> ## Documentation Index
> Fetch the complete documentation index at: https://docs.revbridge.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Campaigns

> Self-optimizing campaigns powered by Multi-Armed Bandit algorithms

AI Campaigns (Autonomous Broadcast) are RevBridge AI's flagship feature. They use **Multi-Armed Bandit (MAB) algorithms** to automatically optimize your marketing campaigns, finding the best combination of message, channel, and timing for each customer.

## How AI Campaigns Work

Traditional campaigns require you to manually:

1. Choose a template
2. Select a channel
3. Pick a send time
4. A/B test variations
5. Analyze results
6. Apply learnings

AI Campaigns automate this entire process:

1. You configure objectives, audience, and channels
2. The AI generates brand-aware content variations
3. The system tests combinations automatically
4. Traffic shifts to better-performing variations
5. Performance continuously improves

<Info>
  The Multi-Armed Bandit algorithm balances **exploration** (testing new variations) with **exploitation** (using proven winners), ensuring your campaigns improve without sacrificing too much performance during testing.
</Info>

## Creating an AI Campaign

Navigate to **Engagement → AI Campaigns** and click **Create Campaign**. The full-screen campaign wizard guides you through 6 steps.

### Step 1: Campaign Type

Choose your campaign type:

| Type                     | Description                                                                                                          | Status      |
| ------------------------ | -------------------------------------------------------------------------------------------------------------------- | ----------- |
| **Performance Campaign** | AI-driven continuous optimization using MAB algorithms. Tests and improves templates, channels, and timing over time | Available   |
| **Broadcast Campaign**   | Single message send to a targeted audience on a specific channel                                                     | Coming Soon |
| **Shopping Campaign**    | AI-optimized product promotions from your catalog                                                                    | Coming Soon |

### Step 2: Objective

Configure the campaign's purpose:

| Field                | Description                                                                                                                                  |
| -------------------- | -------------------------------------------------------------------------------------------------------------------------------------------- |
| **Name**             | Campaign identifier (e.g., "Cart Abandonment Recovery")                                                                                      |
| **Briefing**         | Detailed instructions to guide AI content generation — this is sent to the AI as the primary description                                     |
| **Objective**        | Choose from: Increase Conversions, Boost Engagement, Customer Retention, Reactivate Dormant Users, Upsell & Cross-sell, Improve Satisfaction |
| **Conversion Event** | Select which event defines a conversion for this campaign                                                                                    |

<Tip>
  Use the Briefing field to give the AI specific instructions about tone, offers, or messaging strategy. The more detail you provide, the better the generated content will match your intent.
</Tip>

### Step 3: Audience

Select which customers to target:

1. Choose an existing audience segment from the dropdown
2. Or select a pre-built audience template
3. Review the audience size

<Tip>
  Start with broader audiences. The MAB algorithm personalizes within the audience, so you don't need to create micro-segments.
</Tip>

### Step 4: Channels

**Select channels** to use for the campaign:

| Channel  | Status      |
| -------- | ----------- |
| Email    | Available   |
| SMS      | Available   |
| WhatsApp | Available   |
| Push     | Coming Soon |
| RCS      | Coming Soon |
| Voice    | Coming Soon |

You can select multiple channels. The AI will test and optimize channel selection per customer.

**Select your content strategy** for the campaign:

* **AI Generated**: Let the AI create content variations based on your Brand Dossier and briefing
* **Template Based**: Select existing templates to use in the campaign

When using AI generation, the system creates 6 variations per channel using the Creative Concept Framework:

| Concept   | Strategy            | Best For                  |
| --------- | ------------------- | ------------------------- |
| **CC-01** | Financial Rational  | Discount-driven campaigns |
| **CC-02** | Financial Emotional | Flash sales, urgency      |
| **CC-03** | Social Rational     | Reviews, testimonials     |
| **CC-04** | Social Emotional    | Community, exclusivity    |
| **CC-05** | Personal Rational   | Recommendations           |
| **CC-06** | Personal Emotional  | Storytelling, brand       |

<Warning>
  Make sure each selected channel is properly configured in **Settings → Channels** before launching.
</Warning>

### Step 5: Budget

| Setting                | Description                                                                 |
| ---------------------- | --------------------------------------------------------------------------- |
| **Total Budget**       | Overall campaign spend limit                                                |
| **Daily Spend Limit**  | Maximum daily spend                                                         |
| **Execution Schedule** | Configure when the campaign runs: Fixed Interval, Daily Run, or Distributed |
| **Execution Limits**   | Control how many times the campaign runs before auto-stopping               |

**Configure execution scheduling** with one of three modes:

| Mode               | Description                         | Configuration                                                                 |
| ------------------ | ----------------------------------- | ----------------------------------------------------------------------------- |
| **Fixed Interval** | Run at regular intervals            | Select interval: 1min, 5min, 10min, 15min, 30min, 1h, 2h, 3h, 6h, 12h, or 24h |
| **Daily Run**      | Run once per day at a specific time | Set hour and minute (24-hour format)                                          |
| **Distributed**    | Spread runs across a time period    | Total number of runs + number of days                                         |

**Execution limits** control how many times the campaign runs before auto-stopping:

| Preset    | Runs | Use Case            |
| --------- | ---- | ------------------- |
| Unlimited | ∞    | Always-on campaigns |
| 10x       | 10   | Test runs           |
| 30x       | 30   | Monthly campaigns   |
| 90x       | 90   | Quarterly campaigns |
| Custom    | Any  | Set your own limit  |

When a limit is set, the system shows a preview of the expected campaign duration.

### Step 6: Review & Launch

Review all campaign configuration on a single summary page. From here you can:

* **Launch Campaign**: Start the campaign immediately
* **Save as Draft**: Save without launching for later editing

<Info>
  A completed **Brand Dossier** is required to launch any campaign. If you haven't set up your brand yet, go to **Assets → Branding** first.
</Info>

## Campaign Editing

You can edit campaigns that are in **Draft** or **Active** status.

To edit a campaign:

1. Click the pencil icon or select **Edit Campaign** from the action menu
2. The campaign wizard reopens with all existing settings pre-loaded
3. Make your changes across any step
4. Click **Launch Campaign** to save and launch, or **Save as Draft** to keep as draft

<Warning>
  Campaigns that are already running, paused, completed, or failed cannot be edited. To modify a running campaign, you must restart it.
</Warning>

## Campaign Restart

To restart a campaign that has already been dispatched:

1. Open the action menu (three dots) on any campaign row
2. Click **Restart Campaign**
3. The campaign is restarted on the backend with a fresh execution cycle

The restart preserves the campaign configuration but resets the execution count. A toast notification confirms the restart with the new execution details.

## Campaign Management

### Campaign States

| State         | Description                                 |
| ------------- | ------------------------------------------- |
| **Draft**     | Campaign is being configured, can be edited |
| **Active**    | Campaign is configured and ready to launch  |
| **Running**   | Campaign is dispatched and actively sending |
| **Paused**    | Campaign is temporarily stopped             |
| **Completed** | Campaign has finished                       |
| **Failed**    | Campaign encountered an error               |

### Pausing and Resuming

To pause an active campaign:

1. Click on the campaign
2. Click **Pause**
3. Confirm

The campaign stops sending but retains all learning. Resume anytime to continue where you left off.

### Monitoring Performance

View real-time metrics on the campaign detail page:

| Metric          | Description              |
| --------------- | ------------------------ |
| **Reach**       | Customers targeted       |
| **Sent**        | Messages dispatched      |
| **Delivered**   | Successful deliveries    |
| **Opens**       | Unique opens             |
| **Clicks**      | Unique clicks            |
| **Conversions** | Goal completions         |
| **Revenue**     | Attributed revenue       |
| **CPA**         | Cost per acquisition     |
| **Lift**        | Improvement vs. baseline |

### Template Performance

See how each template variation is performing:

* Open rate
* Click rate
* Conversion rate
* Revenue per send

The MAB algorithm automatically shifts traffic toward better performers while continuing to test others.

### Channel Performance

Compare results across channels:

* Which channel has highest engagement
* Which channel drives most conversions
* Channel-specific delivery rates

## Optimization Details

### How the MAB Algorithm Decides

For each customer interaction, the algorithm considers:

1. **Template Performance**: Historical conversion rates
2. **Channel Effectiveness**: Customer's engagement history per channel
3. **Send Time**: Optimal time based on past behavior
4. **Exploration Needs**: Whether more testing data is needed

The default exploration rate is 15%, meaning 85% of sends go to proven winners while 15% continue testing other variations.

### Convergence

Over time, the algorithm **converges** on optimal settings:

* Early: Heavy exploration, traffic split across variations
* Mid: Winning variations receive more traffic
* Late: Best performers dominate, minimal exploration

<Info>
  You can see the optimization score on the campaign detail page. Higher scores mean the algorithm is more confident in its selections.
</Info>

## Best Practices

### Template Diversity

Provide diverse templates that test different approaches:

* Include emotional AND rational appeals
* Vary subject lines significantly
* Test different CTAs
* Don't use nearly identical variations

### Audience Size

Larger audiences provide more data for optimization:

* **Minimum recommended**: 1,000 customers
* **Optimal**: 10,000+ customers
* Smaller audiences can work but optimize more slowly

### Runtime

Give campaigns time to optimize:

* **Minimum**: 7 days
* **Recommended**: 30 days
* Longer runtime = better optimization

### Budget Allocation

Set budgets that allow for adequate testing:

* Too low: Not enough sends to learn
* Too high: May waste budget on underperformers early

### Frequency Caps

Respect customer preferences:

* Email: 2-3 per week maximum
* SMS: 1 per week maximum
* Consider overall cross-channel frequency

## Advanced Features

### Campaign Priority

When a customer qualifies for multiple campaigns:

* Higher priority campaigns send first
* Frequency caps are respected across campaigns
* Lower priority campaigns may be suppressed

### Control Groups

AI Campaigns can automatically hold out a control group:

* Measures true lift from your campaign
* Control group receives no messages
* Compare conversion rates to prove ROI

### A/B Testing vs. MAB

| Feature          | Traditional A/B     | MAB (AI Campaign)              |
| ---------------- | ------------------- | ------------------------------ |
| Traffic Split    | 50/50 fixed         | Dynamic, adapts to performance |
| Winner Selection | Manual, after test  | Automatic, continuous          |
| Optimization     | Post-test           | Real-time                      |
| Sample Size      | Predetermined       | Adaptive                       |
| Regret           | High (50% to loser) | Minimized                      |