> For the complete documentation index, see [llms.txt](https://docs.sellermagnet.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.sellermagnet.com/ppc-management-beta/ppc-manager-overview/ppc-approval-center.md).

# PPC Approval Center

Review, approve, or reject the PPC changes SellerMagnet queues for you, and read the evidence behind each one before it reaches Amazon.

{% hint style="info" %}
**Difficulty:** 🔴 Advanced · **Reading time:** \~20 min
{% endhint %}

{% hint style="success" icon="rocket" %}
**Open this page in your dashboard:** [**Go to PPC Approval Center \[BETA\] →**](https://dashboard.sellermagnet.com/ppc/approval-center)
{% endhint %}

{% hint style="warning" icon="triangle-exclamation" %}
**BETA:** The PPC Manager is currently in beta. Some features may behave unexpectedly or change based on user feedback. Please report any issues at <info@sellermagnet.com>.
{% endhint %}

## 📋 Overview

The **Approval Center** is your review queue for AI-generated optimization suggestions. Recommendations from the AI engine, Autopilot and your bid rules land here as rows. You decide on each one before it is pushed to your Amazon Ads account - or, if you chose a more automatic mode, you review only what the automation does not apply by itself.

This is the control layer between automation and your live campaigns. It lets you benefit from AI analysis while keeping the final decision on every change that matters. The Approval Center is part of both the PPC Manager and the PPC AI Optimizer plan.

> **Who should start with Manual Approval?** If you are new to PPC automation, manage high-spend campaigns, or work in a team where changes need oversight, Manual Approval is the mode PPC Settings marks as **Recommended**. Almost every suggestion waits here for your decision, so you see what the AI proposes before it goes live.

***

## Page Header and KPI Tiles

The page header shows quick stats - **Pending Review**, the current **Automation Mode**, and **Last refreshed** - plus a marketplace selector (**All Marketplaces** by default) that limits the queue to one Amazon Ads profile.

* **Pending Review:** the number of suggestions waiting for a decision. Snoozed suggestions are not counted.
* **Automation Mode:** the name of your current mode. Click it to open the mode settings in PPC Settings. While the master switch **Stop all automation** is on, it reads **Automation stopped** instead.
* **Last refreshed:** the time the queue was last loaded. Hover over it for the tooltip: "When the queue was last fetched from the database. This is not the time of the latest AI analysis or Amazon performance report." The queue auto-refreshes every 30 seconds.

Below the header, a row of **KPI tiles** summarizes the whole queue at a glance:

| Tile               | Shows                                                                                                                                                                               |
| ------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Pending Review** | Actions waiting on your approval. They expire if left unanswered (see Priority & Expiration below)                                                                                  |
| **Applied**        | Actions confirmed as applied to Amazon                                                                                                                                              |
| **Rejected**       | Actions you declined. Rejected negate, harvest and pause keyword suggestions are not suggested again for 7 days; other rejected suggestions may reappear sooner                     |
| **Expired**        | Suggestions that reached their expiry date without a decision, plus suggestions that left the queue because they could no longer be applied (see Where Did My Suggestion Go? below) |

### Banners and notes above the queue

Depending on your account, some of these appear above the KPI tiles:

* **Sandbox strip:** "Sandbox - changes are not sent to Amazon", with a **Go live in Settings** link. In sandbox mode recommendations are computed but nothing is sent to Amazon.
* **Automation stopped banner:** "Automation is stopped for this account. No automated change is sent to Amazon; suggestions still reach the Approval Center." A **Resume automation** button lifts the stop.
* **Target ACoS line:** when your target ACoS sits far above what the account actually runs at, you see "Target ACoS {target}% vs actual {acos}% over the last 30 days. The bid and budget increases in this queue are growth bets, not efficiency fixes - review them with that in mind." **Adjust target** opens PPC Settings; the close button (**Hide for 7 days**) hides the line for a week.
* **Expiry banner:** on the Pending tab, "{n} recommendations expire within {h} h (≈ {money} at stake)". It offers **Approve all safe ({m})** (only when a safe set exists), **Review** (opens the Pending tab sorted by **Sort: Expiring soon**) and, if the check failed, **Retry**. A second line adds: "Expired suggestions come back in the next cycle; after expiring twice in 30 days they are held for 14 days. Oldest waiting: {d} d."
* **Breaker banner:** when the latest automation cycle ran in protective mode, you see "Your ACoS ({acos}) is above the safety limit, so the engine only suggested cuts and pauses this cycle" (or the same sentence for ads that spent money without attributed sales), followed by how many suggestions that would have raised spend were held back.

### Automation details and expired history

A collapsed section named **Automation details and expired history** sits between the tiles and the queue. Open it to see:

* **Automation Mode:** your current mode with its description and an **Edit in Settings** link.
* **Autonomy ladder:** for each action lane, whether it is **Auto-applying** or **Queued for approval**, and why. Typical reasons are "Not enough scored evidence yet ({n}/{required\_n})", "Win rate below the bar ({win\_rate} of {required\_win\_rate} needed)" and "No evidence yet". If evidence tracking is not set up for the account, you see "Evidence tracking is not set up yet for this account. Every lane stays on manual approval until it is." Below the ladder, a line such as "Profit-aware autopilot covers {covered} of {total} advertised products" links to **Add product costs (COGS)** for the products that still need a cost.
* **Expired unapproved, last 30 days:** how many suggestions expired without a decision (bids, negatives and harvests counted separately), or "none".

***

## Automation Mode

Your automation mode decides which AI suggestions wait here for you and which the automation applies by itself. There are three modes:

| Mode                 | What happens                                                                                                                                                                                                                             | What still waits for you here                                                                                                                                   |
| -------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Suggestions Only** | The AI analyses your campaigns and queues its recommendations here. Nothing changes on Amazon until you click Approve.                                                                                                                   | Everything                                                                                                                                                      |
| **Manual Approval**  | Every suggestion lands here. If the auto-apply switch in PPC Settings is on, the worker also applies bid changes it scores at 55 % confidence or higher on its own, within your bid caps (this needs the accepted PPC Terms).            | Everything else                                                                                                                                                 |
| **Fully Automatic**  | The worker applies bid and budget changes on its own once they score high confidence, within your bid caps and per-run limits. Every change is logged and can be undone for 24 hours. This mode needs the accepted PPC Terms of Service. | Pauses, negatives, campaign state changes and budget raises above the per-run cap. Destructive changes, capped actions and cooldown-held items always wait here |

You can approve from this page in every mode, including Suggestions Only. In Fully Automatic mode each action lane (bid optimizer, budget optimizer and so on) only applies by itself once its measured win rate and sample size clear the bar shown in the **Autonomy ladder**. Until then its suggestions queue here for your approval.

The mode itself is set on the **PPC Settings** page - use the **Edit in Settings** link in the details section, or click the **Automation Mode** stat in the page header, to change it there.

### Stop all automation

Separate from the three modes, a master switch named **Stop all automation** sits at the top of the PPC pages. When you confirm it, Autopilot, bid rules, dayparting, scheduled changes and the budget brake stop sending changes to Amazon right away. Suggestions still reach the Approval Center, and you can lift the stop at any time with **Resume automation**. While it is on, the header stat reads **Automation stopped**. There is no "Paused" mode - this switch is the way to pause automation.

### Choosing the Right Mode

| Factor                                    | Suggestions Only                                                                                                                            | Manual Approval                                                                        | Fully Automatic                                                                                                        |
| ----------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------- |
| **Tag in PPC Settings**                   | Read-only                                                                                                                                   | Recommended                                                                            | Advanced                                                                                                               |
| **Changes that reach Amazon without you** | None                                                                                                                                        | Only bid changes at 55 % confidence or higher, and only if the auto-apply switch is on | Bid and budget changes within your caps and limits                                                                     |
| **What waits in this queue**              | Every suggestion                                                                                                                            | Everything else                                                                        | Pauses, negatives, campaign state changes, capped actions, cooldown-held items and lanes that have not cleared the bar |
| **Reversal of applied changes**           | [Activity Log](https://docs.sellermagnet.com/ppc-management-beta/ppc-manager-overview/ppc-activity-log.md), eligible types, within 24 hours | Same                                                                                   | Same                                                                                                                   |
| **Best for**                              | Learning what the AI suggests before it acts                                                                                                | Most accounts, and anyone who wants to see changes first                               | Experienced users who have set their bid caps and per-run limits                                                       |

> **Recommended path for new users:** Start on **Suggestions Only**, which is the default. Suggestions still arrive here for you to approve one at a time, and nothing is applied unless you approve it. Move to **Manual Approval** when you want to work the queue in batches, and review the [Activity Log](https://docs.sellermagnet.com/ppc-management-beta/ppc-manager-overview/ppc-activity-log.md) weekly. Consider **Fully Automatic** only once you trust its judgment.

{% hint style="danger" %}
**Before switching to Fully Automatic:** Check your AI strategy settings (target ACoS (Advertising Cost of Sale), bid caps, per-run limits). In this mode the worker applies bid and budget changes on its own, so incorrect settings are applied at scale. Destructive changes still wait in this queue, but bid and budget changes do not.
{% endhint %}

***

## Pending Actions

![SellerMagnet PPC Approval Center](https://4139654583-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fih2knSvVpBcq5rIeXVjM%2Fuploads%2Fgit-blob-d36088484d6041d51738b7bd9fa2028436d04d87%2Fppc-approval-center.png?alt=media)

![SellerMagnet PPC Approval Center - the Applied tab showing changes already pushed live](https://4139654583-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fih2knSvVpBcq5rIeXVjM%2Fuploads%2Fgit-blob-76f396146e379477080bf82d0df284a90c42e66b%2Fppc-approval-center-2.png?alt=media)

Each pending suggestion appears as one row in a grid:

### Row Layout

The columns are: a checkbox, an icon for the action type, **Action**, **Campaign**, **Change**, **Match Type**, **Confidence**, **Est. Impact**, **Created - Expires in**, and **Actions**.

* **Priority:** a color-coded left edge (Critical = red, High = orange, Medium = blue, Low = gray) and a dot with the priority label under the action title.
* **Action title:** the action type and its subject, e.g., "Bid Change - men's running shoes".
* **Tag line:** under the title you also see a source tag (AI Engine, AI Optimizer, Autopilot, Bid Rule, Manual or User), a **lane** label and a **risk chip**.
* **Campaign:** the marketplace flag, the Ads account name and the campaign name.
* **Change:** the old value (struck through) followed by the new value, with the percentage change. A decrease is shown in red.
* **Match Type:** the match type where it applies.
* **Confidence:** a bar with the percentage plus a HIGH, MED or LOW chip. Hover over the chip for "AI confidence based on {n} data points". The column tooltip explains: "Probability the new bid improves ROAS by ≥5% based on Thompson Sampling. Higher = more data + cleaner signal."
* **Est. Impact:** the estimated net profit change over the next 30 days if you apply the change. Positive means more profit, negative means less, so a cut that would lose more in sales than it saves in spend shows a negative number. A leading "\~" marks an estimate made without a product cost on file.
* **Created - Expires in:** when the suggestion was created and a colored badge with the time left: "in 3 d", "in 5 h", "under 1 h" or "expired". The badge changes color as expiry gets close.
* **Actions:** the row buttons described below.

Under each row, text lines explain the suggestion:

* **Reason:** why the AI recommends the change (e.g., "ACoS 45% is 1.8x target (25%)").
* **Evidence:** the data behind it, for example "{clicks} clicks, 0 orders - \~{expected\_orders} orders expected at your {cvr\_pct} conversion rate ({market})" or "No sellable offer for {asins}".
* **Profit basis:** how the suggestion was judged, for example "Profit target ACoS {target}% (break-even {be}%, keeps {margin}% margin)". Without a product cost on file you see "No product cost on file: judged against your target ACoS." and a link to add the cost.
* **Queued notice:** on a pending row your automation mode would otherwise send on its own, a line explains why it still waits for you. A bid cut bigger than your per-change bid adjustment limit reads "Queued for you: this cut is larger than your bid adjustment limit ({limit})". A cut on a keyword whose own ACoS already sits well inside (below 80% of) its campaign's break-even reads "Queued for you: this keyword's ACoS is already well inside its break-even, so the cut is not applied automatically". A bid cut held only because the keyword or target does not yet have enough clicks and orders to be sure of it (Fully Automatic mode) reads "Waiting for your OK: the autopilot makes a change like this on its own only when the keyword or target has enough clicks and orders to be sure, and this one does not yet", followed by a hint naming the setting: "In Fully Automatic mode you can let it make cuts like this one: PPC Settings, Automation, 'Let the autopilot apply bid cuts, even with little data'". When both reasons apply, you see the bid adjustment limit line. Any other held row reads the generic "Queued for you: evidence below the auto-apply bar".

**Lane labels:** Bid optimizer, Budget optimizer, Negatives, Keyword harvest, Stock guard, Keyword pauses, Campaign retirement, Campaign state, Placements, Other.

**Risk chips:**

| Chip                | Meaning                                                                                                                                                              |
| ------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Stops traffic**   | A keyword pause, a negative, or a campaign pause or archive. It can show "{amount} sales recorded on it" when the evidence recorded sales                            |
| **Can raise spend** | A bid or budget raise, a harvest, or turning a keyword or campaign on. A raise shows "up to {amount} more per day" (budget) or "up to {amount} more per click" (bid) |
| **Lowers spend**    | A bid or budget cut                                                                                                                                                  |
| **Structure**       | Placement changes and anything else                                                                                                                                  |

> **Reading confidence scores:** Treat confidence as a data quality indicator, not a "correctness" guarantee. A 95% confidence bid decrease means the AI has abundant, consistent data supporting the change. A 40% confidence suggestion means limited data, the recommendation could still be correct, but you should review the underlying metrics more carefully before approving.

### Row buttons

Every pending row has four buttons:

| Button                                                     | What it does                                                                                                |
| ---------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| **Approve** (green)                                        | Opens the pre-check dialog described below, then sends the change to Amazon once you confirm                |
| **Reject** (red)                                           | Asks you to confirm, then moves the suggestion to the Rejected tab. A 60-second **Undo** appears afterwards |
| Shield icon, tooltip "Reject and never suggest this again" | Opens the **Protect from automation** dialog (see below)                                                    |
| Hourglass icon, tooltip "Snooze 7 days"                    | Hides the suggestion for 7 days. In the snoozed view the button becomes **Unsnooze**                        |

The **Protect from automation** dialog says "The suggestion is rejected and automation will leave this item alone." Choose a scope - **All automation on this item** or **Only this kind of change** - optionally add a **Reason (optional)**, then click **Reject and protect** (or **Cancel**). Afterwards you see "Rejected. Automation will leave this item alone - manage it in Settings > Automation." This path has no Undo toast; you manage the protection in Settings.

Negative-keyword suggestions for seasonal terms can also carry a **Pause outside the season** button. It schedules the campaign to be paused outside the season dates and turned on inside them, and declines the negative so the term stays available next season.

### Editing a value before you approve

For pending **Bid Change** and **Budget Change** rows you can approve a different amount than the suggested one:

1. Click the pencil button (tooltip **Edit value**) in the Change cell.
2. Type a number above 0. Press `Enter` (or click away) to keep it, or `Esc` to cancel.
3. The row now shows an **Edited: {value}** badge and a **Use suggestion** button that resets it.
4. Approve as usual. The edited value goes through the pre-check, and the server keeps it within your limits - a value that had to be changed is marked **Adjusted to your limits**.

Editing is not available on other action types, on watch-list rows ("Collecting evidence"), on rows marked "Raise limit reached", or in the snoozed view. Edits are kept only while the page is open. If the server refuses an edited value, the toast reads "{n} not sent - the edited value is not valid, still pending".

### Understanding Priority Levels

| Priority     | Color  | When Assigned                                                                    |
| ------------ | ------ | -------------------------------------------------------------------------------- |
| **Critical** | Red    | Estimated impact above the Critical threshold **and** confidence of at least 85% |
| **High**     | Orange | Estimated impact above the High threshold **or** confidence of at least 70%      |
| **Medium**   | Blue   | Standard optimization: the remaining suggestions that rank above Low             |
| **Low**      | Gray   | Minor tweak: small expected improvement                                          |

Hover over the **Action** column header or the priority filter to see the exact thresholds, shown in your own currency: "Critical = >{critical} impact + ≥85% confidence. High = >{high} impact OR ≥70% confidence."

> **Triage strategy:** During your daily review, handle Critical and High priority items first. These have the biggest financial impact. Low priority items can be batch-reviewed weekly without significant consequences.

***

## Tabs

The queue has **five tabs**:

| Tab          | Badge | Description                                                                                                             |
| ------------ | ----- | ----------------------------------------------------------------------------------------------------------------------- |
| **Pending**  | Count | Suggestions waiting for your decision                                                                                   |
| **Applied**  | None  | Suggestions you (or Autopilot) accepted that were applied to Amazon                                                     |
| **Rejected** | None  | Suggestions you declined                                                                                                |
| **Expired**  | None  | Suggestions that passed their expiry date without a decision, or left the queue because they could no longer be applied |
| **All**      | None  | Every suggestion across all statuses                                                                                    |

Rows on the **Applied** tab carry a **Result** chip once the change can be judged: **Too early**, **Improved**, **No clear change**, **Got worse**, **Not measurable** or **Not control-adjusted**. The tooltip explains: "Whether the change worked: the 14 days after it compared with the 14 days before, measured once 31 days have passed."

> **Read the message after you approve.** After every approval a toast tells you how many changes were applied and, where relevant, how many were not sent, held, already in place, removed or expired - with the reason. Check it before you move on, and use the Result chip later to see whether an applied change worked.

***

## Where Did My Suggestion Go?

The **Pending** tab lists only the suggestions you can act on right now. A suggestion can leave it, or be hidden, for these reasons:

| What happened                                                                                                                                                                                               | Where the suggestion is now                             | How you are told                                                                                                                                |
| ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- |
| It passed its own expiry date, or it is older than 14 days                                                                                                                                                  | **Expired** tab                                         | "{n} suggestion(s) expired and left the queue - they were older than {days} days or past their expiry date."                                    |
| A **harvest** can no longer be submitted: the keyword already exists in the destination ad group, Amazon refuses the keyword text, the destination is an auto campaign, or there is no destination ad group | **Expired** tab                                         | "{n} suggestion(s) left the queue because they can no longer be submitted - the change already exists on Amazon, or Amazon does not accept it." |
| A **negate** is already in place, or a newer negative already covers it                                                                                                                                     | **Expired** tab                                         | The same message as above                                                                                                                       |
| An approval check **held** it until fresh reporting arrives                                                                                                                                                 | Still pending, but hidden until the new reporting lands | "{n} suggestion(s) are waiting for fresh reporting before they can be applied. They are hidden until then - nothing for you to do."             |
| You **snoozed** it                                                                                                                                                                                          | Behind the "{n} snoozed" bar under the toolbar          | Click **Show snoozed** to list them and **Back to the queue** to return. Snoozed suggestions are not counted in the Pending Review tile         |

These toasts appear only when you load or refresh the page yourself, never on the silent 30-second refresh. Because held suggestions are hidden but still pending, the **Pending Review** tile can show a higher number than the rows you see.

Some rows stay visible but cannot be approved. They have no checkbox and a disabled **Approve** button, and they are left out of every bulk path:

* **Collecting evidence:** "Not enough data yet to judge this suggestion with confidence. It stays visible and updates automatically as more traffic comes in - it cannot be approved until then."
* **Raise limit reached:** "This bid was already raised within the last {days} days. Raising it again now would exceed your {pct}% cumulative increase limit (ceiling {ceiling}). It can be approved again once the observation window has passed - a fresh suggestion replaces this one."
* **Currency mismatch:** "Currency mismatch: the evidence and this action use different currencies. Refresh the analysis before applying; no change will be sent."

***

## Filtering and Sorting the Queue

The toolbar above the queue lets you narrow and order what you see:

* **Search:** filter by campaign, keyword, or entity.
* **Priority filter:** show only a single priority - **All priorities**, **Critical**, **High**, **Medium**, or **Low**.
* **Action type filter:** limit to one type (**Bid Change**, **Budget Change**, **Campaign State Change**, **Negate Keyword**, **Harvest Keyword**, **Pause Keyword**, or **Placement Modifier**).
* **Sort:** the sort dropdown defaults to **Sort: Priority**, with **Sort: Confidence**, **Sort: Est. Impact**, **Sort: Newest**, and **Sort: Expiring soon** as alternatives.

Three icon buttons sit at the end of the toolbar: **Refresh**, **Auto** (tooltip "Auto-refresh every 30s", on by default) and a bookmark, **Save as view**, which saves the current search, priority, action type and sort so you can reuse them later.

### More filters & views

A collapsed section named **More filters & views** holds the quick filter chips. Each chip shows a live count, and the section header shows a badge with the number of active chip groups. **Clear advanced filters** resets them.

| Group          | Chips                                           |
| -------------- | ----------------------------------------------- |
| **Confidence** | All, ≥95%, ≥85%, ≥70%                           |
| **Impact/mo**  | All, > 100, 25-100, < 25                        |
| **Type**       | All, Bid, Budget, Pause, Negate, Harvest, State |
| **Source**     | All, AI, Autopilot, Rule                        |
| **Age**        | All, <1h, <24h, <72h, Older                     |

The chips combine with the toolbar filters, and the chip state is kept in the page address, so you can bookmark or share a filtered view.

The same section holds three more controls:

* **Views:** a menu with the quick views **Safe auto-batch** (confidence ≥85%, impact > €100/mo, older than 72h), **Urgent review** (confidence ≥70% and older than 72h) and **Cleanup tier** (impact < €25/mo), followed by **Recent saved** (your latest chip combinations, kept in this browser) and **Save current as view\...**.
* **Approve matching:** approves everything the current filters let through, not just a checkbox selection. It opens the pre-check dialog titled "Approve {n} matching actions?" If the batch includes rows below 80% confidence, it warns "Some of these are lower-confidence - review individually first?"
* **Approve all safe:** appears when a safe set exists. Its tooltip says: "Conservative actions only: non-converting negates and pauses, keyword harvests, and bid cuts. Every action is re-validated at apply time. Eligible reversals are available individually in Activity Log; some actions cannot be undone."

***

## Bulk Actions

When reviewing multiple suggestions:

1. **Select items** with the checkbox on each row, or tick the unlabeled checkbox in the column header to select every selectable row in the list.
2. Use the selection bar that appears above the queue (Pending tab only). It reads " selected" and offers four buttons: **Clear**, **Snooze 7d**, **Reject Selected** and **Approve Selected**.

{% hint style="success" %}
**Efficient bulk review workflow:**

{% stepper %}
{% step %}

### Priority

Sort by **Sort: Priority** (Critical first).
{% endstep %}

{% step %}

### Step 2

Review and individually approve/reject Critical and High items.
{% endstep %}

{% step %}

### Select All

For Medium and Low items, tick the header checkbox and then untick any you want to handle differently.
{% endstep %}

{% step %}

### Approve Selected

Click **Approve Selected**, read the pre-check dialog, and confirm.
{% endstep %}
{% endstepper %}
{% endhint %}

{% hint style="warning" icon="triangle-exclamation" %}
**Approving sends changes to Amazon, and there is no Undo for an approval.** Every approval - single, selected, **Approve matching** or **Approve all safe** - first opens a pre-check dialog (next section), and the change only goes live once you confirm. After that there is no Undo toast, not even for **Approve all safe**. An applied change can only be reversed from the [Activity Log](https://docs.sellermagnet.com/ppc-management-beta/ppc-manager-overview/ppc-activity-log.md), within 24 hours, and only for eligible action types. So check the list before you confirm.
{% endhint %}

***

## The Approve Pre-check

Before anything is sent, the dialog asks the server what Approve would do with each row. Its title reads "Apply {count} action(s) to your Amazon campaigns? This will modify real bids/budgets." (**Approve all safe** and **Approve matching** use their own titles). The dialog shows:

* **Summary lines:** "{n} expected to be sent", "{n} will not be sent (see Pre-check)", "{n} stop traffic (pause or negative)", "{n} cannot be undone (new keywords from harvesting)", and one "Daily budget change: {amount}" line per currency.
* **Breakdown by type:** a row of chips with the count per action type.
* **A table** with the columns **Action**, **Keyword / campaign**, **Change** and **Pre-check**. A ban icon marks actions that stop traffic and a lock icon marks actions that cannot be undone. The dialog lists the first 25 rows, then "+ {n} more".
* **A footnote:** "Checked only when you approve: {checks}. These checks and Amazon's own answer can still stop a change listed here." The checks named there are "search-term and bid evidence from your latest reports", "stock, Buy Box and FBA cover of the advertised products" and "current placement performance".

If the pre-check itself fails, the dialog says "The pre-check could not run - nothing was checked. Every check runs again when you approve." You can still confirm; the checks run at approval time.

Each row's **Pre-check** cell reads **OK**, **Adjusted to your limits**, or one of these reasons:

| Reason                                   | Meaning                                             |
| ---------------------------------------- | --------------------------------------------------- |
| No longer pending                        | The suggestion was already decided                  |
| Expired                                  | It passed its expiry date                           |
| On hold                                  | An earlier approval check holds it                  |
| Conflicts with another change            | Another change targets the same item                |
| Edited value not valid                   | Your edited value was refused                       |
| Still collecting evidence                | It is a watch-list row                              |
| Search term is converting                | The term to negate is now converting                |
| Budget already at or below the cut       | The campaign budget is already lower                |
| Budget changed since this suggestion     | The budget moved after the suggestion               |
| Outside your limits                      | The value falls outside your caps                   |
| Already in place - nothing to send       | Amazon already has this change                      |
| Replaced by a later negative             | A newer negative covers it                          |
| Preview mode - recorded, not sent        | Sandbox mode                                        |
| Would block one of your live keywords    | The negative would block a live keyword             |
| Keyword sync incomplete - will be held   | Keyword data is not complete yet                    |
| Bid changed since this suggestion        | The bid moved after the suggestion                  |
| Too many recent cuts - will be held      | Recent cuts on the item reached the limit           |
| Raise history unavailable - will be held | Recent raise history could not be read              |
| Over your 14-day raise limit             | The raise would exceed your cumulative limit        |
| Matches your brand terms                 | The term is one of your brand terms                 |
| Destination is an auto campaign          | A harvest cannot go to an auto campaign             |
| No destination ad group                  | A harvest has no ad group to go into                |
| Keyword text Amazon refuses              | Amazon does not accept the keyword text             |
| Incomplete suggestion                    | The suggestion is missing data                      |
| Not supported by Approve                 | The action type cannot be approved here             |
| Protected from automation                | The item is protected (see Protect from automation) |

***

## Undoing a Decision

| What you did                            | Can you undo it?                                                                                                                                                                                                             |
| --------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Rejected** a suggestion               | Yes, for 60 seconds: click **Undo** in the toast. It returns the rows to pending ("Returned {n} rejected actions to pending.")                                                                                               |
| **Snoozed** a suggestion                | Yes, for 60 seconds with **Undo** in the toast, or any time with **Unsnooze** on the row (click **Show snoozed** first)                                                                                                      |
| **Rejected and protected** a suggestion | No Undo toast. Manage the protection in Settings > Automation                                                                                                                                                                |
| **Approved** a suggestion               | No. There is no Undo after you approve                                                                                                                                                                                       |
| Change already **applied** to Amazon    | Only from the [Activity Log](https://docs.sellermagnet.com/ppc-management-beta/ppc-manager-overview/ppc-activity-log.md), within 24 hours, for eligible action types. Keyword harvests and some other types cannot be undone |

***

## Suggestion Types

The AI may recommend any of the following:

| Type                       | Description                                                                                                                                  | Typical Trigger                                                                                       |
| -------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- |
| **Bid increase**           | Raise bids on high-converting keywords                                                                                                       | Low ACoS and good conversion, with room to grow impressions                                           |
| **Bid decrease**           | Lower bids on high-ACoS keywords                                                                                                             | ACoS above target with enough click data                                                              |
| **Keyword pause**          | Pause keywords with consistently poor performance                                                                                            | No conversions after significant spend                                                                |
| **Keyword harvest**        | Add high-performing search terms as new keywords                                                                                             | A converting search term worth its own keyword                                                        |
| **Create manual campaign** | Create a new manual exact-match campaign from an automatic campaign's converting search terms, for a product that has no manual campaign yet | Several search terms in an AUTO campaign sold well and no manual campaign advertises that product yet |
| **Budget adjustment**      | Change a campaign's daily budget                                                                                                             | A campaign hitting its budget cap, or one that underspends                                            |
| **Negative keyword**       | Block wasteful search terms                                                                                                                  | Search term with clicks and no conversions                                                            |
| **Placement adjustment**   | Increase or decrease placement bid modifiers                                                                                                 | One placement converting clearly better or worse                                                      |
| **Campaign state change**  | Turn a campaign on or off                                                                                                                    | Retiring a campaign, or pausing one with no sellable offer                                            |

The wording of the **Reason** and **Evidence** lines in your queue shows the exact trigger for each suggestion.

### Use Case: How Each Suggestion Type Impacts Your Account

The examples below are illustrative. The numbers in your queue come from your own data.

#### Bid Increase Example

> **Suggestion:** Increase bid on "stainless steel water bottle" from EUR 0.85 to EUR 1.05 (+23%)
>
> **AI Reason:** "This keyword has a 14% CVR and 12% ACoS over the last 14 days, well below your 25% target. Current impression share is 18%, indicating significant room to grow."
>
> **What happens if you approve:** The keyword wins more auctions, gets more impressions, and (based on historical conversion rates) should generate additional profitable sales.
>
> **What happens if you reject:** Performance stays the same. You may miss out on profitable sales volume, but your current ACoS remains stable.

#### Negative Keyword Example

> **Suggestion:** Add "free water bottle" as a negative exact match in Campaign "Water Bottles - Broad"
>
> **AI Reason:** "This search term received 87 clicks (EUR 43.50 spend) with 0 orders over the last 30 days. Adding as negative prevents future wasted spend."
>
> **What happens if you approve:** The search term is blocked. Future searches for "free water bottle" will not trigger your ads. Estimated savings: EUR 43.50/month.
>
> **What happens if you reject:** The search term continues to receive clicks and spend with no conversions.

#### Keyword Harvest Example

> **Suggestion:** Add "insulated water bottle for gym" as an Exact match keyword in Campaign "Water Bottles - Exact"
>
> **AI Reason:** "This search term appeared 5 times in broad match, generating 4 orders with a 9.2% ACoS. Adding as exact match gives you bid control over this high-performing term."
>
> **What happens if you approve:** A new exact match keyword is created, allowing you to set a specific bid and track performance independently. A harvest cannot be undone afterwards, so check the destination campaign in the pre-check first.

#### Create Manual Campaign Example

> **Suggestion:** Card titled "New manual campaign 'Water Bottles - Insulated - Exact' with these search terms as exact keywords", listing each search term with its **Bid**, **Orders** and **ACoS** columns.
>
> **AI Reason:** "{n} search term(s) from the automatic campaign 'Water Bottles - Broad' sold {orders} times ({sales} in sales), and no manual campaign advertises this product yet. Create a manual exact-match campaign with them."
>
> **What happens if you approve:** A new manual, exact-match campaign is created with these search terms as its keywords, bid at the lowest of the terms' own bids, placement boost at 0%, with the bid strategy set to fixed bids (your bid is used as entered, not adjusted up or down by Amazon). The daily budget shown is the automatic campaign's own budget, and you can edit it before approving. The automatic campaign keeps running unchanged - nothing is negated in it at this step. Once the new keywords reach at least 100 impressions and 5 clicks, SellerMagnet suggests blocking those same terms in the automatic campaign as negative exact, so the two campaigns stop competing against each other.
>
> **What happens if you reject:** No campaign is created; the search terms keep converting inside the automatic campaign only, and the suggestion may come back in a later cycle.
>
> **Before you approve:** this suggestion is re-checked at the moment you approve it, not only when it was created - a campaign already created from an earlier approval of the same row, a manual campaign for the product that appeared since, a campaign name already in use, and your budget limits are all checked again. If one of them no longer holds, nothing is created and the row explains why instead (for example "A manual campaign for this product exists now, so no new one was created" or "A campaign named '{name}' already exists, so no new one was created"). Like a keyword harvest, creating the campaign **cannot be undone** from here, and it is never included in **Approve all safe** because it can raise your spend. The same stock guard checks apply as for a harvest.

***

## Priority & Expiration

* **Priority levels:** Critical, High, Medium, Low - based on estimated impact and confidence (see Understanding Priority Levels above).
* **Expiration:** a pending suggestion expires if you do not decide in time. The Pending Review tile describes this as 7 days. Each row shows its own countdown in the **Expires in** badge, and that badge is the time to trust. Anything older than 14 days is closed whatever its badge says.
* **What expiry does:** an expired suggestion is not rejected. It gets the status **Expired** and stays on the **Expired** tab. The app states: "Expired suggestions come back in the next cycle; after expiring twice in 30 days they are held for 14 days."

{% hint style="warning" %}
**Why suggestions expire:** PPC performance data changes constantly. A bid decrease suggested based on last week's ACoS data may no longer be relevant if this week's performance shifted. Expiration ensures you never accidentally apply an outdated recommendation. If a suggestion was genuinely needed, a fresh, updated version can come back in the next analysis cycle.
{% endhint %}

{% hint style="danger" %}
**Do not let Critical suggestions expire.** Critical priority means significant money is being wasted right now. If a Critical suggestion expires without action, the underlying problem (e.g., a keyword spending EUR 50/day with zero orders) continues unchecked. Review Critical items within 24 hours.
{% endhint %}

***

## Daily Review Workflow

Follow this workflow each day you review the queue:

**Triage**

1. Open the Approval Center and check the **Pending** tab badge count and the expiry banner.
2. Sort by priority (Critical first), or use **Review** in the expiry banner to sort by **Sort: Expiring soon**.
3. Approve or reject all Critical items right away: these are time-sensitive.

**High Priority Review** 4. Review High priority suggestions. Read the reason and Evidence line for each. 5. Approve those that align with your strategy. Reject any that conflict with upcoming promotions or strategy changes, or snooze them for 7 days.

**Bulk Process the Rest** 6. Select the remaining Medium and Low priority items with the header checkbox. 7. Quickly scan the list. Untick anything unusual. 8. Click **Approve Selected** and read the pre-check dialog before you confirm.

***

## Keyboard Shortcuts

Power users can work the queue without touching the mouse. A **Keyboard shortcuts** link in the page footer opens a shortcuts overlay (you can also press `?` to open it). The keys do not act while you are typing in a field. The available keys:

| Key     | Action                                                                                                                     |
| ------- | -------------------------------------------------------------------------------------------------------------------------- |
| `j`     | Move highlight down                                                                                                        |
| `k`     | Move highlight up                                                                                                          |
| `Enter` | Tick or untick the checkbox of the highlighted row (the overlay calls it "Open detail view", but there is no detail panel) |
| `a`     | Approve highlighted (opens the pre-check dialog)                                                                           |
| `x`     | Reject highlighted (asks you to confirm)                                                                                   |
| `s`     | Snooze highlighted for 7 days                                                                                              |
| `/`     | Focus the search box                                                                                                       |
| `?`     | Show the shortcuts overlay                                                                                                 |
| `Esc`   | Close the overlay; otherwise clear the selection, then clear the highlight                                                 |

{% hint style="success" %}
**Fast triage:** Use `j` / `k` to move down the sorted queue and `a` / `x` to approve or reject each highlighted item. Rows that cannot be approved ignore the `a` key.
{% endhint %}

***

## What Happens After You Reject or Snooze

* **Reject:** the suggestion moves to the Rejected tab and you get a 60-second **Undo**. Rejected negate, harvest, and pause keyword suggestions are not suggested again for 7 days; other rejected suggestions may reappear sooner.
* **Reject and protect:** use the shield button when you never want this suggestion again. Automation then leaves the item alone, either for this kind of change or for all automation on the item. Approving a protected item is refused with the pre-check reason "Protected from automation".
* **Snooze 7 days:** the suggestion disappears from the queue and comes back after the snooze, so it does not expire unseen. A suggestion that is too close to expiry cannot be snoozed: "{n} too close to expiry to snooze - approve or reject them now."

If you are unsure whether a rejection is a lasting preference or a one-time exception, use **Snooze** for the one-time case and **Reject and protect** for the lasting one.

***

<details>

<summary><strong>⚠️ Common Mistakes to Avoid</strong></summary>

{% hint style="danger" %}
**Mistake 1: Letting the Pending queue build up.** If you skip reviews for a week, suggestions start expiring and the savings or uplift they carried are lost for a while. **Fix:** Set a daily calendar reminder, and watch the expiry banner. If you cannot review regularly, consider Fully Automatic, but remember that pauses, negatives and campaign state changes still wait for you.
{% endhint %}

{% hint style="danger" %}
**Mistake 2: Approving everything without review.** While bulk actions save time, approving 100% of suggestions blindly defeats the purpose of Manual Approval. The AI is good but not perfect: it may suggest changes that conflict with your promotions, inventory changes, or strategy shifts. **Fix:** At minimum, individually review Critical and High priority items. Only bulk-approve Medium and Low, and read the pre-check dialog every time.
{% endhint %}

{% hint style="warning" %}

* **Mistake 3: Rejecting suggestions without understanding why.** A rejected negate, harvest or pause keyword suggestion is not offered again for 7 days. If you reject a valid suggestion by accident, use **Undo** within 60 seconds. **Fix:** Read the Reason and Evidence lines before rejecting. If the suggestion is sound but the timing is bad (e.g., you are about to run a promotion), **Snooze** it instead.
* **Mistake 4: Switching between modes frequently.** Changing from Manual Approval to Fully Automatic and back every few days makes it hard to tell which changes you approved and which the worker applied. **Fix:** Commit to one mode for a few weeks before evaluating whether to switch.
* **Mistake 5: Not checking whether applied changes worked.** An approved change can be applied and still not help. **Fix:** Look at the **Result** chip on the Applied tab, and reverse a change from the [Activity Log](https://docs.sellermagnet.com/ppc-management-beta/ppc-manager-overview/ppc-activity-log.md) within 24 hours if it was a mistake.
  {% endhint %}

</details>

## 🔧 Troubleshooting

### No pending suggestions appearing

| Check                                      | Solution                                                                                                                          |
| ------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------- |
| Did the suggestions leave the queue?       | See "Where Did My Suggestion Go?" above: check the Expired tab, the "{n} snoozed" bar and the toasts shown when you load the page |
| Is the AI Optimizer enabled and analysing? | Open the AI Optimizer to run an analysis on demand. New findings depend on arriving reports and your saved analysis schedule      |
| Is your Amazon API connection active?      | Verify the Amazon Advertising API connection in Settings                                                                          |

### Applied a suggestion but nothing changed in Amazon

| Check                                                              | Solution                                                                                                                                                                                    |
| ------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| What did the message after approving say?                          | It reports applied, failed, held, "already in place", "not sent" and removed rows. Rows that were held or not sent did not reach Amazon                                                     |
| Was the keyword/campaign modified between suggestion and approval? | If the item changed, the pre-check or the approval refuses the row (for example "Bid changed since this suggestion" or "Budget changed since this suggestion"). Wait for a fresh suggestion |
| Is the account in sandbox mode?                                    | In sandbox mode changes are recorded but not sent to Amazon. See the Sandbox strip                                                                                                          |

### Suggestions seem wrong or irrelevant

| Check                                           | Solution                                                                                                                                           |
| ----------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------- |
| Are your strategy settings correct?             | Verify target ACoS and bid limits in your PPC settings                                                                                             |
| Is your target ACoS far above your actual ACoS? | Then the Target ACoS line warns that increases are growth bets. Adjust the target if that is not what you want                                     |
| Is a product cost missing?                      | Without a product cost the row says "No product cost on file" and is judged against your target ACoS. Add the cost to get profit-based suggestions |
| Has your product catalog changed recently?      | New products, price changes, or inventory issues can make old data misleading                                                                      |

### Suggestions expired before I could review them

| Check                                 | Solution                                                                                                                                                                                           |
| ------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Are you reviewing frequently enough?  | Check the **Expires in** badge on each row and the expiry banner. Use **Sort: Expiring soon** to see what lapses first                                                                             |
| Can you extend the expiration window? | No, it is system-defined. Expired suggestions come back in the next cycle, and after expiring twice in 30 days they are held for 14 days. Use **Snooze** for suggestions you want to look at later |

***

## ❓ FAQ

<details>

<summary><strong>What happens to pending suggestions when I change the automation mode?</strong></summary>

Changing the mode does not clear the queue, so review the Pending tab before and after. Whatever your new mode does not apply on its own - in Fully Automatic mode that includes pauses, negatives, campaign state changes and capped actions - keeps waiting here for your approval. If you do not want some pending suggestions, reject or snooze them first.

</details>

<details>

<summary><strong>Can I approve a suggestion but modify the amount before applying it?</strong></summary>

Yes, for pending bid and budget changes. Click the pencil button (**Edit value**) in the Change cell, enter a number above 0, and press `Enter`. The row then shows **Edited: {value}** and you can undo the edit with **Use suggestion**. The server keeps the value within your limits and marks it **Adjusted to your limits** in the pre-check. You cannot edit other action types, watch-list rows, rows marked "Raise limit reached", or snoozed rows.

</details>

<details>

<summary><strong>How many suggestions does the AI generate per day?</strong></summary>

It depends on your campaigns, the reports arriving from Amazon and your saved analysis schedule. The **Pending Review** tile shows how many are waiting now, and the **Age** chips under **More filters & views** show how recent they are.

</details>

<details>

<summary><strong>Can I set up notifications for Critical priority suggestions?</strong></summary>

This page has no notification setting for priority levels. Check the Pending tab regularly, watch the expiry banner, and use the **Critical** priority filter or **Sort: Priority** to see the urgent items first.

</details>

<details>

<summary><strong>Does rejecting a suggestion permanently block that type of optimization?</strong></summary>

No. Rejected negate, harvest, and pause keyword suggestions are not suggested again for 7 days; other rejected suggestions may reappear sooner. To stop a suggestion for good, use the shield button (**Reject and protect**) and choose whether it covers all automation on the item or only this kind of change.

</details>

<details>

<summary><strong>Can multiple team members review suggestions?</strong></summary>

Yes. The page and its approve and reject actions need a login and access to the project; there is no separate approver role. Because every approval changes your live campaigns, agree within your team who reviews the queue.

</details>

***

## ➡️ What's Next?

{% content-ref url="/pages/kw3mb3OSwMEFnPlRPVag" %}
[PPC AI Optimizer](https://docs.sellermagnet.com/ppc-management-beta/ppc-manager-overview/ppc-ai-optimizer.md)
{% endcontent-ref %}

{% content-ref url="/pages/UR1mVi1OEupeXaCZA7zK" %}
[PPC Activity Log](https://docs.sellermagnet.com/ppc-management-beta/ppc-manager-overview/ppc-activity-log.md)
{% endcontent-ref %}
