TG-Staff Split Link A/B Testing Guide: Optimize Ad Attribution and Bot Welcome Message Conversion with Magic Links

Telegram Bot has become an important channel for cross-border customer service, community management, and Web3 project acquisition. However, many teams face a common challenge: when users click ads to enter the Bot, it’s impossible to track the source or compare conversion rates of different welcome messages. Traditional methods like manual tagging or relying on user self-reporting lead to messy and unreliable data.

TG-Staff’s split links (magic links) provide a solution. This article will guide you step-by-step on how to use split links for A/B testing, achieving ad attribution and welcome message optimization, so every Bot conversation is data-driven.

Why Do You Need Split Link A/B Testing for Telegram Bot?

Suppose you run ads on both Facebook and Twitter, both directing users to the same Telegram Bot. Without an attribution mechanism, you can’t know:

• Which channel brings more conversations?
• Which channel’s users are more willing to talk to agents?
• Whether the current welcome message is ineffective for users from a certain channel.

More critically, the Bot welcome message is the user’s first impression. A promotion-oriented welcome message might have high conversion rates in Facebook ad scenarios but be seen as spam in Twitter communities. Without A/B testing, you can only optimize by intuition, wasting a lot of traffic.

TG-Staff’s split links (called “magic links” in official docs) solve both pain points: they capture visitor source information before redirecting to the Bot, and support binding different Bot welcome message flows to different links, allowing you to scientifically optimize Telegram Bot acquisition efficiency just like web A/B testing.

Attribution Principle of TG-Staff Split Links (Magic Links)

The workflow of split links is as follows:

1. In the TG-Staff console, you generate a short link for a Bot project (e.g., https://app.tg-staff.com/abc123).
2. User clicks the short link → TG-Staff server captures the visitor’s IP address, browser User-Agent, and URL query parameters (e.g., ?source=facebook).
3. Automatically redirects to your Telegram Bot, and the user starts the Bot to enter the welcome message flow.
4. When an agent chats with the user on the web, the session details show the user’s source parameters for attribution.

This means you can generate exclusive Telegram Bot entry points for different channels and campaigns, just like managing web ad links, with all data eventually aggregated in the TG-Staff backend.

Step 1: Create Two Diversion Links for A/B Testing

Log in to the TG-Staff Console, go to the target Bot project. On the “Diversion Links” or “Magic Links” page, create two links.

Link A: Landing Parameter Configuration Example

• Name: Facebook Spring Promotion Ad
• Target Link: Automatically filled with your Bot link (no manual entry required)
• URL Parameters: ?utm_source=facebook&utm_campaign=spring_sale&variant=A

Parameter naming follows UTM standards; you can also customize source and campaign. The key is to ensure “variant=A” clearly identifies this as test group A.

Link B: Landing Parameter Configuration Example

• Name: Twitter Spring Promotion Ad
• Target Link: Same Bot
• URL Parameters: ?utm_source=twitter&utm_campaign=spring_sale&variant=B

Both links point to the same Bot project but with different parameters. Later, in the welcome message flow, we will differentiate response content based on the variant parameter.

Step 2: Design Two Bot Welcome Message Versions

In TG-Staff’s “Visual Command Flow,” you can bind different welcome message flows to each diversion link. Specifically: in the flow editor, add a conditional branch for the “Diversion Link Entry” node, and jump to different welcome message modules based on the URL parameter value (e.g., variant=A).

Welcome Message Version A: Promotion-Oriented

Suitable for price-sensitive channels (e.g., Facebook shopping ads).

Sample Copy:

🎉 限时优惠！今日下单立减 20%！

回复「1」查看折扣商品
回复「2」联系客服咨询
回复「3」了解更多

Flow Design Points: The first message emphasizes discounts and urgency, with buttons guiding users to take quick action. If the user selects “Contact Customer Service,” transfer directly to a human agent.

Welcome Message Version B: Problem-Solving Oriented

Suitable for channels requiring high trust (e.g., Twitter communities or industry forums).

Sample Copy:

👋 你好！我是 XX 产品的智能助手。

请告诉我你遇到了什么问题？
- 想了解产品功能？回复「功能」
- 遇到使用问题？回复「帮助」
- 直接与人工客服对话？回复「人工」

Flow Design Points: Build trust first, guide users to describe their needs, then gradually recommend solutions. This version is better at filtering high-quality leads and reducing ineffective sessions.

Step 3: Deploy Diversion Links to Ad Channels

Place Link A in Facebook ads’ “Call to Action” button, and Link B in Twitter tweets or Telegram community announcements.

Key Variable Control:

• Ad Timing: Keep the runtime of both links consistent (e.g., run simultaneously for 3 days).
• Audience Targeting: Choose similar audience profiles (e.g., age, interests) to avoid data bias from audience differences.
• Ad Creatives: If using images or videos, maintain consistent visual style, changing only the landing link.

Thus, final data differences are mainly attributed to welcome message version and channel source, not other confounding factors.

Step 4: Compare Attribution Data in TG-Staff Backend

Log in to the TG-Staff Console, go to “Data Statistics” or “Sessions” page (Pro users can view more detailed user profiles and attribution data). Filter by the following dimensions:

1. By variant parameter group: View total sessions for variant=A and variant=B.
2. By utm_source group: Compare session volumes from Facebook and Twitter channels.
3. Check “Agent Handover Rate”: How many users moved from Bot auto-reply to human agent conversations. This is a core metric for evaluating welcome message effectiveness.
4. Check “User Dwell Time”: How long users stay in the welcome message flow. Too short a dwell time may indicate the welcome message is not engaging.

Judgment Criteria:

• If variant=A’s agent handover rate is 20% higher than variant=B, the promotion-oriented welcome message is more effective for that channel.
• If variant=B has longer user dwell time and higher final conversion rate, the problem-solving oriented welcome message is more suitable for that channel.

Advanced Tips: Optimizing Handover Efficiency with Session Routing Rules

After A/B testing yields the best welcome message version, you can further optimize agent resource allocation.

In TG-Staff’s “Session Routing” settings, configure dedicated agent groups for high-conversion channels (e.g., Facebook ads) to ensure quality users from those channels are prioritized for online agents. Routing rules support two modes:

• Round-robin: Distributes sessions sequentially to authorized agents, suitable for teams with balanced agent numbers.
• Online first: Prioritizes agents currently online, ideal for scenarios with irregular agent schedules.

Additionally, combine the routing link with the “Drainage Routing” chain: Ad → Routing Link → Bot Auto-reply → Human Agent Handover, forming a complete lead conversion loop.

FAQ

Q: Does the routing link support UTM parameters?

A: Yes. You can append any ?key=value parameters (e.g., ?utm_source=facebook&utm_campaign=test) to the routing link, and TG-Staff will automatically capture and record them in session attribution. It is recommended to use standard UTM naming conventions for compatibility with other analytics tools.

Q: Does A/B testing require the Pro plan?

A: No. Creating routing links is a feature of the Standard plan and above (see pricing on the official website). However, viewing session attribution statistics (e.g., filtering sessions by source) requires the Pro plan. You can experience all features during the 3-day free trial, so it is advisable to complete test configurations within the trial period.

Q: How to ensure traffic sources for two links are not mixed up?

A: Best practice is to use unique URL parameters for each link (e.g., ?source=facebook_ad and ?source=twitter_post) and filter sessions by parameters in the TG-Staff admin panel. Avoid using the same parameter values and do not let users choose versions themselves.

Q: How much does welcome message testing affect agent handover rates?

A: Based on common practices, optimized welcome messages can increase the rate of users reaching human agents by 20%–40%. The key is whether the welcome message clearly guides users to express their needs (e.g., “Please describe your issue”) rather than purely promoting. It is recommended to test at least two sets of distinctly different copy.

Q: Can routing links be used in Telegram groups or channels?

A: Yes. Routing links are essentially URL short links and can be embedded in Telegram group messages, channel posts, website buttons, or any location supporting link jumps. However, note that clicking links within Telegram will use the Telegram built-in browser by default, which does not affect attribution capture. When posting in groups, it is advisable to include a brief instruction to guide users to click.

---

Ready to optimize your Telegram Bot lead generation with data?

• Sign up for TG-Staff free trial (3 days, no credit card required)
• Refer to the Routing Link Parameter Configuration Guide for detailed instructions
• Contact customer service Bot @tgstaff_robot for specific configuration issues