The Internet is the most disruptive engine in modern society, changing everything we do, from shopping to socializing. From shopping to the way we socialize. As the use of social platforms gradually increased, growth hacking also began to spread, and startups began to change the way they think about marketing and growth. Growth hackers who emphasize data, products, and strive for "lean" are challenging the basic assumptions of marketing. In this process, how do we build our own Growth Model? Let’s take a look at this article.1. What is Growth Hacking?An interesting phenomenon: everyone’s understanding of the concept of Growth Hacking is very different, and it is always changing.1.1 The Ever-Changing Growth Hacker
Alex Schultz, Facebook's vice president of user growth, created a website that made paper airplanes in his early years. He put twenty or thirty "paper airplanes" on the website page, and then when he searched through the search engine, he found that the relevance ranking went up. This was the initial SEO attempt.
Dropbox saw rapid growth when it introduced a feature that allowed users to get more storage space by inviting friends.
When Hotmail was first launched, it added a CTA to sign up for use below everyone’s emails, and it saw huge growth in a short period of time.
Later, in addition to using increasingly sophisticated SEO, emails and text messages to attract new users, you can also use social invitation mechanisms, public account dissemination (scan the code to follow) and specific interactive design (text highlighting, sharing, downloading) to drive growth.Now people are starting to use data-driven operations to achieve growth, such as establishing new indicators, building A/B testing frameworks, etc.1.2 The Essence of Growth Hacking: Technology ArbitrageAfter some complicated thinking, I think hacking is a better word I have seen. Any system has rules that are recognized by everyone and rules that actually exist. Hacking is to achieve seemingly impossible goals while recognizing the actual rules.From the perspective of growth hacking, the common approach in the above cases is to gain stronger execution by exploring cutting-edge technologies and tools , and then make full use of it to gain corresponding competitive advantages before the understanding of these tools or rules spreads. This mainly focuses on areas such as customer acquisition, activation, retention, monetization and recommendations. Simply put, growth hacking is about technical arbitrage in the above areas.Some tools and insights of technical arbitrage can hardly reproduce the effect of the past after being widely spread. For example, inviting friends to give space on the online disk, forwarding lottery, CTA, etc. have now become standard features of Internet business operations, but not many of them have grown big because of it.What is really important is to constantly explore and verify new ways of making products and operations. It is precisely because of this that different companies explore different directions at different stages, which leads to the confusion in the definition of growth hacking mentioned at the beginning.1.3 The Flying Geese Model in ExplorationThe more cutting-edge the practice, the higher the investment and the higher the return, because there are fewer competitors and even fewer people who know about it. After a period of time, such as employees leaving their jobs to start new businesses and sharing technology, this practice has gradually spread, been proven to be very effective, and has become popular. As the threshold continues to lower, standardized workflows or tool software will emerge. The important watershed is SaaS tools, which will gradually become the standards and norms of various industries. This is the characteristic of the flying geese model.In my personal opinion, the most cost-effective approach for startups is to achieve a level that can be supported by currently popular SaaS products.2. Growth Hacking and Data DrivenOne of the more important areas of growth hacking and the most worthwhile one is data-driven.Why?First, as summarized in the flying geese model above, the most cost-effective frontier that most companies can keep up with is data-driven. The tools involved are being improved and SaaS-based at a very fast pace in the past two years.Second, data analysis itself is an indicator for measuring the results of all subsequent product improvements or growth hacking. This is in line with what management guru Peter Gluck said: If you can't measure it, you can't improve it.So how to achieve data-driven?Based on previous practical experience, we can divide this process into three parts.
The first step is planning, determining what events you need to collect.
Then comes collection, how to collect data into the corresponding database or product.
The last step is data analysis, which allows us to draw some useful conclusions from the collected data.
3. How to plan data collection?The planning is mainly divided into two parts:The first is which events are mainly collected; the second is which dimensions are used to analyze the collected events.3.1 What kind of events are recorded?The first thing you need to make sure you remember is the events that constitute your key metrics (these key metrics are closely related to your product), such as registration, activation, purchase, etc.Next, break down the things that support these key indicators. For example, registration includes: opening the registration page, entering the account, password and verification code, or using other social accounts to log in.During this process you will find that some events have an obvious order and some users will give up on certain events. This process is called funnel analysis. The various main steps that make up the funnel analysis can also be recorded in this way as an extension and improvement of this key indicator.3.2 What dimension to choose?The most important principle is to record those dimensions that are likely to make a difference.For example, if I place a certain function in different places on the page, it will have different effects, so I need to record their page attributes as a dimension. For example, on an e-commerce website, the goods sold can be recorded in multiple dimensions such as number, price, and place of origin.For startups today, the most cost-effective approach is to use SaaS services and complete the collection of raw data through third-party SDKs.4. How to collect data?Traditional data collection is a very resource-intensive process, which may take up 1/3 to 1/2 of our time.4.1 What should I pay attention to when collecting data?First of all, in the early stages of a product, the needs of the “statistical layer” should be placed directly into the product planning and iteration structure. Advance planning can prevent a lot of hassles and reconstructions later. Because later you will find that the collection requirements are somewhat contradictory to the requirements of program reuse.Secondly, when implementing the product automation testing process, try to include the statistical return results of the statistical data. In the long run, it can help us save a lot of technical problems.Finally, if you are collecting data for the first time, try to start with the SDK that comes with the SaaS tool, which can help you save a lot of small details.4.2 How to choose a data collection tool?The first recommendation is Mixpanel, which has more comprehensive event collection than Google Analytics and is very user-friendly.There is also an overseas tool called segment, which also embeds data once and then converts the collected data into other solutions that you can accept. You just need to control it through the background switch. It’s also a very easy tool to get started with.The above solutions can serve as a starting point for you to understand the more popular SaaS products.4.3 How to process data?If you only use the most cost-effective SaaS tools, you may not need to worry about the data processing process; but it may not meet your data requirements.You may have heard of ETL (extract, transform, and load). Many times we have to do some data processing work of our own. This is mainly due to performance considerations, the data format captured by the SDK of the SaaS product is inconsistent with what we need, or the need for post-processing.5. How to conduct data analysis?Plan and collect the required data, and then start data analysis. The main purpose is to view the correlation between data by separating dimensions.5.1 Two major goals of data analysisThe first is to provide reports and dashboards. According to the business needs of each colleague in the company, produce the results he needs; at this time, you need to focus on key indicators and generate reports or dashboards for visual analysis. When making indicators, try not to use historical data as inventory. For example, how many registered users has the company had since its founding? This number may seem impressive, but it is not comparable. We are more concerned about the trend of variables in each period, comparing the data of this period with the previous period to identify problems.Then comes hypothesis testing to guide the business. Once our reports or dashboards reflect the entire business development situation, we will have better experience in using the tools and can make some assumptions based on business experience to guide your products and operations. Then watch the data trends to verify whether your hypothesis is valid; thereby continuously optimizing your products and operations.5.2 Common ways to start data analysisThe report should cover key rows (registration, activation, purchase payment, etc.) and key paths (conversion, funnel, retention, etc.), with a focus on "anger point" behaviors.What are user “anger points”? If a user performs operations on the same element on a web page more than 5 times in a row within 10 seconds, it is often the case that he keeps clicking the button with the mouse, which is a "anger point" for the user.There are two possibilities in this situation. One is that the product is really cheap, that is, one action requires pressing the same button 8 times, and those 8 times are switching to different states. Another situation is that the product does not perform in line with the user's cognition and expectations. In this case, the user infers that there is a network problem or the product is stuck. In some cases, he does not really understand the logic of your product interaction. It is very likely that there are some confusing aspects in your design, or some prompts are not in place. In fact, it is worth going deeper into this process.5.3 Hypothesis TestingA relatively standard step for hypothesis testing is shown in the figure above. We observed a phenomenon and, based on our business experience, speculated on what mechanism caused this phenomenon, or what mechanism was related to this phenomenon. If the assumption is true, we can better optimize our products by promoting this mechanism. If not, the product or operation may deteriorate or remain unchanged. We analyze whether the hypothesis is valid by continuously observing the data before and after and comparing the differences in data changes.5.4 Teambition Case StudyThe picture above is the interface of Teambition's online project creation. We conducted a funnel analysis specifically for this project and observed that some users exited after operating at [Project Type] and had not completed the project creation. Can we improve this situation under such circumstances? Based on past experience, too many options on the interface tend to distract users and affect core operations. The core of creating a project is to let users create a project, rather than having to choose a cover.So how do we verify this? If we can omit non-critical factors and let users customize non-critical information after they have created the project, will there be any improvement?Based on the above assumptions, we have simplified the interface for creating a project. The left side of the picture above shows how to create an enterprise project, and the right side shows how to create a personal project. It is very simple and convenient and can be completed in less than a minute.Through the subsequent funnel analysis, we found that the conversion rate was indeed greatly improved.6. Summary: Build your own Growth Model6.1 Pirate Model-AARRRAccumulate a large number of such insights to support the update and iteration of your product. The verified hypothesis or the summary of all your decisions becomes what everyone calls the growth model, which is also a very popular word now, the Pirate Model - AARRR, which means customer acquisition, activation, retention, monetization and recommendation if you break down each first letter.
Acquisition is to obtain website traffic mainly from website channels.
Activation allows users to truly experience the iterative and innovative value of your product and be willing to come back to use it and understand your product.
Retention means long-term positioning of your product, rather than just registering and leaving. This is actually something that people in the industry are starting to pay more and more attention to. Because the era of low-cost traffic has basically ended, and in the process of products with normal revenue, the conversion rate of purchases or the efficiency of converting traffic to sales, etc.
Referral allows users to introduce the product to their friends, colleagues, and other acquaintances.
In fact, for each of the above steps, companies have their own goals and trade-offs. During the user registration process, should we let users fill in as little information as possible to increase the registration conversion rate, or should we let users complete the information to facilitate follow-up by the sales department? In fact, companies can formulate corresponding strategies based on their own development stage; for example, start-ups may pay more attention to conversion rates.Combining every point in the AARRR model that your company's actual business concerns is called your own Growth Model. Nowadays, being able to build and iterate the Growth model relatively well is a relatively scarce and valued skill in the industry. Another thing is to be able to test and improve the Growth model in real time after the iteration is completed.6.2 Build a data-driven teamThe last point is to cultivate the team’s data awareness and tool usage capabilities. This is very important. Internal awareness plus their ability to manipulate data themselves to get the conclusions they want is, in a sense, the leverage of this advanced technology within your organization.There are two main factors driving image data:The first is that the construction speed of enterprise data infrastructure design is very slow. The traditional manual tracking process takes 1-2 months. By the time a new version is released, the project is basically aborted.The second is that data users and collectors are staggered. Self-service queries are encouraged. Business, operations, and products frequently use data and need to use data to support decision-making; however, data engineers are required to collect and process data. By the time data engineers run out the queued data reports, the business side has missed the best time to make decisions. This will also severely limit the use of data technology and corporate decision-making.In this case, it is recommended to use a third-party SaaS SDK to simplify data tracking, and try to encourage self-service query and even self-service data tracking to facilitate business personnel to complete data query and data-driven decision-making process by themselves: this is a very important thing.
ASO optimization service click link: ASO optimization service introduction
APP Top Promotion (www.opp2.com) is the top mobile APP promotion platform in China. Welcome to follow our official WeChat public account: appganhuo
[Scan the top APP promotion WeChat QR code to get more dry goods and explosive materials]
This article was written by @GrowingIO and published by (APP Top Promotion). Reprinting this article requires the consent of Top Promotion , and please attach the link to this article!
