Source: Buy volume Jianghu
This article mainly shares the corresponding account diagnosis and optimization ideas around the three major pain points of "low volume, high cost, and poor quality" when the Douyin channel carries out information flow advertising, which is more suitable for the delivery diagnosis of AD delivery, especially for the delivery of sales lead collection, and other delivery methods are for reference only.
1. Goal alignment
Before the account diagnosis, we need to understand where the most important problems in the current account are, and then disassemble and find out the causes, and formulate optimization strategies.
Generally speaking, it is the three core difficulties of "no quantity, high cost, and poor quality", and the corresponding meanings are understood by everyone, so we will not explain them one by one.
There may be times when we encounter a situation where two or three of these exist at the same time, then we still need to make a priority and find the core points that need to be solved.
For example, when the volume is not large, the cost is high and the quality is poor, because your data accumulation is still too small, and the account is not enough to reduce the cost, so many times, the idea of starting the volume first and then controlling the cost and quality is right, but it does not mean that we need to raise the price significantly or pile up the automatic bid to get the volume, and then we must have a bottom limit for the cost to rise, otherwise the amount obtained after the cost is increased by two or three times, it will be difficult for the project to maintain.
Overall, if these are all at the same time, I think the priority to solve is not quantity> poor quality> high cost.
Because first of all, there is room for optimization and reference significance after the other two, and then if your conversion quality is good, I think the cost is a little higher, because OCPM is generally difficult to have two or three times the cost of floating, all of which are floating within 1.5 times the target cost, so that the quality is OK, and there is still a lot of room for testing.
The following is a detailed discussion of the diagnostic ideas for these conditions.
Second, the idea of diagnosis without quantity
1. Bidding space
Of course, I know that many times people actually feel that there is no room for price increase for various reasons, and all plans are run with the largest target cost, and if it is Party B, the bid is stuck by Party A.
Here I would like to mention a few ideas to help us check whether the bid in the account is in place, and how to formulate a bidding strategy.
At least align the market: It is necessary to understand the level of the market, and sometimes we calculate the bid by ourselves, such as how much the front-end bid must be through the funnel inverted indicator, or pat the brain and think that it may be this range, but sometimes if you really understand the bid of the market, you may still find that the bid is lower now. In addition, it is also necessary to consider that even if we align the market, if it is really difficult to get volume during the cold start period, we have to go out to a level higher than the market to accumulate data first;
Aligning the cost of accurate customers: In the end, we still want to accurately target users and achieve conversion, but if we have no problem with quality, the back-end conversion rate is relatively high, especially higher than that of peers, and I can offer a higher price when it is fed back to the front-end. For example, the cost is now 100, the proportion of accurate customers is 25%, the cost of accurate customers is 400, and the transaction rate is also very good, but the cost of accurate customers of peers is 600 or even higher, so we can further raise the price in the front-end bidding, and we are not afraid of being higher than our peers, because we have more accurate materials, or do a stronger conversion team, there is no need to only stare at the conversion cost at the account level;
Part of the high-price strategy: Continuing the idea of the previous point, through data analysis, we will find that the quality of the conversion crowd in some regions or part of the creative direction is better and the transaction rate is higher, so you can build a separate plan or open an account to place and give a higher bid to grab the volume;
2. Infrastructure maintenance
When the account is not enough, doing a good job of infrastructure maintenance is the most basic basic skill, whether it is the existing material more on the plan, or in the title, orientation and other dimensions of testing, it is necessary to maintain a certain number of plans every day, every day, as well as the history of the plan to clean up, how much is appropriate? From this point of view, when the account is not large, the large-scale production of high-quality materials is the core solution.
3. Material quality evaluation
This point is also the core, the underlying reason for the lack of quantity is that the system's estimation and evaluation of the material is not very good, and the amount of planning can actually be increased because of the amount of material, and the amount of material can often be increased by the plan.
When there is no quantity, your material update needs to maintain high-quality, multi-quantity, and high-efficiency output, so that the batch of materials should account for 90%, and there is another 10%, you need to brainstorm with the team, do some innovative materials and dots, return to the user's life scene, and strive to find a breakthrough point of the material.
So how to diagnose the quality of the material, there are two aspects that need to be combined, the first is the real data indicators, in terms of importance: bidding indicators (click-through rate and conversion rate> playback indicators (3-second playback and completion) > interaction indicators = negative indicators.
The second is some subjective analysis, whether the attractiveness of the material, whether the selling point is strong enough, whether the user scenario can really impress users, etc., is necessary for the refined dismantling and analysis of the video script, which requires some experience accumulation and understanding of user insights.
Here is the core content of the above article:
The easiest to run is not necessarily the one with the highest clicks + the highest conversion, but the closer it is to the industry median. The closer you get to the median, the more "familiar" the system will be, and the more confident it will be to give you traffic.
What does high click + high conversion mean much higher than the industry? It means that the probability of starting volume is large, but it requires the learning time of the model, because because you are much higher than the industry, if the system gives you traffic according to the average CPM, the system is at a loss, and he must improve the CPM to ensure his maximum return, and even lower the index for the pan-crowd, but this means that the system needs the model learning time to remain stable. But theoretically, this model also means that the quality of the material is good, given enough exploration time and model learning, and the probability of later volume is still very large.
For creatives with high click-through rate and low conversion rate, what to do is not to increase the conversion rate, but to reduce the click-through rate to see if it is too inducing clicks, or if the conversion guidance is too strong.
Low-click, high-converting creatives don't have to be to increase click-through rates. Instead, compare it with industry data to see the difference range. For example, if the industry has a click-through rate of 1.2 and a conversion rate of 6, your material has a click-through rate of 1 and a conversion rate of 6.5, which seems to be a little bit worse, but it is actually very close. Unless it is, the CTR is 0.5 and the conversion rate is 9, which is worth increasing the CTR.
4. Operation logs
The main purpose is to judge the necessity of changing accounts, and we need to pay attention to our delivery process, one is to look at the negative operation, and the other is to look at the adjustment process of the running volume plan, and the key is to judge whether there are frequent adjustments (including bidding and orientation) that cause the account crowd to be a little disordered, otherwise even if we do the correct operation in the future, it will take time for the account to adjust the crowd, and if it is fast, it is best to change the account.
5. Internal competitive environment
When all the things that can be done are done, there is still no amount for a long time, and there is a high probability that this account will face greater internal competition, and we will sometimes encounter one or two hundred different materials that have been produced, and several accounts still have no movement.
If you are Party B, objectively speaking, Party A must have opened a lot of accounts under his qualifications in other agents, if the accounts of other agents are running, you will be difficult to start, this is an objective reason.
If it's Party A, pile up one or two hundred materials, and each agent can't afford to do it, which is very rare in itself, it must be that your product is not good, or there is a problem with the conversion link, or there is a problem with all your material directions, don't vote quickly, make a big adjustment.
Third, the idea of high cost diagnosis
1. Bidding space
Everyone knows that the price can be reduced when the cost is high, but everyone also knows that the price reduction may mean that there is no quantity.
There are several ways to think about the bidding strategy when the cost is high:
If the demand for the magnitude is not urgent, you can run the volume plan to reduce the price directly, of course, the range is not too large, generally about 5%.
If you need to ensure the magnitude of demand, it is more prudent to create a new low-price plan for the running volume material, and build a lower-bid plan for all new plans in the future, and the historical running volume plan will not be left for the time being.
If the supply of materials is relatively sufficient, you can open a new account and make the materials different and the pages are different (to avoid squeezing), and then put them at a low price to reduce the overall cost.
2. Check whether there is a crowd deviation in the direction of the material
Generally speaking, the high conversion cost is due to the low conversion rate from the data indicators, but this does not mean that it is a problem with the landing page (frankly speaking, as long as there is a basic understanding of effect advertising, everyone's landing page is not a big problem), and it may be more due to the deviation of the crowd model attracted by the material.
The biggest manifestation is that the material is too "digging up the pain points of life" or "awakening the demand", such as placing an advertisement for a certain professional skill, and the material side is telling about low salary, wanting to be promoted, Life is confused, so the people attracted are actually very wide, but whether they really make up their minds to start learning a skill, in fact, is very difficult, you really need to pass one or twenty seconds, to let users decide a major decision that they have never thought of, and the more accurate group of people is the group of people who have related ideas or ideas, that is to say, the entire account of the crowd is biased, and the cost will of course remain high.
If you feel that the direction of the material is okay after checking or adjusting, but the cost has remained high for a long time, you can return to the data analysis of click-through rate and conversion rate, mainly to solve the problem of conversion rate, there may be problems with the placement of dots, problems with the attractiveness of the selling points of the product, and problems with the "hook products" of conversion.
3. Operation logs
The same purpose as looking at the operation log under the diagnosis idea of insignificant amount, check whether there are too frequent bids and directional adjustments, resulting in the disorder of the overall system of the population model, which is in the re-exploration period, at this time, there will be a certain cost fluctuation, and it is necessary to be patient to slowly accumulate and correct the population model, if you are in a hurry, it is recommended to change the household directly.
4. It is not feasible to stop investing in high-cost people, and it is not feasible to give benefits
There is a way of thinking on the market that it is too naïve to check the crowd portrait of the account to exclude those who are related to high-cost region, age or device information, thinking that if you don't invest in high cost, the cost will be low, which is too naïve, combined with my own experience, this operation is actually a negative operation, because if your bid is not adjusted, as the delivery progresses, the system must become more and more stable for "what people are easy to convert and keep the cost in the bid range", and if you go to the "assistance system" in the middle of the way"Excluding the high-cost crowd, in fact, it increases the learning cost of the system, even if it is to adjust the region, it is easy to have a higher cost and no quantity.
In addition, try not to put out the welfare, whether it is to send discounts or free information, although this improves the conversion induction, but often the quality of users is relatively poor, can only solve short-term problems, and in the long run, the gains outweigh the losses.
Fourth, the idea of poor quality diagnosis
1. Judge whether it is or not, and the difference
Generally speaking, the conversion of delivery is definitely not 100% in terms of efficiency, then we need to understand the average efficiency of the industry corresponding to the channel in this industry, and then see how much our difference is, like form collection, the range of intention rate of 30-50% is considered normal, and there are differences in various industries, so if the small partner who converts the undertaking says, most of the clues are invalid, You must know that this is normal, and this is also the case with peers. Of course, we don't necessarily have the back-end data of our peers, so at this time, it is more important to find other Party A students in the industry through communities, planets, etc., and ask about relevant data.
2. Check the direction of the material to judge the account model
The principle is the same as the idea of the second point of population bias under the diagnosis idea of "high cost", if the crowd model corresponding to the material is biased, or relatively front-end, even if the front-end account cost is OK, the conversion quality of the back-end will be relatively poor.
Although our materials seem to be introducing products, we can split the user's intent stage to think about which stage of our material will be most attractive to the crowd, and the closer to the intention of the transaction, the better the quality of the user. The core idea is to "dig less pain points and brighten more products".
3. Channel and regional optimization
The quality of the crowd will be much worse under different channels, such as the byte system, the quality of the crowd from these traffic sources such as Douyin, Toutiao, search promotion, and pangolin needs to be counted separately, and the lead collection placement must not be cast on the panshan, and all the fake amount.
Generally, the best effect is still Douyin, and if you find that a large number of sources are Toutiao and search promotion, and then the quality is not good, you can adjust only Douyin and turn off search fast investment;
In addition, there is a clear difference in the quality of regional optimization, mainly because many products have different local policies in different regions, as well as user cognition, and after a period of time to accumulate a certain amount of data, you can carry out separate high bidding for high-conversion regions, or directly reduce the conversion of poor regions.
4. Raise the threshold for conversion, increase the depth of conversion, and only look at the cost of in-depth conversion
Raising the conversion threshold is actually to screen the real high-intent people in the operation, such as no longer putting free materials, free evaluations and other content on the conversion point, reducing the discount, etc., adding questions and options to the form options, directly submitting the form to 1 yuan, 9 yuan paid forms, etc., so that after the operation, it may cause the click-through rate and conversion rate to decline, and higher bids are needed to run the volume and the conversion cost will inevitably rise, but we actually need to pay attention to the depth of conversion cost, that is, how much money it costs to obtain an accurate customer.
For example, before the adjustment, the conversion cost of 100 seemed to be low, but the effective rate was 20%, and the actual effective customer acquisition cost was 500, but after the adjustment, the conversion cost rose to 150, but the effective rate was 40%, and the actual effective customer acquisition cost was 375, which is more cost-effective and a successful optimization.
Of course, this process needs to be tested, and you don't know how much the front-end cost will rise after adjustment, and how much the efficiency can be improved.
5. Crowd sourcing strategy
DMP crowd strategies can be considered for quality issues, but the actual effect requires a testing cadence and real-time feedback.
There are two ways of thinking, one is to target high-quality people, and the other is to exclude low-quality people;The first is more difficult, because optimizers often do not have high-conversion crowd packs, and if there are already used them, it is difficult to calculate what groups are really high-quality through the background, unless you find a third-party reliable crowd packet data platform that can supply them.
Focusing on how to exclude low-quality people, there are two approaches that are relatively feasible:
Through the tag market, the obvious impossible groups are screened out for exclusion, such as career prediction, spending power, college students, behavioral interests, etc., based on the insight and understanding of the target user's crowd, delineate non-audience groups, and then exclude the delivery. Because the system has its own exploration algorithm, the target audience of your circle may not be accurate, and even the effect may not be as good as the system itself, and the system cannot determine the special accuracy of the exploration in the process of exploration, so it is helpful to exclude some wrong people;
Through the establishment of data monitoring backhaul, data collection of the display group and advertising contact group to generate an audience package, and then upload it to the Tencent advertising knowledge background to analyze the portrait of the audience package (there are also huge but not Tencent's comprehensive), through the display of the account and the click group of the portrait analysis, judge the quality of the account crowd, if the analysis results deviate greatly from the target audience, you can consider packaging and uploading the "low-quality conversion group" in the huge number of "converted people", and expand the exclusion test in a small amount.
Why not just exclude the low-quality people from the converted population? Because it is a very dangerous behavior to exclude the converted crowd, especially for stable old accounts, the crowd model of the account is based on the converted crowd, but you have to exclude some of it, which can easily cause cost fluctuations or order of magnitude attenuation, so after passing Tencent's crowd portrait check, we found that it is indeed realized that the crowd model is crooked, so it is recommended to do this to further correct the crowd model, which is best done with material adjustment and account strategy.
