Brief introduction: in the process of supporting the operation and development of almost all the core businesses of ant, we have made many attempts in platform construction, business support, platform operation, AI innovation and overall operation of AI, and we have gained a lot of gains and insights. I would like to share them with you here.
In the past few years, my team and I have been responsible for the design and operation of relevant platform products within ant group.
These platform products include a / B test platform, machine learning platform, financial knowledge mapping platform, NLP platform, intelligent copywriting platform, financial vision (CV) platform, search platform, robot platform, labeling platform, etc. of AI department, as well as related platform products of risk control team. These platform products support the operation and development of ant’s almost all core businesses.
During the whole process, we made a lot of attempts in platform construction, business support, platform operation, AI innovation and AI overall operation, and gained a lot of gains and insights.
Recently, I spent some time to sort it out and write this article.
The content of this article covers seven aspects, such as “requirements management, platform design, product verification, platform collaboration, human confrontation, cross-border thinking, challenge / growth”. There are not only some abstract and methodological summaries, but also many real and meaningful cases.
The length is relatively long, about 15000 words. If you are interested, you can collect it and watch it slowly.
I hope this article can inspire you, and I look forward to further discussion and exchange with you.
1、 Demand management: “role dislocation” and “selfless state”
1. Mining demand, vigilance “role dislocation”, put an end to “behind closed doors”.
The first step to do a good job of a product is to grasp the demand. We must know who the real users of each product and function are.
For C-end products, this problem is easier to solve, because designers and users often overlap. But for technology platform products and b-end products, the two are often misplaced, that is, the designer may not be the real user.
For example, Alipay’s product manager pays and manages Alipay every day in his daily life. He is a typical Alipay user, so the designer and the user are the same person. In the technology platform and b-end products, product designers can use their own products, but they are basically limited to testing and verification. The real users are others.
Therefore, the designer’s reasoning and judgment on product requirements may be different from the real situation. Even if he uses it, there is still a difference between the test oriented use and the real use.
Thus, it is precisely because of the dislocation of this role in technology platform products that it is easy to lead to problems in demand control.
Now, let’s start with a little story about our platform.
One day in December last year, we had a meeting with the relevant students of the platform to review the product design.
In the meantime, for a product design detail marked on the page, the students in the product manager, ued, front-end and back-end positions expressed their opinions and argued endlessly.
Suddenly, I realized a serious problem that all the students in the conference room are not users of this feature.
Because the specific tagging work is done by hundreds of tagging personnel of outsourcing companies, they are the real users of tagging pages.
If you are not a real user and are not in that scene, it is difficult to understand the real situation. Therefore, we can only judge and deduce according to our own experience and professional ability.
You can’t build a car behind closed doors. So, we immediately arranged for relevant students to go to the outsourcing company to do on-site research.
At the beginning, we had a small-scale preliminary communication with the leaders of several labeling teams. At that time, they casually asked about the use of the product, and they unanimously said “no problem, it’s very useful.”.
This is a normal answer. After all, it is very difficult to obtain valuable information and understand the needs of users with such a simple and direct method.
In the industry of product manager, we often say that before the invention of automobile, if you directly ask the user what he wants, he can only say “I want a faster horse”.
When the former head of nailing didn’t recruit students to share the “road of nailing entrepreneurship”, he also talked about this issue.
His view is that seeing users can’t just “talk about the matter” and only ask shallow questions about the use of products（ Even if you ask such a question, you can’t ask straightforward questions such as “what do you need” that are difficult to obtain real needs.
The right way is to leave the specific products behind and learn more about the background, business, status and other overall, background and context information of customers. We should show “interest” in customers and want to be friends of customers.
Only in this way can customers be willing to chat with you more and deeply, and only in this way can you capture valuable information. In addition, by observing the specific behavior and operation of customers, we can capture the real demand and have insight.
So, after the meeting, we asked to go upstairs to the office area marked with employees to see the specific situation.
When we stand behind the taggers, carefully observe their operation, and have a deep conversation with them, we will have a new discovery.
Many previously unimaginable usage methods and scenarios, product design details, in the continuous operation of the tagging personnel, it shows up. Before the product review meeting, we debated the question, naturally have the answer.
In half a day, we recorded dozens of valuable feedback and findings, and dealt with and followed up one by one in the follow-up work.
It can be seen that if you are not a real user and you do not observe the operation of a real user with your own eyes, you can not expect many problems.
Everyone’s IQ is not bad. When we encounter problems, we are often used to talking about methods and logic. We often fight with each other in the conference room, even slap the table and stare at each other. In the end, no one can convince anyone and get effective conclusions.
At this time, you might as well ask yourself “who is the real user?”, Try the “stupid method” again, go out of the office, go to the customers, ask them, chat with them, and see how they use our products.
At that time, many problems became clear.
2. Meet the needs, constantly “from shallow to deep”, cultivate “selfless realm”.
Then, let’s think more deeply.
Now, if you have made clear who the users are and found out the general context of the requirements, you should also consider the problem of “whether the understanding of the requirements is deep”, that is, the problem of shallow and deep requirements.
In other words, it is also a matter of means and ends – “shallow needs” are often only means, while “deep needs” are ends.
For example, for the financial vision platform we are responsible for, there is a user feedback “I need a model report”, that is, after the model is trained, some indicators such as “accuracy rate, recall rate, AUC” are displayed in the form of charts.
If you just do this requirement, it’s not enough.
Why? Because the user’s model report is just a “shallow demand” — he really needs to look at various indicators, but what he wants most is that after the new model is trained, he needs to compare the effects of different versions of the model — not only to know how many indicators are, but also to know the specific changes of indicators, which are increased, which are decreased, and the specific values.
Only in this way can we meet the deep needs.
The reason is the same, similar problems will be encountered in the C-end products.
If you pay attention, you will find that many product managers of e-commerce websites and auto guide products have already felt the deep demand.
For example, the car website basically has a “model comparison” function: it can not only list the configurations and parameters of different models one by one in a table, but also provide intimate functions such as “highlighting different configurations and hiding the same configuration”. This is the deep level to meet the needs of users.
Therefore, for a demand, ask more why and ask yourself “is this the real purpose of the user? What does he want to do with this function. Only in this way, it is possible to touch the deep-seated needs of users, and it is possible to make functions that make users feel very intimate.
In addition to shallow and deep analysis, the idea of “divide and rule” can be adopted to meet the needs of users in depth. The product is divided into “n-stage Rockets” from modules and functions, and each “rocket” is used to meet the needs of different types of users, or the needs of the same user at different stages.
For example, although the functions of our map, NLP, CV, search, robot, annotation and other platform products are different, we still find the commonness, that is, we abstract the “five stage rocket” of demand grading and business enabling, including “function embedding, API calling, data training, model customization, algorithm development” and so on. The service side can choose different access modes according to the specific situation.
- The first level, function embedding: through iframe and other means to achieve the lowest cost, a function module of the platform is embedded into its own system.
- The second level, API call: directly call the mature API provided by the platform, such as calling the OCR recognition API such as ID card and driver’s license.
- The third level is data training: the model of the platform meets the requirements, but it needs to provide its own training data or dictionary data to solve the specific scene requirements.
- The fourth level, model customization: the field model of the platform does not meet the requirements, so we need to configure the algorithm parameters, and then train a new model to meet our own needs.
- The fifth level, algorithm development: the most advanced situation is that the business side understands the algorithm and wants to develop new algorithms. The platform provides a series of deep-seated capabilities such as “algorithm development, data management, model training, model testing and release” to improve the efficiency of algorithm development.
The above-mentioned “five stage rocket” meets the needs of different types of users and different stages of the same user.
I remember many years ago, I attended an advanced training course in management. The training has lasted for several days with a lot of contents, but I have forgotten almost all the training contents except a “universal four step method” introduced by a teacher.
The so-called four step method is the four steps of “classification – sorting – finding rules – Application”. No matter when you are learning new domain knowledge, taking over a new job, or coming to a new environment, you can try this universal four step method. I believe the most complex problems can be solved easily.
User layered, five stage rocket is an application of “classification sorting”.
After talking about “demand / user stratification, five stage rocket”, is it to meet the user’s demand 360 degrees without dead angle?
The answer is no, because we haven’t achieved “selfless state”.
The so-called “selfless” realm is that when meeting the needs of users, we should not only consider “who I am and what I have”, but forget ourselves and see what users need and what is most useful to users.
For example, although you are a product manager of AI technology platform, you can’t only see AI, algorithm and model in your eyes. To achieve “selflessness”, you need to achieve: if there is a non algorithm and non AI product strategy, if it can help the business, you should also do it.
In the eyes of business students, it doesn’t matter whether there is an algorithm or not, whether it is high-tech or not is not important – but whether there is a business effect is the key. As the saying goes, no matter white or black, catching mice is a good cat.
For example, our intelligent copywriting platform can intelligently generate thousands of marketing copywriters. In the past, we have been iterating products and improving algorithm capabilities, trying to generate more intelligent, accurate and personalized copywriting.
However, we all know that the improvement of the algorithm can not be achieved overnight, and the algorithm effect is slowly polished and optimized.
In this process, the product manager can’t wait.
Therefore, we are thinking that no matter how advanced the algorithm is and how intelligent the platform is, what we produce is still copywriting. With the development of advertising industry, copywriting has existed for hundreds of years, so there must be mature methodology and model.
As Internet practitioners, we advocate innovation and subversion, but we must also be awed by the industry.
As a result, our product manager students went to study some classic books on marketing and Advertising Copywriting, and summed up the so-called “18 kinds of high-quality copywriting sentence patterns / templates”, which includes not only the experience of copywriters, but also the scientific principles of advertising, psychology and other fields.
After the production of these “high quality sentence patterns” and “copywriting rules”, with algorithms and technologies, we can output more effective copywriting to the business.
We believe that machines can not completely replace human beings. Machine intelligence, industry knowledge, expert experience and other human wisdom will complement each other.
2、 Platform design: platform products must also be understood in seconds
After talking about requirements, let’s talk about design.
In the Internet industry, products for C-end users are not only abundant, but also generally free, and the acquisition cost is basically zero.
If you don’t pay, you won’t “cherish”.
Therefore, for users, the product must be easy to use, that is, it must be “understood in seconds”. If the user can’t understand or use it for a few minutes or even dozens of seconds, he will give up and the product will have no chance.
In fact, it’s the same for mid platform and platform products, except that users can only endure unpleasant experiences, because it’s their essential work to use your products to solve their business needs.
However, this does not mean that the product can be made casually, because it has other options. You know, companies often have similar products in-house, let alone external, free, open source or fee based solutions.
Therefore, you also need to work hard on the platform design. You must be able to quickly grasp the users, let the users quickly start, access and go online, and help the business get the business results.
How can we achieve “second understanding”? It can be considered from “product framework, terminology system, help guide, product demo, unified interaction” and other aspects.
1. Clear product framework
As soon as users open the platform page, they should clearly perceive what the platform can do, what the product framework is like, what functional modules are included, the relationship between modules (including, sequence, etc.), what to do in the first step, what to do in the second step, and so on.
This may not seem profound, but the common problem is that product managers do not think enough about the framework in the early stage of product design. It is often in the PRD and visual evaluation stage that we find that the module design is unreasonable and the process is not clear. At this time, rework, change, the cost is big.
What’s worse, frequent rework and change will make the product manager lose his professionalism and authority. In the future, how to raise demand for technology and improve resources?
In order to avoid such a tragic thing, the product manager should work hard on the stage.
A good habit is to reconstruct in the brain first, and then start painting. Many product managers are used to drawing demos as soon as they come up. This is wrong – the cognitive and computing resources of the brain are limited. If you look at “this”, you will lose “that”. When you fall into all kinds of details, it is impossible to think about problems from the fundamental and framework.
Then what shall I do? We can make full use of brain map. Specifically, instead of considering any demo diagram, you should first sort out the product hierarchy of the whole platform, including navigation and modules at all levels, pages and core function blocks contained in each module. After drawing the brain map, from the user’s point of view, repeatedly comb and simulate, until the horizontal and vertical logic and process are no problem, then start to do the specific demo and PRD.
2. There is a system of terms as the name suggests
After sorting out the overall framework of the product, we should also pay attention to the “terminology / concept system”, that is, the naming of the core concepts in the product and the design of the logical relationship between the concepts.
The reason why this is important is that the concept and terminology system is the result of the precipitation of knowledge in each field, and it is also the medium for people to learn new things and communicate.
The concept is complex, the product is bound to be complex; Simple concept, simple product.
For example, the same is the man-machine interaction instructions and ways, WeChat’s “shake” can let users “as the name implies”, and immediately have a sense of the earth to do, and our Alipay “cheese”, it is difficult to understand and put into action.
For another example, when Steve Jobs released the iPod, he did not directly and abstractly say that “the storage space is as high as 4.8g”, but that “put 1000 songs in his pocket”.
It can be seen that unreasonable naming of new concepts in products or direct exposure of obscure underlying terms will cause great trouble to users.
For another example, in the A / b experimental platform, the initial concept system from top to bottom is “business domain – business line – Product – Experiment”.
We found that it is difficult for users to distinguish between “business domain” and “business line”, and the “products” in it are not products like “payment, loan, flower and yu’ebao” as we all understand them, so there are a lot of problems.
Later, with the help of the well-known “physical laboratory and chemical laboratory”, we transformed the conceptual system into this: Darwin is an “experimental platform”, in which “XXXX laboratory” and “yyyy laboratory” can be created, and various “experiments” can be done in each laboratory. In this way, it is easier to understand.
In addition, we also modified the role naming in the lab.
In the previous experiment authority management, there were two common roles of “Administrator” and “member”. We also changed them to “laboratory director” and “researcher” by referring to the post names of laboratory staff in real life.
Interestingly, “researcher” has the hierarchical meaning of “high P / Organization Department” in Alibaba system. Such a small modification of the copywriting also contains the “humanistic care” of the platform designer, which is to give a little warmth and glory to the students who practice data-driven innovation and pursue scientific and rigorous way of doing things with a / b experiments.
Moreover, future operation activities are better, such as “top 10 researchers, top 10 laboratories” and so on.
In a word, in the terminology system of product design, first of all, we should “do not add entities if it is unnecessary”. Secondly, we should try to use the existing concepts in our minds instead of directly copying technology or creating new concepts.
3. With the right help and guidance
Even if you work hard on conceptual design, there is no guarantee that users will not have any questions.
Therefore, we need to design a “help system” for further explanation and elaboration.
This is not to say that you should write a lengthy product document. The document should be written, but it’s not the point, because most people don’t read the product document carefully before they start to operate – only when they encounter problems can they check the manual.
The “help system” here refers to the help system of productization, that is, “document productization”. Specifically, it is to embed the key points in the help document into the product page as far as possible, so that the product can realize “self explanation”, rather than putting them outside the product and only saving them in the help document.
The specific measures include the following aspects:
There is an auxiliary description on the page
The common situation is that our page is too clean and empty to put an explanation. When users encounter problems, they are at a loss. Therefore, you can make small word explanation under the title and tip bubble tips on the concept. For complex situations, you can also add a “learn more” link after the help text – jump directly to the corresponding place in the help document instead of asking the user to look it up from scratch.
New function online, with tips and notification
The platform is constantly improving iteratively, but it is often found that users do not know the new features. Therefore, we can make appropriate tips and notification: large iterations can mask pop-up windows, small changes can produce small red dots, and so on.
4. There is a simple and intuitive demo of the whole process
You can’t learn how to swim by watching the teaching video, and you can’t learn how to drive by watching “subject one” of optics.
It’s better to play with it than to talk about it for a long time.
The current situation is that many technology platforms have no demo and experience capabilities at all. Then, it’s hard for users to get started.
Therefore, the platform must build a “full process, somatosensory, simple and easy” demo to let users experience it by hand.
The whole process means that your demo should cover all links and steps of the platform. Somatosensory refers to the intuitive results (rather than just displaying abstract values, JSON code output, etc.). It’s easy to do, which means it’s simple enough and can be done in a few minutes (so you need to build several groups of demo corpus, maps, data sets, etc.).
For example, in NLP platform and financial vision platform, users can easily experience the effects of financial ner / text classification and ID card / bank card OCR online.
It can also complete the whole process of “project creation, data upload, data marking, model training, model testing” and other links.
It’s worth pointing out that for the platform demo, the simpler the better. Don’t overestimate people’s patience.
I remember that after the first version of the full process demo of the financial visual platform was launched, when members of the project team had a specific experience, they found that it was still very cumbersome and even had to give up.
To complete the demo, you still need to write a bunch of forms, such as project name / introduction, model name / introduction, dataset name / introduction, and prepare your own training data. You have to search online and download dozens / hundreds of pictures
Later, we made a great simplification. Users who can click the mouse should not be allowed to enter words, such as automatically filling in various names and profiles. In addition, some test data sets are built into the platform for users to use, etc.
After some simplification, users can complete the whole process and feel the demo in a few minutes.
5. Standard / unified interactive experience
In addition to the design of each platform, we also need to consider the experience consistency of different platforms, that is, the unification of platforms.
Doing this well can not only reduce the learning cost of users and switch smoothly between different platforms, but also reduce the repeated work of ued, product managers and technical students.
Firstly, the general framework and modules of the platform can be abstracted and unified, including portal page, project management, permission management, data management, task management, release management, etc.
Secondly, the details of the experience will be unified, specific to the component design, naming, color, location and so on.
When we precipitate a set of classic product framework and interaction standards, the product iteration speed and user experience will be greatly improved.
3、 Product verification: if you don’t use it deeply, you can’t do it well
1. In depth verification, not skimming
If a product manager wants to make a good product, he must use it more by himself.
The reason is very simple, but what we want to talk about here is the “depth” of use, which is very different from the depth of use.
For example, if you are asked to design the turn signal in the navigation product, you may think that it is similar to “turn right at the intersection 500 meters ahead”, so it is no problem.
You see, it not only contains the distance, but also makes the direction clear. It feels perfect. However, when you deeply use the product and drive your own car, you will find that this is not the case – it is difficult to accurately grasp whether you have reached 500 meters, and it is likely that you have made a wrong right turn at a junction at 300 meters.
Therefore, the current navigation prompt will not only say “turn right at the nth intersection 500m ahead”, but also prompt “passing the nth intersection-1” at the intersection where you should not turn right. Only by doing this, can you ensure that users will not go wrong.
For our annotation platform, the depth of use is reflected in the number of times to do data annotation – annotation several times and dozens, hundreds of times, your perception is completely different.
Some of the design details of the tagging page are not obvious when you do tagging once or twice. When you do it dozens or hundreds of times, the smallest problems will be exposed and magnified.
For example, there is an image classification task, you just need to mark “right” or “wrong”.
The previous design was to show a large picture on each page and switch to the next page after answering the questions. When we have marked dozens of them ourselves, we feel that the efficiency is very low.
So, we changed it to one page to display 20 pictures. The taggers only need to glance at them, check out the “right” or “wrong” ones, and then submit them as a whole (at the same time, it also reduces the waiting time for each page to refresh the page and load the pictures). Such a simple change, in fact, there is no technical difficulty, but the labeling efficiency is directly improved many times.
2. Doing business on your own makes a big difference
If we really want to do a good job in a platform, we should not only be “taggers” as mentioned above, but also be “business side”. There is still a big difference between supporting business and enabling business and doing business on your own.
Now, let’s use the case of our garbage intelligent classification project “sorting treasure” to illustrate.
In July 2019, many cities across the country began to implement waste classification.
Based on the AI technology capabilities of precipitation image, NLP and atlas, our students rapidly developed the technology and products of intelligent waste classification, and the project is named “sorting treasure”. Users can experience “AI garbage sorting” through convenient interaction such as “photo taking, voice search” and so on Alipay small program and intelligent garbage collection box IoT device.
This project is not started by Bu of various businesses to meet our needs. This time, we have a dual identity. We are not only the platform side, but also the “business side” for the first time.
After we started our business, we found that garbage classification seems simple, but in fact it contains many complex links, from “acquisition of training data, sorting of item categories, maintenance of garbage classification standards, correction of online return data” to “adjustment of item category weight and priority, and confirmation of labeling results”, Then to the cooperation with internal departments, docking with outsourcing ISV, coping with holidays and special items, and so on.
After a flurry of twists and turns, the project was finally stumbling up.
In this process, we encountered a lot of problems that we didn’t know before, including unreasonable product design of the platform and technical problems such as too long training time.
What’s more, we can see the “vacuum zone” between different processes, different systems and different teams. This is the problem of “no care, no play” brought about by the division of labor and boundaries of large companies. These problems, which are hidden and greatly affect the efficiency, need to be found and solved through the mechanism of products and processes.
This practice of “doing business by ourselves” has changed the perspective of our platform students, deeply realized the difficulty of business students, and directly promoted the iterative improvement of the platform, as well as the improvement of team cooperation and process setting.
4、 Platform collaboration: connecting, generating value
A lot has been said before, but most of them focus on the individual of a certain platform.
If a platform exists in isolation, it may be degraded to a tool, and its value and energy will become very limited.
Therefore, to do a good job and expand the platform, we need to jump out of the platform itself and look at it from the perspective of connection, overall situation and ecology.
If different platforms cooperate and connect, it will produce the effect of “1 + 1 > 2”. If the “control flow and data flow” enclosed in the platform are extended into a closed loop, a lot of innovation will burst out.
Next, several methods and cases are introduced.
Cross link, with exposure and flow
This is the simplest way of platform collaboration. Each platform should not only fulfill its own mission, but also consider doing something for brother platforms, such as exposure and traffic. Therefore, we add a menu of “Ai product matrix” in the navigation bar of each platform product, listing the logos, names and links of seven or eight products. The data shows that this small menu can bring considerable exposure and transformation to other platforms every day, and the ROI of making this menu is very high.
Reuse of platform capabilities to eliminate waste
In the process of continuous iteration and upgrading of the platform, for a new requirement, we should not do it by ourselves, but first see if other platforms have ready-made capabilities that can be reused, even if it is “saving the country by curve” or “expedient”.
For example, the knowledge update of knowledge mapping platform and the copywriting release of intelligent copywriting platform need to go through the marking and confirmation process. We found that the tagging ability of the tagging platform is enough. So, instead of redeveloping, we quickly solved this problem by connecting platforms.
Feedback and closed loop to achieve common development
If a platform only has one-way output capacity, but does not get feedback from the downstream, and does not form a closed loop, it is not a perfect system and platform.
For example, our annotation platform has marked hundreds of millions of data, which makes the training of various models possible. As the saying goes, there is no intelligence without artificial intelligence.
In this process, the annotation platform only outputs value and helps intellectualization, but it does not benefit from intellectualization.
Later, we considered to form a closed loop of this chain, that is, let the model trained from the marking data feed back to the annotation platform, so as to realize “intelligent auxiliary Annotation”.
In this way, the whole platform is changed from “pure manual annotation” to “intelligent auxiliary Annotation”, which greatly improves the annotation efficiency and reduces the annotation cost.
Precipitation of data assets to create greater value
If a platform has data precipitation, then these data need deep mining, so as to produce more and greater value.
For example, each business is initially connected to the knowledge mapping platform. In order to solve its own business problems, it has to build schema and guide data from scratch. But with the development of the platform, more and more knowledge is accumulated. Then, the subsequent platform can directly benefit from the previous precipitation of knowledge, without having to rebuild itself. This is the value of platform data.
Another example is the annotation data in the annotation platform. After the completion of model training, the life cycle will end. It’s a pity that no one will take care of it.
Now we plan to precipitate and open up the data, so that the data can generate greater value.
First, label data is open to the inside. There is a cold start stage when the business has just been connected to the AI platform, and the most important thing is the marking data. Therefore, we can sort out and open the sea volume annotation data in the annotation platform, so that the business can search in the platform first to see if there is any existing data, and if there is, it can be reused. If not, consider rebuilding the data.
Secondly, label data is open to the outside world. We can open up some data that do not involve privacy and our core technology capabilities, so as to create greater value for the society.
For example, in the “sorting treasure” project of intelligent garbage classification, hundreds of thousands of marked garbage image data are deposited. After we open the API of relevant models, we can open some of the data, which will contribute to the intelligent garbage disposal of the whole society.
Access to the open platform to realize the combination of the strong and the strong
Here, let’s talk about the specific measures of opening up. If you open up to the outside world directly, it will be more troublesome to do, and there are many docking and maintenance things. We should consider connecting our capabilities to existing and large platforms, such as Alipay applet platform / open platform, Ali cloud platform and so on. With the help of these large platforms, many things about customer acquisition, docking and operation and maintenance can be found out.
Here, I would like to share another idea to consider platform collaborative innovation, that is “graphic method and exhaustive method”.
At the beginning, platform collaborative innovation happened sporadically. It was not systematic and systematic to think of one and do one.
Later, in order to exhaust all the possibilities of “connection” and “collaboration”, we drew a big picture and matrix of system collaboration, put all the platforms into it, and comprehensively thought about what is not connected between the platforms and what is the possibility of collaborative innovation.
This method, you can also refer to when doing other work.
5、 Human confrontation in the platform
It is often said that where there are people, there are rivers and lakes. It’s a platform, it’s also a lake.
When people with different roles and demands participate in it, human nature is displayed.
Therefore, we need to think about human affairs, and we need to operate and govern the platform.
1. Misuse of platform
First, correct incorrect usage on the platform.
Why does this happen?
The reason is that although product managers will try their best to prevent most of the mistakes in product design, and there are corresponding rules to inform users in the play of the platform, people will not “abide by the rules” as you imagine, and they will intentionally or unintentionally “use wisely”, “misuse” or even “abuse”.
For example, when I was in charge of the A / b experimental platform last year, we conducted in-depth analysis of all the experiments in the platform, and found many amazing phenomena.
- Hundreds of experiments have only one version: normally, two or more versions are needed for comparative experiments, but many experiments have only one version. One of the big “magical” or “misuse” is that users only use the platform as a gray platform.
- The flow of hundreds of experiments is 0: some users do not have the ability to use the platform, but do their own streaming, which we did not expect.
- Hundreds of experiments run for less than 3 days or more than 30 days: normally, experiments need to run for about a week. But many students run the experiment for one or two days, and push the experiment completely or offline as soon as they see the data change, which is actually unscientific. Some experiments have been running for dozens of days, but the reason is that someone forgot to deal with them. Maybe the experimental scenes no longer exist.
It can be seen that our understanding of the A / b experiment is still not enough, leading to a variety of “strange” usages on the platform. So, we need to do more work in platform training and product design.
In addition to the A / b experiment platform, we also found many problems on our financial knowledge map and other platforms.
We know that in the Schema Specification of knowledge map, the same entity can only have one type.
For example, for “company”, the most common entity type in the financial field, it’s OK to globally define a type such as “company”. Different business domains can have different business scenarios, but they should share one type.
However, the reality is that in order to be simple and easy to control, business students often want to create a type by themselves. As a result, duplicate types like company1 and company2 appear on the platform.
On the graph platform, in addition to schema duplication, data duplication and inconsistency also exist, which need to be managed one by one.
However, the issue of platform governance is both a science and an art – it can neither be laissez faire nor too strict. Especially in the early stage of platform construction, if the restrictions are too rigid, it is difficult for the business side to understand and cooperate, and even lose customers.
Therefore, we should grasp the strength.
2. Abuse and violation
The above mentioned problems of platform governance are not too bad.
Next, I’d like to introduce some “abuse and violation” behaviors that need to be highly valued and seriously dealt with.
They are two real cases in the annotation platform: “task release” and “collusion”.
First, the abuse of the “task release” function.
Considering that there are many changes in the outsourcing tagging personnel, the product manager designs a “task release” button on the tagging page to prevent the task from being stuck in one person’s hands.
However, later, the leaders of the tagging group reported that they “want to cancel this button”, saying that this button was used by many tagging personnel to “pick the job”: when they encountered a difficult tagging problem, they would click “task release” to skip it.
So, we took this function back from the front-line tagging staff and only opened it to the team leader (this problem was discovered when we went to the outsourcing company for field research, which was not expected by the team members before).
The second is illegal behavior, which refers to the collusion of personnel to “work hard”.
For a period of time, the algorithm feedback annotation speed has decreased. We analyzed the following report and found that the labeling speed of multiple tagging personnel in individual group decreased, including those who had done faster before.
After the investigation, it turns out that there are some individual evil horses who are not only lazy themselves, but also abet and collude with others to reduce the speed of marking and “slouch” collectively.
Of course, the most fundamental reason for the problem of “colluding with foreign workers” lies in the performance management scheme of these marked personnel, which used to adopt the monthly salary system instead of the piecework system, with performance bonus but very little.
Recently, we are setting up a special task difficulty grading standard, and improving the overall management plan of outsourcing personnel.
3. It’s too smart. It’s not working
Finally, another very interesting thing.
We know that if a product is not considerate, smart or intelligent enough, users will not like it. On the contrary, if it is too smart, sometimes it will not work.
People are restless and anxious. If they feel “too magical to know what happened inside”, they dare not use it.
For example, in the products of the model service platform, some students designed the function of “model one click deployment”, that is, to automate the complex and tedious feature processing in the process of offline model deployment to online.
However, after several months of development, we found that we could not find a business party at all, because we all said that we did not dare to use it. Finally, this “smart” one click deployment function can only be reluctantly offline.
(it should be noted that it is not that there is a problem with the product direction of “simplified model deployment”, but that the above-mentioned “black box, let the user have no bottom in mind” solution needs more consideration and consideration from the user’s point of view.)
6、 Cross border, cross border, cross border
The so-called cross-border is to break through the original industry practices and conventions, and realize innovation and breakthrough by grafting the ideas and technologies of other industries.
Charlie Munger, the world-famous investor and Warren Buffett’s golden partner, is a man of great wisdom. He highly values cross-border thinking
- You have to think in an interdisciplinary way.
- You must always use all the concepts you can learn from freshman courses in various disciplines.
- If you can master these basic concepts skillfully, your solution to the problem will not be limited.
To do a good job in the design, operation and promotion of technology platform, you also need cross-border thinking and playing methods – for example, you can combine marketing thinking with product and technology cross-border.
The so-called marketing thinking, in short, includes several key points such as “cognitive law, brand system, material carrier and communication path”. First, we should obey people’s cognitive law of new things (simple and intuitive), build a brand identification and memory system (logo and naming), and constantly plan creative activities and materials, And exposure and dissemination in the right place.
Then, for the operation and promotion of the technology platform, we can also use the theories and methods in the above marketing field.
Specifically, we can start from the following aspects:
Platform products need brands
We sorted out the brand identification system of all platforms, and referred to the practice of “Ali zoo”, respectively named Zhizhi financial knowledge mapping platform, whale language NLP platform, tuying financial vision platform, thousand sturgeon search platform, rhinoceros robot platform. The choice of each animal reflects the characteristics of the platform products as far as possible (Picasso intelligent copywriting platform The name of alphaq intelligent annotation platform has a certain degree of recognition, so it has not been modified).
In addition to the name, we awesome UED students have also designed logo with very high degree of isolation and memory. With the name and logo, communication, dissemination and promotion will be much easier.
Product system needs brand
We should not only give each platform a degree of memory and recognition, but also consider how to remember and spread multiple platforms as a whole. Also considering Ali’s martial arts culture, we have packaged the overall brand concept of “Ai Zhongtai Tianlong Babu” to spread eight AI technology platform products. Later, it was found that the influence of “Tianlong Babu” is very high, and many people use “Tianlong Babu” to refer to the AI technology platform family as a whole.
Operational activities need brands
Do operation, do promotion, also need to have a brand system. So, we constructed an image of “Ai commissioner”. All the articles, videos and posters we publish internally are included in this system. For example, all intranet article titles, the beginning and end of the article are in a unified format, adding the name and image of “Ai commissioner”, which is not only convenient to form a unified cognition, but also convenient for everyone to retrieve information in the future.
In addition, in the design of operation activities and materials, there are also brand marketing ideas. No matter how advanced the technology and platform are, interaction, creativity and interest must also be considered in communication.
To this end, we have customized interesting Coke bottles with platform name and slogan, and awarded “appointment letters” to students who have marked product experience.
It can be seen that the integration of marketing, technology and products, and the design of product brand system, operation activities and materials from the perspective of users will achieve good results.
7、 Challenge and growth of platform product manager
After reading this, you may find it interesting and easy to build a platform.
In fact, it’s not. It’s hard for everyone.
For the product managers of the technology platform, they will face all-round challenges of “heart, brain and body”.
In terms of professional skills, in addition to the ability of “demand management, product design, project promotion” required by the product manager position, we also need to “understand technology”. To understand the R & D process, you need to understand the terms and principles of various algorithms and models, because you need to have a dialogue not only with the development team of the platform, but also with the users of the platform, most of whom are also technical classmates.
This is not to ask you to understand technology better than technical students and do technical things instead of technical students, but to understand the essence of technical points, to know what this technology can and can’t do, what is the difference between this technology and other technologies, and what is the development vein of this technology.
When you work hard to figure out these problems, you won’t be too passive.
However, “lack of initiative and weak sense of achievement” still plagues the product managers of the technology platform.
To solve this problem, we can consider the following aspects.
In depth understanding of business needs, improve business sense
The platform ultimately serves the business. No matter how powerful the platform is, it will not help the business and can not be based on it. Therefore, when you have a full grasp of business needs, you can reasonably plan the direction of platform construction, and you will have a sense of achievement.
Consider what unique value you can bring to the team
The success of a project and a platform requires not only professional ability, but also the ability of communication, coordination, promotion, BD and sales. It is no exaggeration to say that to do a good job in products, product managers should not only be product managers, but also play the role of “little CEOs”. When you achieve something through your own efforts, you will be very happy and win the recognition of the team.
Anything has room for innovation and improvement
For annotation platform, you can follow the old way of “manual annotation”, or innovate in the direction of “intelligent aided Annotation”. For the intelligent copywriting platform, you can only rely on the path of algorithm improvement, or you can take the initiative to innovate and product domain knowledge and industry experience to achieve product manager driven. For the acquisition of user feedback and the iterative evolution of products, you can use the traditional way of “face-to-face conversation and questionnaire survey”, or try the new method of “analyzing user logs and using big data + AI”. We should believe that as long as we start from the end, start from business and start from users, we can find opportunities for product innovation.
Always in awe of products and users, do everything seriously
We used to use this sentence to encourage our team members: we should use the mentality of making hundreds of millions of dau products to polish the technology platform of hundreds or thousands of daus. Serious people will not suffer. Every effort you make today will produce value and improve yourself. There is no way to go in vain in life. Every “need” counts.
Finally to the end, here is a summary of the content of the article:
Demand management: “role dislocation” and “selfless state”
The more basic and simple a question is, the more difficult it is to answer and the easier it is to be ignored intentionally or unintentionally. The first step in making a product is to answer these basic questions: who are the users and what are their real needs. To meet the needs of users in depth, we need to ask why and understand the real purpose of users. Forget yourself and think from the perspective of users.
Product design: platform products must also be understood in seconds
If a product is in a mess at a glance and can’t figure out what’s going on, it’s basically a failure. Therefore, we should start from “product framework, concept system, help system, demo experience, interaction unification” and other aspects to realize “second understanding”.
Product verification: if you don’t use it deeply, you can’t do it well
If you want to make a good product, you need to do a good job in product verification. Product managers should try every means to use their products with high frequency and depth. If you have a chance, you have to “do some small business” by yourself, and then you will marvel “ah, there are so many problems.”. In this process, you will have many unexpected gains.
Platform collaboration: connecting, generating value
The value and energy of a single platform is limited. When you break through the boundaries of the platform and create more connections and closed loops, you will create a thriving system and ecology.
Human confrontation in the platform
Where there are people, there is humanity. For the platform with multiple roles, it is necessary to operate, guide and govern, so as to make the whole platform develop smoothly and healthily.
Cross border, cross border, cross border
In the face of complex and changeable environment, it needs diversified talents, complementary skills, and cross-border integration of different industries and fields. Cross border will produce chemical reaction, cross-border will produce innovation.
Challenge and growth of platform product manager
There is no easy word in the dictionary of adults. Only when there are problems and difficulties can the platform, team and individual be promoted and developed. Product manager position is a complex, not a single skill can be based on, product manager students need to constantly meet the challenges, constantly cultivate themselves.
Believe in the power of the platform, believe in the power of the product.
We’re just starting, we’re moving on.
Author: Bai Ling, senior product expert of ant group, has been responsible for the product work of ant AI platform and risk control platform.
This article is the original content of Alibaba cloud and cannot be reproduced without permission