Information flow oCPC: Spent hundreds of thousands of dollars with no effect?

I saw a question: It said that someone invested in Toutiao's oCPC , spent 1,500 yuan, got more than 6,000 clicks, but didn't get any customer information. He asked why oCPC had no effect. My first reaction was that this might not be oCPC’s fault, because I suddenly remembered that a few years ago, when I first started investing in oCPC, I also spent hundreds of thousands of dollars before I found some tricks.

1. It took more than 100,000 yuan to reach the threshold of oCPC

I remember when Toutiao first launched oCPC, I was very excited after hearing about it, because advertising format based on conversion goal bidding has been available in foreign countries for a long time, but it has never been launched. Now that it is available in China, I naturally want to try it.

So I immediately persuaded the company to open an account, produce materials, and put them online. But the result was like an ice bucket challenge - the data of 0 conversion in 3 days gave me a chill.

That was the first time we were doing information flow and oCPC, and we were still using the thinking of search and display advertising. However, due to the special nature of our company's industry, Toutiao's advertising consultants had neither feasible suggestions nor effective strategies to offer, and there were no peers to refer to. They only suggested that we keep trying.

So I kept guessing, adjusting, summarizing, and guessing again... It wasn't until the third week that I was sure I had found the direction and it was working.

From opening an account to investing, to seeing results; from feeling completely disappointed to feeling elated, I spent more than 100,000 yuan in total.

Because the company had sufficient promotion budget at the time, there was no need to worry about spending hundreds of thousands of dollars to test a new channel .

However, for companies with insufficient budgets, they may have to stop advertising if the trial fails, thus losing a potentially better advertising channel.

Therefore, I wrote down the little experiences I have summarized, which may be of some help to students who are new to oCPC.

2. The principle of oCPC: control the bidding of the crowd and increase the value of a single click

The principle of oCPC can be explained in one sentence:

Increase bids for people with high estimated conversion rates to gain traffic ; lower bids for people with low estimated conversion rates to reduce impressions, ultimately making the average conversion cost lower than the set target price.

The core is to reduce conversion costs by controlling crowd bidding and increasing conversion rates (or, increasing the value of a single click).

For example:

When developing financial management products, two days after the system was launched, it was found that the conversion costs of IT men from BAT and middle-aged women in Beijing, Shanghai, Guangzhou and Shenzhen were better than the target price. When users who met the labels of these two groups were found, the bids for them were increased in the hope of obtaining more impressions, clicks and conversions.

There are also IT guys and housecleaning ladies with a net worth of tens of millions in third- and fourth-tier cities, but the overall conversion rate of the population is low and the cost is high. In order to achieve the set target conversion price, the system may lower the bid for IT guys and housecleaning ladies in third- and fourth-tier cities, hoping to reduce impressions and clicks and control costs.

In Xiao Cao's opinion, oCPC is a way of delivering advertisements that reduces the workload of advertising optimization .

From its principle, we can see that oCPC turns the boring and time-consuming work of crowd analysis and price adjustment into a procedural, background and real-time process.

When placing advertisements, we can focus on optimizing creativity, landing pages , customer service reception, and placement strategies, without having to analyze crowd data or pay attention to real-time prices.

As shown in the example above, advertisers can focus on producing targeted creative materials without having to spend time calculating the conversion costs of men in different industries and middle-aged women in different regions. The time required for statistics and calculations, as well as the low conversion consumption caused by adjustment delays, are all considerable costs.

3. Why oCPC: Media don’t like CPA, and we don’t like CPC

Since the ultimate goal is to achieve a conversion cost that meets the target price, can’t we just invest directly in CPA? Why the whole oCPC?

Because the media is not happy about investing in CPA.

The final conversion data is in the hands of our advertisers. Even if a third party approved by both parties is involved, the media is unwilling to bear the risk of CPA: no matter how high-quality and accurate the traffic is, it will be useless if our landing page is a piece of shit.

Then it’s still CPC, advertisers won’t be happy about that.

The click data is in the hands of the media. If the media says there are 100 effective clicks, then they will charge for 100 clicks. But where do the clicks come from? Sorry, we have blocked malicious clicks. The lack of conversions must be because your landing page is too bad. Do you still want to win the game after you die on the landing?

So to sum up: it is a good idea to launch oCPC, the media will not suffer any loss, and we feel it is appropriate to place advertisements - we still have to accept the conversion costs we set ourselves.

Since payment is based on conversion, you can invest in oCPC with confidence? The answer is no, otherwise there would not be the questions and experiences at the beginning.

4. How to make oCPC more effective

In other words, how can we make the background crowd analysis and automatic price adjustment effective?

If it is manual optimization, you need to count enough people and historical performance data of the advertisements, and then divide them into four quadrants. Lower the price for people with high consumption and low conversion, and raise the price for people with low consumption and high conversion.

oCPC also requires sufficient historical performance data of the population and ads to complete the work of estimating conversion rates and dynamic bidding.

Sufficient historical performance data requires: sufficient money and sufficient quantity.

Pay enough money, that is, bid high enough to get enough impressions

Students who have done searches generally understand that bidding should be high.

If your bid is not high enough, Baidu and 360 will not give you the opportunity to display, and you will not get any traffic; if your bid for the Through Train is not high enough, your product will be ranked low, and Taobao users, especially Taobao Mobile users, will not be able to find your product, and you will also not get any traffic. It won’t be shown until all the ads ahead of it have consumed their entire budget—possibly late at night.

The volume must be sufficient, that is, the limit must be high enough to get enough conversions

The purpose of sufficient quota is to accumulate data as quickly as possible.

If it is a bidding game, if the limit is not enough, there may not be enough malicious clicks; if it is a through train, if the click volume is not enough, it will affect the entire promotion plan.

Take the example of the Through Train: for dresses, the daily limit is 50 or 100. Dozens of clicks are not enough to test the models. By the time the models are tested and distributed, winter is here, and you are still selling cheap dresses.

The same goes for oCPC.

The quota was not enough, resulting in insufficient clicks, and conversion data could not be accumulated. The system's testing cycle for population and variables became longer... It took two months to test the right population, and I still sold a lot of dresses.

We understand the principle, but what should we do? Two steps.

To avoid different ways of dividing levels for different media, the account levels here follow the order of plan → unit → creativity from top to bottom.

Phase 1: Subdividing units and stabilizing budgets

One plan can be subdivided into multiple units, such as 3 to 5 units, corresponding to different groups of people, targeting, and landing pages.

The goal is to first find out large directional combinations that can achieve stable conversions manually, which can also reduce the learning cost of the system.

Don’t over-segment, which will affect data accumulation (20 conversions/5 units VS 10 conversions/10 units). Moreover, if you segment too finely, you will have to do a lot of units, which is tiring.

The initial budget does not need to be too high. It can be gradually adjusted based on the conversion situation, and ultimately a budget that ensures a stable number of conversions will be determined.

If there are no conversions or the conversions are too few, you need to consider whether the bid is too low, the limit is too small, the targeting is too detailed, or the creativity and landing page are not good enough.

There is another situation in which there is no conversion: the product is not suitable for information flow delivery , or the media does not match. This will be discussed later.

The second stage: gradually increase the volume and update steadily

After entering the oCPC stage, if the click-through rate and conversion rate are stable, the volume can be gradually increased.

The purpose of increasing the volume is to allow the system to obtain more data for learning, further optimize the conversion cost of the current unit, and increase the conversion volume.

The method of increasing volume is to gradually increase the target conversion bid under the premise that the conversion cost is acceptable.

If the cost and effect remain stable after the price adjustment, continue to raise the price until the price becomes unacceptable or the price increase affects the effect.

After the price adjustment, on the premise that the effect is stable, relax the unit targeting conditions, such as adding interests and crowd attributes, and continue to increase the volume.

After relaxing the targeting, if the effect remains stable and you want to increase the volume, you can split the unit and take out one of the targeting conditions of the unit as a separate unit to run the volume.

In order to ensure click-through rate and conversion rate, creatives, landing pages and units need to be updated regularly.

It is understandable that creative ideas and landing pages should be updated regularly to prevent aesthetic fatigue, but why should units also be updated regularly?

Because information flow is delivered based on the population, and the population has an upper limit on size and conversion, it will be exhausted over time, and it is necessary to relax the targeting standards or rebuild the population.

If it is manual optimization, it is necessary to constantly create new population variable combinations, and increase the price and display of variable combinations that meet the conversion cost to ensure stable conversion;

oCPC automatically mines combined population variables. What we can do is to gradually relax the targeting, create new units, and provide the oCPC system with more population data for testing and learning.

Let’s take the example of the financial product above: during the period of advertising , the number of middle-aged women in Beijing, Shanghai, Guangzhou and Shenzhen was fixed. After several rounds of advertising, the conversion cost soared. If you still want to target the middle-aged women at this time, you need to relax your targeting and explore the second- and third-tier cities.

The purpose of gradually increasing the volume and steadily updating is to allow the system to continuously have new population data to learn from, ensuring that when the effectiveness of the old plan decreases, a new plan can take over.

No conversion? Maybe it's not really suitable

At the end of the first stage, it is mentioned that the product is not suitable for information flow delivery or the media does not match. What does this mean?

If the product cannot be targeted to a certain audience, then neither manual nor system methods can find a targeting method that can produce stable conversions, and it is not suitable for information flow delivery.

If the users covered by the media do not match the target customers of the product, it is also not suitable to place ads in this media. Or, the media has never placed a certain type of product, so there is no relevant data for the oCPC system to learn, and the product is not suitable for placement in this media, or the early conversion and testing costs will be very high.

Speaking of which, when we started doing oCPC, the industry was special and we hadn’t heard of other peers doing information flow at the time. Toutiao had no data accumulation on the population and conversion data, which is probably one of the reasons why it cost tens of thousands of dollars to test.

Until I left that company, the delivery of various types of information flows had to at least be set with a special targeting setting to achieve a more ideal effect. And that directional condition is the result of hundreds of thousands of trial and errors.

However, the number of users covered and the advertising data accumulated by the current domestic media that can invest in information flow can already meet the needs of most industries, and the optimization algorithms can also stand the test.

Or take Toutiao for example. The fact that it dared to cancel the first stage of oCPC and directly enter smart bidding shows that it is very confident in the scale and diversity of data accumulation, as well as the accuracy and reliability of the intelligent algorithm.

Back to the basics, the core of oCPC is to control the bidding of the crowd and increase the conversion rate.

If you want oCPC to be more effective, you need to manually optimize and adjust the fundamentals of the advertisement, such as display, creative click-through rate, and landing page conversion ability, to cooperate with oCPC.

At the same time, it is also necessary to manually optimize oCPC-related data, such as target price and audience targeting, to make the system more accurate and efficient.

If manual CPC can produce good results, then in theory oCPC will do better than manual CPC.

If manual optimization is not possible, or if a plan is launched but not followed, no matter how hard you try, the result will be 0.

The author of this article is @小曹同学, and it is compiled and published by (Qinggua Media). Please indicate the author information and source when reprinting!

Product promotion services: APP promotion services, advertising platform, Longyou Games