How to measure the effectiveness of advertising?

"Half of my advertising dollars are wasted, but I just don't know which half."

Retail tycoon John Wanamaker's words decades ago captured the difficulty of measuring advertising effectiveness.

Indeed, as an important part of economic operation, how to measure the effectiveness of advertising has always been the focus, hot topic and difficulty of the industry.

Before we delve into this topic, let’s first review how advertising effectiveness measurement has changed.

1. The long history of advertising effectiveness measurement

Modern advertising first emerged and flourished in the United States. In 1923, Claude Hopkins, a famous American advertiser, published a very important book in the field of advertising - "Scientific Advertising", which systematically expounded his views on performance advertising .

This advertising industry expert on results invented a series of advanced advertising methods such as coupons, free samples, mailing catalogs, test marketing, etc.

For example, in order to compare the effectiveness of two advertising copies, he would print the two copies on the same page of the same newspaper but leave different mailing addresses for receiving free samples. He would then count how many free sample requests the two copies received. This was almost the earliest and most accurate measurement of effectiveness in the advertising industry.

Hopkins sharply criticized another school of advertising at the time, the artistic school, saying:

"Some advertising people have abandoned their responsibilities. They have forgotten that they are salesmen and have started to think of themselves as actors. They have begun to pursue applause rather than sales."

His ideas had an important influence on the subsequent famous advertising man David Ogilvy, who said a famous saying in "Confessions of an Advertising Man" - "We do advertising to sell products, otherwise it is not advertising."

In his opinion, whether or not an advertisement can generate sales in the end is the criterion for measuring the quality of an advertisement. "The difference between one advertisement and another is measured by the scale of sales power, and the difference can be 19:1."

In a sense, what Ogilvy did was the "performance advertising" of that era.

Therefore, contrary to what many people believe, traditional advertising is not a field that relies entirely on creativity, genius, and inspiration. Instead, the leaders in this field have long focused on measuring advertising effectiveness.

Of course, the advertising effect at that time was still only a post-measurement and was not directly related to advertising spending. It was not until the great Internet era that advertising effect and advertising spending were truly directly linked.

2. Measuring Advertising Effectiveness in the Internet Era

The birth of the Internet is undoubtedly a technological revolution for advertising. It has changed the billing rules for advertising, redefined the accuracy of advertising, and brought about a qualitative leap in the measurement of advertising effectiveness.

First of all, the first change that the Internet brings to advertising is payment based on performance. Why can't traditional advertising do this? The reason is that although traditional advertising also focuses on advertising effects, the advertising effects at that time were relatively vague and difficult to quantify.

The significance of the Internet is to digitize the entire process of advertising. This digitization process allows every aspect of advertising to be quantified and measured. The direct result of this precise measurement is payment based on performance.

Thus, CPC, the earliest pay-per-performance billing method, was born - users only pay when they click, and no money is charged if they do not click.

Please note that along with the emergence of pay-per-performance, a new advertising settlement model has also emerged - bidding.

In fact, pay-per-performance and bidding are twin brothers that appear together. The underlying reason behind this is that pay-per-performance is essentially in conflict with the maximization of interests of advertising media.

For example, an advertiser pays 1,000 yuan and wants 1,000 clicks, which is equivalent to 1 yuan per click. But if he provides a very bad material, the click-through rate will be very poor.

In order to achieve these 1,000 clicks, the advertising platform needs to provide unlimited traffic to achieve the goal. If the purchase is not made according to the bidding, the advertiser will have no motivation to optimize the advertising creativity.

With the further development of the Internet, the connotation of "effect" in pay-per-performance has gradually become richer. Since clicks can be called an effect, can installation be called an effect? ​​Can activation be called an effect? ​​Can payment be called an effect?

The answer is both, provided that advertisers can obtain these data. Some of the above in-depth performance data rely on the platform itself, and some rely on feedback from advertisers. In short, when it comes to measuring effectiveness, Internet advertising can reach the very end of the conversion chain.

With this data, one important thing the advertising platform can do is optimize the results.

Effect optimization is the essence and core technology of Internet advertising. That is, if the platform knows that a certain user has converted in response to a certain advertisement, it can make simulated estimates based on the characteristics of both parties, and match the same advertising content to people with the same characteristics next time. This is how advertising accuracy is achieved.

3. How to achieve a closed data loop for performance advertising?

We just talked about that the most basic work done by Internet advertising platforms is CTR estimation, that is, click-through rate estimation. This is because the data can be obtained by clicking on the platform itself.

Although clicks are important, for advertisers, clicks are not the goal, they are just a means. What advertisers care about is actually the subsequent conversion link, but if the advertising platform does not have a subsequent conversion link, it cannot perform corresponding optimization.

Therefore, with the development of performance advertising, it has become an urgent task to build a closed loop of data feedback between advertising platforms and advertisers.

So how do we do it specifically?

Facebook made the first attempt, and the first thing it launched was a tracking feature called "pixel".

At that time, the landing page of the advertisement was mainly presented in the form of H5. Facebook's pixel was a blank pixel + a statistical reporting code. Advertisers were required to place this "pixel" in the corresponding position of the advertisement landing page. In this way, when users visited the advertisement landing page, they could click on different areas and Facebook would receive such in-depth interactive behavior data.

This is the simplest solution to implement. After all, the interactive behaviors on the advertising landing page belong to the shallow conversion behaviors in the conversion field. If you want to count the download, activation, payment of the App, the collection, payment, repeat purchase and other deeper behaviors of the e-commerce, "pixels" will not work.

Therefore, the advertising platform simply developed a special API interface, usually called a "callback" interface, which is specifically used to return conversion data, and the types of conversions have become very rich.

For example, the downloading, installation, activation, and payment mentioned above, e-commerce collection, adding to shopping cart, and payment, offline store visits, trials, and ordering, can all be transmitted back through this unified interface.

At this point, some people may have questions - the conversion data that follows is actually very important data, are companies willing to pass this data back to the advertising platform? In fact, there are indeed such concerns, and we will talk about the solutions for typical platforms for such situations later.

It can be seen that ROI is actually the core indicator that advertisers ultimately care about, and the other conversion indicators in the middle are the leading indicators of ROI. There is a strong correlation between the two, but this strong correlation fluctuates and changes.

Therefore, when the advertising platform cannot directly optimize for ROI, advertisers need to grasp and control the relationship between conversion indicators and ROI on their own. This is an important pain point for advertisers and is also the direction that advertising platforms have been working towards.

IV. Evolution of Internet Advertising Effectiveness Measurement

Today, in which direction will the measurement of Internet advertising effectiveness evolve?

One of the directions is conversion-based algorithm optimization.

For Internet advertising, the effect is a typical conversion funnel that decreases layer by layer. Some conversion funnels are short, and some are long. For example, tool products in the Internet service industry may have a relatively long funnel.

A typical conversion process for tool products is: ad exposure, ad click, app download, app installation, app activation, and browsing monetization ads.

There will be varying degrees of loss from the previous link to the next link.

In theory, now that bidding by performance is supported, you can bid separately for the above conversion events - for example, 3 yuan for a valid download, 5 yuan for a valid activation, no matter which conversion event you use to bid.

Today's mature advertising systems are able to perform targeted optimization, that is, push advertising content to users who are most likely to convert within the range acceptable to customers.

However, there is a problem here. The premise for the advertising platform to optimize conversion events is that advertisers must send the conversion event data back to the advertising platform, because the advertising platform needs these conversion data to feed back the system to identify and estimate which users are easy to convert.

But for any advertiser, conversion data such as audition, ordering, and repeat purchase are the most core data and the company's most important assets, so they also have concerns about the security of this data being transmitted back.

So, how to solve this problem?

Tencent Advertising's solution may be worth referring to. First of all, Tencent's advertising ecosystem is very complete. There is Tencent Advertising for App promotion and buying traffic, and there is also a platform for App traffic monetization - "Youlianghui". The benefits of this are very obvious.

That is, you do not need to send back revenue data, because Youlianghui already knows your revenue data accurately and in real time, which can provide a very favorable operating space for optimizing advertising effects.

So, how does Tencent optimize advertising effectiveness through this data closed loop?

The answer is a bidding model based on ROI estimation, that is, App advertisers can set a conversion cost for shallow conversion behavior and a minimum acceptable first-day monetization ROI (= monetization amount on the first day/consumption on the day of advertising).

The system will calculate the bid in real time based on the target cost of shallow conversion behavior, estimated advertising frequency, and ECPM, and try to control the advertising monetization ROI to be no lower than the expected ROI set by the advertiser.

The purpose of doing this is to improve the ability to obtain volume while ensuring ROI.

This is actually an upgrade of Tencent Advertising for optimizing the effectiveness of this type of advertising. Before the upgrade, a double-bidding model was adopted to estimate the next-day retention rate. The logic behind it is that there is a strong correlation between next-day retention and monetization. Therefore, the next-day retention indicator can represent the monetization efficiency in a sense.

But after all, the two are not equal. This approach will directly filter out some users with low retention but strong willingness to pay. These users are actually the users that the App should strive for.

The benefits are obvious:

First of all, advertisers do not need to send back all revenue data, because Youlianghui itself stores monetization data, and there is no need to send back monetization revenue from other channels, which eliminates advertisers' data security concerns to the greatest extent.

Secondly, the system automatically adjusts bids in real time based on estimated revenue, eliminating the need for pitchers to manually adjust bids based on first-day ROI and reverse-calculating retention rate targets. It also avoids incorrectly filtering out customers with low retention but high willingness to pay.

Again, due to the continuity of data, this solution can effectively avoid the problem of advertising plans’ ability to run in volume declining after a certain period of time.

It is precisely these leading advantages in ROI monetization that allow Tencent Advertising to still bring advertisers sufficient and excellent returns even after performance advertising has entered the deep water zone.

Performance advertising is a "results-oriented" industry that requires a sufficiently impressive ROI to gain recognition from advertisers. As a long-established performance advertising platform in China, Tencent Advertising has always been at the forefront of innovation in advertising effectiveness measurement.

Compared with competitive media, Tencent's advertising monetization ROI is more flexible. It does not force advertisers to set a monetization ratio, nor does it need to send back all media monetization data. A good estimate can be obtained by using large serial single-layer concurrent requests.

More importantly, in response to the pain point of advertisers experiencing a drop in ROI 7 days after using monetization, Tencent Advertising’s monetization ROI optimization strategy takes into account the overall LTV recovery curve, steadily increases LTV, and ensures continuous recovery.

Taking an advertiser in the tool industry as an example, the test account started in June and continued to run for 18 days. The data in terms of volume-generating effect, second-time retention rate, first-day and 7-day monetization ROI were all better than the second-time retention double bidding ads. The ROI on the first day exceeded the bid by 120%, and the sales volume can still be maintained.

There is no doubt that today's advertising effect prediction is no longer the same as in the traditional era due to the advancement of system capabilities. In addition to the result-based algorithm optimization just mentioned, our definition and understanding of effects are also constantly changing. For example, should the effect be determined by one indicator or multiple indicators?

If there are multiple indicators, which one takes priority? What are the weights of other indicators? Should we look at short-term or long-term performance indicators? If we look at long-term effect indicators, how can we solve the problem of immediate data feedback required by the algorithm?

These are practical problems in measuring advertising effectiveness in the Internet era, and solving these problems still requires our new generation of advertisers to continue to explore and develop in practice.

Author: Wei Xi

Source: Weixizhibei (ID: weixizhibei)