompt Tracking — and complete that run. One is enough.
4. Come back to Growth Actions. The chip updates, and the Generate button unlocks.
5. Type your domain into the field under STEP 2 · DOMAIN TO GROW.
6. Click Generate growth queue (3 credits), or press Enter.
7. Wait for Building your growth queue… to clear.
8. Work from the top card down. Use the filters and sort to reshape the list.

Reading the Results#

The funnel chip. FED n / 4 counts how many of the four funnel tools you have a saved run for. A fed card gets a checkmark. The chip is a coverage gauge, not a gate — one run unlocks generation, and Brand Perception and Narrative Drivers also count toward unlocking even though they are not among the four cards.

The status line. Under the button, once a queue exists: Updated <time ago> · N actions · sources: <tool names>. The sources list is the important half — it tells you exactly which of the six runs the queue was built from. A queue built from one source is thinner than one built from four, and the line is how you know which you are looking at.

The filter bar. Category chips (All, plus Content, Backlinks, Technical, AI Visibility, Brand, Keywords with counts) and a Sort group: Best first, Impact, Least effort, Win %. Best first is the default and the only one that uses the composite priority. Switch to Least effort when you have a short window, and Impact when you have a quarter.

An action card, collapsed. A rank number, the category, the title, and three metric tiles.

The colors are consistent: green means favourable, amber means acceptable, red means the number argues against doing this first.

An action card, expanded. Click the header to open it.

• The description — what the play actually is.
• Why this — the rationale, tied to the specific finding it came from. If this reads generically, the queue was built from thin source data; run another source tool and regenerate.
• Steps — two to five concrete actions. This is the part you paste into a ticket.
• Target keywords — the terms the play is aimed at, when the source data had them.
• From: — the source tools behind the action, at the foot of the card.

The card buttons. Open <tool> jumps to the Metric Vault view most useful for executing or tracking the play, carrying your domain across. Copy brief puts the whole card on your clipboard as plain text — category, title, description, why, steps and keywords. Mark done greys the card and sinks it to the bottom of the list; Undo brings it back. Dismiss removes it from the queue entirely. Done and dismissed state is stored per domain in your browser, so it survives a regeneration but does not follow you to another machine.

What good and bad look like overall. A healthy queue has several actions under two days of effort in the top five — that means there is real work you can start today. A queue where everything is above four days usually means you have only fed it one source tool and the model reached for big structural plays; feed it another and regenerate. A queue with no AI Visibility category at all normally means the Visibility Overview run is missing or stale.

What to do with it. Take the top three cards. Copy each brief into your tracker. Mark them done as they ship. Regenerate when you have run new source tools, not weekly out of habit — the queue only changes when the evidence does.

Examples#

Screenshots#

Tips#

• Run Visibility Overview first. It is the richest single source and the one most actions are built from.
• Sort by Least effort when you have a week rather than a quarter.
• Use Copy brief rather than retyping. It carries the steps and keywords intact.
• Dismiss the plays you will genuinely never do. A honest queue is a shorter one.
• Regenerate after you add a new source run, not before — the queue can only be as good as the evidence behind it.

Best Practices#

• Feed at least two source tools before you rely on the ordering. One source produces a plausible queue; two produce a defensible one.
• Enter the same domain you ran the source tools on, so the newest matching results are used rather than an unrelated run.
• Treat effort as the scheduling number and impact as the argument. Ship the low effort items and use the high impact ones to justify headcount.
• Keep the queue for one domain per workspace target. Done and dismissed state is keyed to the domain.
• Pair a stale-looking queue with a fresh Brand Health run, which feeds several sources at once.

Common Mistakes#

• Expecting a queue without running anything. Growth Actions never invents data; with no source runs it refuses and charges nothing.
• Entering a different domain from the one you researched. You will get the most recent runs for those tools regardless of target, and the actions will be aimed at the wrong site.
• Regenerating repeatedly. Each regeneration costs 3 credits and, without new source data, produces much the same list.
• Reading impact as guaranteed traffic. It is a composite score, not a forecast.
• Assuming Done and Dismiss sync across devices. They are stored in the browser.

Limitations#

• Six fixed source tools. Site audits, backlinks and PPC results are not read.
• Only the newest saved run per source tool is used, and each payload is trimmed before it is sent, so very large results are sampled rather than read whole.
• Source runs older than 90 days have been purged and no longer count.
• The queue is capped at 20 actions.
• Effort, impact and win probability are model estimates, calibrated by instruction and clamped to sane ranges, not measurements.
• Done and dismissed state lives in your browser, per domain.
• There is no scheduling and no alerting on the queue.

Troubleshooting#

FAQs#

Do I have to run all four funnel tools? No. One is enough to unlock generation, and the screen says so. More sources make a sharper queue, which is what the FED n / 4 chip is tracking.

Am I charged if I have not run anything? No. When there is no source data the request returns the unlock message before any credit is taken.

Why is this 3 credits when most tools are 1? It runs one large synthesis over your combined tool output, which costs more on the backend than a single-endpoint tool. Credits track backend cost — see How credits work.

Where do impact, effort and win probability come from? They are produced by the synthesis under explicit scoring instructions — a copy edit is roughly half a day to a day, a new page two to three, an outreach push three to five — and then clamped and re-ranked by Metric Vault. Read them as calibrated estimates.

How is the ranking calculated? Priority weights impact most heavily, adds win probability, and subtracts a penalty that grows with effort up to a cap. That is why a smaller, faster win can outrank a larger, slower one.

Can I export the queue? Yes. Export PDF, Export Excel, Export CSV, Export JSON and Share public link sit at the top of the view and activate once a queue exists.

Does marking an action done tell my teammates? No. Done and dismissed state is stored in your own browser, keyed to the domain.