Learn how to use Tencent advertising crowd packs in one article

Today I will show you how to play the crowd pack and share my experience

Currently, the industries that the crowd is more popular with are: games/finance/education and other advertisers with strict assessment . You can find more bigwigs in these industries to communicate with.

Today's article mainly focuses on the commonly used first-party/second-party data . The frequency of daily use of third-party data is not very high, so let's put it aside. Let's first take a look at these concept explanations

First-party data

Advertisers’ own data, such as APP advertisers (activation/registration/online time/paid amount, etc.)

Second-party data

Media platform data/vertical platform data, such as Tencent advertising platform, can be extracted as needed: advertising interaction behavior (converted users, uninterested users, etc.)

After reading the above explanation, you should have some understanding of the classification of data assets. Let’s get straight to the point. What are the ways to use them?

Directional gameplay 1: Expand the crowd

For example, if the advertiser's assessment target is: form/retention/payment/etc., find the advertiser and ask for the corresponding target population data, package it and upload it to the media backend for expansion

The data types mainly include: QQ number/mobile phone number/IMEI/IDFA

Be careful not to step into the pit

1. The minimum number of people to be expanded is 500,000. Generally, it is easier to get exposure if the number of people is in the millions (referring to the estimated exposure of the nationwide campaign)

2. The base number of recommended seed population should be in the range of 20-50 times the expansion level. Too large an expansion level will easily lead to inaccurate expansion pack population, while too small an expansion level will easily lead to poor exposure.

The following two methods are more friendly to the data asset management capabilities of small and medium-sized advertisers.

Directed gameplay 2: Expanding the seed population

Logic: Crowds of data from one party help machines model

This function does not require a large number of people, the number can be between 100-100,000

It would be better if the media recommendation base is 2K-10K data. Generally, I also put it at 10W+, because I always believe that the larger the amount of data, the smaller the error~

Orienteering Game 3: Tag Square

This function is based on Tencent's analysis of each user: user behavior (behavior/interest), population attributes (consumption capacity/work status, etc.), industry labels (games/finance/education, etc.) to perform label segmentation

You can filter and target targeted groups based on known crowd portraits.

Targeted gameplay 4: attract new customers/increase activity/increase estimated click-through rate/conversion rate

Attract new customers: exclude already converted audiences

Pull activation: exclude users who have converted but are inactive

Improve estimated click-through rate/conversion rate

Exclude audiences who have been exposed/clicked/converted

Exclude uninterested audiences and other features

This orientation is suitable for use with the first three orientations, and the effect is more fragrant.

A summary guide to avoiding pitfalls

1) How many packages can be bundled for each plan?

Method 1: There is no particularly accurate answer. You can judge by estimating whether the exposure has reached the level of tens of millions. In the early stage, use a package that is closer to the assessment standard to survive the cold start. After there are dozens of conversions (I start with 50), gradually increase the number of bundled packages.

Method 2: When planning to build, deploy by binding multiple precisely layered packages

2) What to do if you suddenly can’t run while shooting?

Check your pulse first, find the media to confirm whether there is any brand buying traffic today, and observe for a whole morning.

After excluding the above possibilities, it can be confirmed that the crowd of people who bought the package has washed you, and the exposure has declined (watch me use a big trick)

You can increase the bid by a few billion points (just a few 0.1 points) to give the machine a positive feedback stimulus. You can also add some new packages or turn on automatic expansion.

3) I don’t know what to do with the crowd portrait of my family

You can package previous conversion data or find and extract your own conversion population.

Then, use the DMP data management platform to find out your audience profile through crowd insights.

(Whispering: The people I often upload and analyze are slightly different from the real people portrait, but I still follow the media's)

Author: Science Delivery

Source: Science Delivery (wzdn0123456)