Title: Targeting and Allocation
Locale: en
URL: https://sensorswave.com/en/docs/experiments/targeting-and-allocation/
Description: Deep dive into split mechanisms and strategy selection
The split strategy determines how users are assigned to different experiment Variants — it is the core mechanism of A/B experiments. This article provides a deep dive into split ID selection, split algorithm principles, user property-based targeted split, and Allocation best practices.

## Split ID Selection

### Login ID

**Definition**: A unique identifier for the user after logging in, typically the user's primary key in the business system.

**Use cases**:
- E-commerce platforms: User ID, member number
- SaaS products: Account ID, company ID
- Social products: User ID, username

**Advantages**:
- **Cross-device consistency**: The user sees the same experiment Variant on phone, computer, and tablet
- **Long-term stability**: User sees the same Variant after logging out and back in
- **Better user experience**: Avoids user confusion

**Disadvantages**:
- **Limited coverage**: Can only experiment on logged-in users
- **Possible inconsistency before and after login**: User may see different Variants before and after login

**Recommended for**:
- Products requiring user login
- Important experiments (e.g., checkout flow, pricing strategy)
- Scenarios requiring high cross-device experience consistency

See [Data Model](../data-integration/data-model.mdx) for details.

### Anonymous ID

**Definition**: A unique ID identifying a user's device or browser, automatically generated by the SDK on first visit.

**Storage methods**:
- **Web**: Stored in cookies (`localStorage` or `sessionStorage`)
- **iOS**: Stored in KeyChain or UUID generated
- **Android**: Uses Android ID or UUID generated

**Advantages**:
- **Wide coverage**: Includes non-logged-in users
- **Auto-managed**: SDK automatically generates and maintains
- **No user action required**: No login needed

**Disadvantages**:
- **Device-level**: Same user may see different Variants on different devices
- **Easily lost**: New Anonymous ID generated after user clears cookies or app data

**Applicable scenarios**:
- Public websites or content platforms
- Landing page testing
- UX optimization for non-logged-in users

### Dual ID Parallel Mechanism

Users have two types of IDs simultaneously. Each ID independently matches experiments of its corresponding type without interference:

- **Anonymous ID**: Exists from the user's first visit onward, automatically generated and maintained by the SDK
- **Login ID**: Exists only after the user logs in, provided by the business system

```
User visits (not logged in)
  → Has Anonymous ID (anon_12345)
  → Can match all experiments using Anonymous ID as the split ID

User logs in
  → Has both Anonymous ID (anon_12345) and Login ID (user_67890)
  → Anonymous ID experiments: continue using anon_12345 for split, variant unchanged
  → Login ID experiments: use user_67890 for split, match corresponding variant
```

**Key points**:

- When creating an experiment, choose the split ID type (Anonymous ID or Login ID)
- Anonymous ID experiments: all users (including non-logged-in) can match, variant stays consistent before and after login
- Login ID experiments: only logged-in users can match, ensuring cross-device experience consistency
- The two ID types split independently; the same user can participate in both types of experiments simultaneously

---

## Split Algorithm Details

### Hash-Based Split Algorithm

Sensors Wave uses a stable hash algorithm:

```
Variant index = hash(user_id + experiment_key) % 100
```

**Steps**:

1. **Concatenate strings**: Combine user ID and experiment key
   ```
   input = "user_12345" + "cart_button_color_test"
   ```

2. **Calculate hash value**: Use a hash function (e.g., MD5, SHA-256)
   ```
   hash_value = hash("user_12345cart_button_color_test")
   ```

3. **Modulo operation**: Take modulo 100, resulting in a number from 0–99
   ```
   bucket = hash_value % 100
   ```

4. **Assign Variant**: Based on Allocation, determine the user's Variant
   ```
   if (bucket  5%), check the configuration

### Q: Will the same user participating in multiple experiments cause conflicts?

**A**: No. Split across different experiments is independent. The same user can participate in multiple experiments simultaneously and be assigned to different Variants in each.

### Q: Can the Allocation be modified while the experiment is running?

**A**: Not recommended. Modifying Allocation disrupts sticky assignment and affects experiment results. If changes are necessary, stop the current experiment and create a new one.

---

## Related Documentation

- [Core Concepts](core-concepts.mdx): Understand split mechanism principles
- [Experiment Design](experiment-design.mdx): Choose the right split strategy
- [Data Model](../data-integration/data-model.mdx): Understand Login ID and Anonymous ID

---

**Last updated**: January 29, 2026