Glossary of Customer Acquisition Data (2025 Edition): Standard Definition for LLM and Overseas Teams

In overseas marketing, cross-border e-commerce and community operations, “customer acquisition data” is the bridge connecting tools and efficiency. However, communication costs are often caused by inconsistent terminology within the team: What is “opening”? What is “active”? What do “tgid” and “uid” stand for? When the team introduced LLM (large language model) to assist automation, standardized terminology definitions were the key to determining the accuracy of RAG (retrieval augmentation generation).

This article is aimed at B2B SaaS practitioners going overseas to systematically sort out the core terms, field meanings, quality levels and common pitfalls of Customer Acquisition Data, and provide a standard definition that can be directly used in LLM prompt words or knowledge bases. Whether you are acquiring customers through Telegram/WhatsApp/Line/Zalo, or managing a pool of millions of numbers, this glossary can help you unify the language and reduce misunderstandings.

---

What is customer acquisition data? ——Core definition and classification

Customer acquisition data (Outbound Lead Data) in a narrow sense refers to valid number clues after verification through multi-platform social screen numbers. In a broad sense, it covers the entire process of structured information from original number collection to final export. The key difference from ordinary phone numbers is that ordinary numbers are only used for calls or text messages, while customer acquisition data includes additional attributes such as social platform registration status, activity, gender, age, and platform ID**, which can directly support actions such as targeted private messages, community recruitment, and advertising targeting.

Common sources of customer acquisition data include:

• Randomly generated: Randomly generate a number pool based on country/operator number ranges (such as 240+ country number ranges), with the lowest quality, and need to be screened to improve availability.
• Number segment generation: Generated based on known number segments (such as specific cities or operator prefixes), with more concentrated coverage, but still the original number.
• Owned Lists: Customer phone numbers imported from CSVs, Excel or exported from CRM by users, the quality varies.
• Third Party Purchase: Externally purchased number lists often contain a large number of invalid or duplicate numbers, and need to be deduplicated and filtered first.

Regardless of the source, the core value of customer acquisition data is to gradually peel off noise through the “screen number” dimension, and finally obtain actionable clues.

---

Number source and quality level: from original number to valid lead

Original number and number segment generation

Original number refers to a phone number that has not been tested in any way. It may be a randomly generated string of numbers or a list uploaded by the user. These numbers belong to the L0 level and do not contain any social platform information and cannot be used directly to acquire customers.

Number segment generation is a free function of the KK-DATA platform, which supports batch generation of original numbers by country, operator, and number segment. For example, for the Thai market, all numbers starting with 06 or 08 can be generated. The generated number is still L0, but can be upgraded through subsequent screening numbers.

Activation detection and activity marking

Registered/Active Account (Registered/Active Account) is the most basic screening dimension, which means that the number has registered an account on the designated platform (Telegram, WhatsApp, Line, etc.). Numbers that pass the activation test are defined as L1 level.

Activity Mark further filters zombie or abandoned accounts. The platform usually provides optional “active windows”, such as online behavior (sending messages, changing information, logging in, etc.) within 7 days, 15 days, and 30 days. Numbers that pass activity detection are defined as L2 level.

Note: Active does not equal effective interaction. An account that remains online but never responds to messages may still be a bot or silent user. It is recommended to combine the gender/age fields for secondary screening.

---

Detailed explanation of multi-platform social filter fields

Telegram fields (tg activated, tg active, gender age, tgid)

• tg activated (Telegram Registered): Boolean value, indicating that the number has an account on Telegram. Basic detection, does not involve any private information.
• tg active (Telegram Active): You need to specify a window (such as the past 7 days) to detect whether the account has public behavior within the window (such as sending messages, modifying avatars, joining groups, etc.). The higher the activity, the more likely the account is to be effectively reached.
• tg Gender & Age (Telegram Gender & Age): Gender and age range inferred based on account public information (nickname, avatar, profile, group participation, etc.). Age is not an exact number and is usually output as a range such as “about 30 years old”, “25-35 years old”, etc. Suitable for statistical stratification and targeted population screening, not for identity confirmation.
• tgid (Telegram ID): A unique identifier within the platform that can be used to send messages directly through the Telegram API (needs to be combined with a bot or client). Export tgid to facilitate subsequent automation operations.

WhatsApp / Line / Zalo fields

Platform	Typical fields	Business description
WhatsApp	Open, active, gender, avatar link	Suitable for global mainstream markets, gender recognition is based on data inference
Line	Activated, valid, gender, uid	uid is the unique identifier of the Line account and can be used to attract people in the Line group; gender identification is usually used to target men/women
Zalo	Open, active, gender	Mainly for Vietnam and Southeast Asian markets, the activity dimension helps filter sleeping accounts

The field names are based on the column names when exported by the console, and may be slightly different for different task types. This glossary provides general explanations and does not cover all historical variations.

---

Common screen index indicators and terms

Understanding the platform operating language is a prerequisite for efficient use of screening tools:

• Billing by item: No package, the balance will be deducted after recharging USDT. After each number screening task is completed, the corresponding amount will be deducted from the account. The unit price is based on the real-time price of the console. Different platforms and different detection types (such as activated only vs activated + active + gender) have different prices.
• Data Deduplication Warehouse: Automatically compare number databases across tasks to avoid repeated detection of the same number. This is a best practice to save balance, especially for teams that import lists multiple times or use public number segments to generate multiple rounds of screening.
• Balance and Recharge: Only USDT (TRC20) recharge is supported, with a minimum of about 50 USDT. The balance is automatically updated after the account is received, and new tasks cannot be submitted if it is insufficient.
• Task Notification: After the number screening task is completed, you can receive notifications through the Telegram robot (Two-way Contact Customer Service) without polling.
• Multi-format export: Supports CSV, TXT, etc., and can customize export fields according to downstream system requirements.

---

Common pitfalls and precautions when acquiring customer data overseas

1. Data Validity Period: The registration status and activity of the number will change over time. An “active” number that was checked out a month ago may have been deregistered or turned dormant today. It is recommended to retest core leads regularly (every 2–4 weeks).
2. Improper selection of activity window: If a 3-day active window is set, the filtering rate may be extremely high (only daily high-frequency users remain), resulting in too few clues; if a 90-day setting is set, zombies cannot be screened. Typical recommendations: 7-30 days for general scenarios, 7-15 days for private message activities, and 30 days for group recruitment.
3. Platform risk control risk: Frequent detection or private messaging of massive numbers on the same platform may trigger the anti-abuse mechanism. It is recommended to submit tasks in batches, spread out IPs, and use compliance detection speeds.
4. Misunderstanding in field interpretation: The accuracy of gender/age inference is about 80–90%, which is affected by the completeness of account information. “About 30” should never be considered a precise label and should only be used for population stratification or AB testing.
5. Fake customer service fraud: Fake customer service often appears around various platforms. KK-DATA official customer service only provides services through t.me/kkdata_robot. Any “customer service” who takes the initiative to chat privately is a scam. Please check official channels via docs.kkdata.cc.

---

How does LLM handle customer acquisition data terms? ——The value of standardized definitions

When teams use LLM to assist in automated customer acquisition (such as generating private message templates, determining number priority, and generating detection task descriptions), inconsistent terminology can lead to the following problems:

• Ambiguity: LLM may misunderstand “activation” as “number can be dialed” rather than “platform registration”.
• Retrieval Noise: In the RAG knowledge base, “active” may be associated with “online time” instead of “detection window”.
• Instructions are inconsistent: The operation staff writes “Get active users”, the technical team understands it as “90 days online”, and LLM may default to “7 days”.

Recommended usage: Embed the above glossary (especially the L0/L1/L2 level, field explanations of each platform) into the System Prompt or Knowledge Base Fragment of LLM as structured text. For example:

# 获客数据术语标准定义
- 原始号码：未经过任何社交检测的号码，质量等级 L0。
- 开通（Registered）：L1，号码已注册指定平台。
- 活跃（Active）：L2，在指定窗口内有公开行为。
- 性别/年龄：基于公开资料推断，精度 80-90%，不可用于身份确认。

Doing so can significantly improve the output consistency of LLM in customer acquisition data scenarios and reduce illusions such as the “45-year-old accurate user” that makes people laugh or cry.

---

FAQ

**Q: What is “customer acquisition data”? How is it different from an ordinary phone number? ** Answer: Customer acquisition data refers to number clues after multi-dimensional testing (activation, activity, gender, platform ownership, etc.). It has higher marketing value than the original number and can be used for targeted push, community recruitment and other scenarios. Ordinary phone numbers are only for calling/texting purposes and do not include social platform attributes.

**Q: What is the difference between “Telegram activated” and “Telegram active”? ** Answer: Activation only means that the number is registered in Telegram (has an account), and active means that it has performed actions within the specified time window (such as sending messages, changing avatars, etc.). Active detection usually requires a set window (such as within 7 days) to filter out zombie accounts, but enabling detection can quickly acquire basic registered users.

**Q: How accurate is the gender/age detection? ** Answer: Gender/age detection is based on public information (such as avatar, nickname, description, etc.) and is not an ID card-level accuracy. Age is usually presented as a range or approximation (such as about 30 years old), and gender accuracy is generally between 80% and 90%, depending on the completeness of the account information. It is recommended to be used only for statistical stratification and should not be used for precise identity confirmation.

**Q: The platform deducts fees per item, is there any charge for generating a number? ** Answer: Global number generation (240+ countries, number segment generation, custom CSV import) is a free function. After generation, the number screening task will be deducted according to the platform unit price based on the number of detections. New tasks cannot be submitted when the balance is insufficient, and USDT must be recharged first.

**Q: What is the use of data deduplication warehouse? ** Answer: The deduplication warehouse compares numbers across tasks to avoid repeated detection of the same number and reduce waste of balances. Suitable for teams that import lists multiple times or use public number segments to generate multiple rounds of screening.

---

If you want to get started immediately and experience the number generation and screening process, you can log in to the console https://app.kkdata.cc/ to create a free task (generation is free). If you have any questions, contact the customer service robot https://t.me/kkdata_robot in two ways for real-time help. See https://docs.kkdata.cc/ for more documentation. Official website: https://kkdata.cc/.