Last time, we talked about the recommendation system. Today, we will talk about advertising briefly. I’m not going to get into too much technical detail, but I’m going to try to give you an overview of what advertising is on the Internet and how the advertising system works. As an advertising system, what are its goals? Knowing these things may come in handy when you’re looking for a job or changing jobs.
Advertising and Bidding
You may more or less feel that advertising is very important for Internet companies, is also the profit model of many Internet companies. Since most Internet companies don’t make money directly from users, but platforms need to, advertising is a good model. In fact, good advertising is multi-win, merchants and platforms have earned money, users bought their favorite products. Users who don’t spend money also enjoy the platform’s services, which is why advertising accounts for so much of Internet companies’ revenue, even more than half.
As users, when we surf the Web, we see ads. These ads are not fixed or random, but they are pulled from advertising libraries through search engines. In order to maximize the profit of the platform, advertising will be conducted in the form of bidding. Take a simple example, for example, when we open a web page, there is only one advertisement field, but there are three candidate advertisements, such as shampoo, car and medical beauty respectively. There are three candidates for advertisement. Which one should I choose? It’s very simple of course whoever bids the highest for the platform chooses.
That is to say, within the advertising platform, each exposure of display ads is actually equivalent to an auction. Advertisers will give their ads a bid, the logic of the bid is different in different scenarios. The best thing is to be able to dynamically bid based on the advertiser’s current budget as well as the traffic already received and expected, but this is difficult to do and not all systems have this capability. The simple thing is a fixed bid, and here we don’t need to be too detailed, just a general understanding of it.
Let’s assume that all ads are fixed prices, like $3 for shampoo, $10 for cars, and $50 for beauty treatments. So obviously, medical beauty will win, so this AD will show medical beauty. But there are two problems here, and we’ll take them one at a time.
The first problem is that the profit of the medical beauty industry is relatively large, so the advertising bid is high, so it is not shampoo and car can not be displayed? That’s not exactly true, because the bidding is multiple times, based on recall candidates. Shampoo may be underpriced, but it doesn’t always have to compete with medi advertising. So it’s also possible to get demonstrated.
The second problem is that the user may be male, who is not interested in medical beauty, and he will not consume no matter what you do. Isn’t that money wasted on advertising? So there’s the issue of AD selection, and we can’t just choose ads based on the price, because even if we show users don’t click, the platform won’t make any money. How can platforms make more money? Of course you want to expect the most. What is the expectation? It’s equal to bid x clicks, so the core is clicks, and with clicks we can estimate its expectations. How do you get clicks? Using models to predict, of course.
So we’ve strung this thread together. Why do we need clicks for AD scenes? Because you need to calculate the expected revenue that advertising will bring to the platform.
Second bidding strategy
There is also a very interesting strategy in advertising called second bid strategy. The meaning of this strategy is also very simple. Take the shampoo, car and medical beauty advertisement as an example. The shampoo is 3 yuan, the car is 10 yuan, and the medical beauty is 50 yuan. Apparently, we did the reverse, and then the car and then the shampoo. Instead of considering the impact of CTR on the AD, let’s just consider bidding. The AD should show Medi.
The usual auction is the first bid strategy, which means the highest bidder wins, and that makes perfect sense, and that’s the way it works in everyday auctions. But in the second bidding strategy, the highest bidder also wins, but the highest bidder does not have to bid, he only has to bid the second highest price plus a base, such as a penny or a dime (depending on the situation). This is called the second bid strategy. In that example, the AD for medical beauty is 50 yuan, but it only costs 10.1 yuan.
But our intuitive look will certainly feel wrong, ah, this is not less money, is the platform stupid, clearly can collect more money must earn less?
In fact, the platform is not silly, buy never sell fine. It’s very simple. Let’s think about it. Let’s say you’re a Medley advertiser and you bid on an AD for $50. So will you bid 50 again on the next bid? Obviously not, the reason is very simple, you bid 50 to get the advertisement, you don’t know how much other people pay, but you are sure to be lower than you, then you must try to lower the price next time, see if you can get the advertisement for less money. In fact, all advertisers do this. If they all bid down, is this auction still called an auction? How do platforms make money?
In order to prevent the situation of everyone price, so the platform adopted the second bidding strategy. So even if the advertiser shot the advertisement, also can’t press the price. Because the second advertiser’s price also fluctuates, it may also raise it. And the highest bidder has no demand to lower the price, because even if the price is lower, he will still be the second price of the advertisement, there is no profit for him. In this way, although it seems that the platform does not grab the maximum benefits, in fact, it can avoid the situation that everyone will lower the price together, so that the profits will be higher.
At present, basically the mainstream algorithm platform is to take such a strategy, such as a certain degree of search advertising, Taobao commodity advertising and so on.
Advertising logic
For advertising in the Internet era, it is very different from traditional advertising. For example, traditional flyers or TV ads, signboard ads and so on, these ads are fixed, there is no user customization function. No matter who gets the leaflet, it is the same content. It is impossible to say that different people get the leaflet and have different results. So the traditional advertising is sea, the conversion rate is very low, and it is often difficult to statistical transformation. For example, when we put an advertising sign somewhere to sell a certain household appliance, it is difficult to calculate how much of the sales of this household appliance is contributed by this advertisement.
But advertising in the Internet industry is different, because all of our content is electronic, it can be a huge step forward. The first is personalization, which can push ads to us according to our needs, rather than random push or push according to the needs of advertisers. The second is that it’s a great way to measure conversion and subsequent effects, because all of the information on the Internet is connected, from users’ views to clicks to purchases. We can also train advanced models to predict users’ preferences, so as to screen out high-quality advertisements for users and earn profits for the platform.
Speaking of which, why is taobao the only one of so many e-commerce companies in China making money, and earning a lot? Because Taobao has been a platform, play is flow, make money is flow. Taobao buys traffic from other platforms and sells that user traffic to merchants. What we buy on Taobao is not taobao’s own goods, but the goods of the merchants. To sell goods, merchants need to publish advertisements on the platform to obtain traffic, which is the revenue source of Taobao, not the money to buy things. Buy the thing is not good to find Taobao, Taobao also never favoritism, because Taobao does not rely on selling goods to make money, and the user is the source of flow, so Taobao will not be partial to businesses, basically can do notary. And like some east has been the main self-support, this is to make money by selling goods, the first two years have been adulterated and refurbished machine and so on, this series of problems will have later, because in essence some east to make money by selling goods, it and Taobao with low competition, in order to make money high quality must be difficult to maintain.
As we said earlier, in order to maximize profit, you need to invest in the goods that have the highest expectation of profit. We all know that the expectation is equal to the probability times the value, the probability we don’t know, nobody can evaluate it exactly. We can only use some machine learning or deep learning model to predict Click Through Ratio as closely as possible, which is the number of clicks divided by the number of exposures. So we can write the revenue of the platform as follows:
Here R refers to Revenue, or return Revenue, and bid is the bid made by the merchant. The formula assumes that ads are paid per click. But the ads in the platform may not be charged per click, for example, some good placement ads may be charged per exposure (view). For example, the location of taobao’s home page:
We see that there will be advertisements in the top column of the home page. These advertisements are in the best position and belong to the content that users must see. These high-value locations may not necessarily charge per click, but per exposure. What is the advantage of this? As well as making more money, these advertisers can be forced to improve the quality of their ads. This is the same reason that YouTube ads can be skipped. Because ads can be skipped, advertisers are forced to improve the creativity and quality of their ads, so that viewers are willing to watch them.
When we search on Taobao, there are advertisements in the search results. For example, if I search for a phone, the first one is an AD. Search results ads do not have their own unique copywriting and no special images, generally according to the click deduction fee.
There are also hidden corner ads, which are worse placed and may be charged for conversion, or purchases. The platform only makes money if the user actually buys the product. Such as some circle of friends, small website advertising is generally this kind, because the conversion rate is not high, if the advertisers are not willing to pay according to exposure or click.
For pay-per-sale ads, we can’t just predict CTR. Because some products may have a high click-through rate, but the conversion rate is not always high. For example, because I buy clothes for my wife, Taobao will give me some sexy little sister. These obviously eye-catching images or headlines tend to get a lot of hits because humans are visual creatures and can’t resist clicking on them. Those who have done advertising know that the click-through rate of these beauty pictures is very high. In this case, we can’t just estimate the click-through rate, we also need to estimate the conversion rate.
The technical name for Conversion Rate is CVR(Conversion Rate), which is the probability of placing an order for each click. The advertising revenue at this time is:
So the probability that the user clicks on it times the probability that the user clicks on it and buys it, is equal to the probability that the user sees it and buys it. In addition to the need to estimate CTR, the AD team has the additional task of predicting CVR.
conclusion
So just to summarize, the logic of an advertising system that we talked about in previous articles is very much the same as a recommendation system, except that targeting is different from recommendation and search, because targeting involves calculating revenue, so there’s an operation that multiplicates bidding and computes expectations. Two core values, CTR and CVR, are required to calculate accurate expectations. CTR is more widely used, and the predictive logic of CVR and CTR is basically the same. So like recommendations, the core of the advertising team is CTR estimation, and the day-to-day work of an algorithm engineer is actually very, very close.
And just to give you a little bit of insight, do you know what the richest advertisers in the advertising industry are? I said two. One is decoration and the other is medical beauty. The telephone of a owner that is considering to decorate recently is the quotation that decorates a company is about several hundred pieces, the telephone of a young girl that loves beauty is in cure beautiful company even more, the likelihood is thousands pieces. That is to say, they will pay hundreds or even thousands of dollars just to get your phone. Then the problem comes, not to mention the quality is not good, but the profit, which is how much profit?
That’s all for today’s article. I sincerely wish you all a fruitful day. If you still like today’s content, please join us in a three-link support ~ (like, view, forward)
Original link, ask a concern
This article is formatted using MDNICE
– END –