What are the common big data analytics models?

Listen to this article

For example, which data models in data analysis can be used directly, especially for some Internet platform products. There are 11 general data analysis models that can be used directly, including event analysis, attribute analysis, channel analysis, session analysis, retention analysis, attribution analysis, heat map analysis, distribution analysis, interval analysis, path analysis, and funnel analysis.

The content of the article:

• Event Analysis
• Attribute Analysis
• Channel Analysis
• Session Analysis
• Retention Analysis
• Attribution analysis
• Heat map analysis
• Distribution Analysis
• Interval Analysis
• Path Analysis
• Funnel Analysis

Event Analysis

Events refer to the behavior of users in apps, websites and other applications, i.e. who, when, where, how and what they did. The event analysis model is mainly used to analyze the user's behavior in the application, such as opening the application, registering, logging in, and paying for the order. User behavior is measured by base metrics such as number of activated users, number of triggers, and access duration, and supports metric calculation to build complex metrics to measure business process.

So what problems can the event analysis model solve? For example:

1. Track the daily number of users, visits and product use time, has the trend changed? What factors caused the change?
2. What is the difference between the distribution of household appliances purchased by users in St. Petersburg and Moscow?
3. Today, the product launched a topic, how is the user participation in each period?
4. What is the number of paying users and ARPU in the last six months?

The event analytics model can track user behavior across platforms in real time, attribute indicator changes across dimensions, and combine custom indicators into new indicators to achieve more powerful analytics. Analysys Ark smart analytics products support dimension filtering and department conditions, and support group comparison by analyzing user groups.

Attribute Analysis

Attribute analysis is based on an analysis of the proportions of user-defined attributes or preset attributes. It can calculate the proportions of metric attributes such as the number of users across various attributes, and then get the preliminary conclusions of the analysis. For example, by analyzing the proportion of users with gender characteristics, we can quickly get statistical results on the number of users of different genders.

Using attribute analysis, you can quickly see the distribution of users across different attributes, which is useful for calculating the total number of users with different characteristics. In the process of using attributive analysis, it is necessary to choose a reasonable method of measurement. Commonly used measurement methods include: number of users, multiplicity, sum, maximum, minimum, average, etc.

For example: the indicator we select is "Average Cumulative Consumption", the dimension is "Membership Level", and the user selects "All Users", then the result we get is "What is the Average Consumption of All Users with different membership levels.

Similar to the Event Analysis Model, the Attribute Analysis Model can perform multi-dimensional and multi-user comparisons and display statistical results in various charts. The tag feature scenario can also perform statistical comparison of tags across versions.

Channel Analysis

Channels, i.e. various points of contact between businesses (products) and users, such as search engines, social networks, advertising platforms, offline meetings, etc.

The channel analysis model is used to analyze the sources of user visits (including visitors) and evaluate the quality of the channel using key metrics such as the number of visiting users, the number of visits, the duration of the visit, and the bounce rate. It also supports custom conversion goals to measure the conversion effect of channels.

So what problems can channel analysis solve? For example:

1. What about the number of visitors and page views of each channel in real time?
2. What about the number of user registrations attracted by each channel?
3. Social media, search engines, external links… Which channel has the highest retention rate?
4. Wechat source users are more concentrated in the official account or in the circle of friends?
5. Which search terms drive the most traffic and convert well?

The channel analysis model can clearly represent the performance of each channel by identifying key metrics and conversion rates, selecting the analysis platform and channel parameters to evaluate the actual output effect of various channels, and finally selecting a combination of high-quality channels to improve the overall ROI.

Session Analysis

A session, i.e. a session, refers to a set of sequence of user actions that occur on a website/H5/widget/app over a period of time. For example, a session may contain multiple page views, interaction events, and so on. Sessions have temporal attributes. Sessions of different durations can be created according to different cutting rules.

The session analysis model contains many metrics to measure the quality of session access, including number of visits, number of visits per capita, total visit duration, duration per visit, depth per visit, bounce rate, bounce rate, number of exits, exit rate, and duration of visit per visit. per capita, total time on page, average time on page.

Unlike event analysis, session analysis additionally supports subdivisions of some dimensions to meet the needs of session analysis in certain scenarios, including:

1. Channel Source Grouping: used to distinguish the channel sources of each visit, applicable to Web/H5/Mini only;
2. Number of pages viewed: with a step of 5 as an interval, the distribution of the number of pages viewed each time is calculated;
3. Landing page: used to distinguish the landing page of each visit, and the quality of the visit to different landing pages can be assessed;
4. Exit Page: used to differentiate the exit page on each visit, evaluate the exit situation on different pages, and find a page with a high exit rate for optimization;
5. Access duration: 0-3 seconds, 3-10 seconds, 10-30 seconds, 30-60 seconds, 1-3 minutes, 3-10 minutes, 10-30 minutes, 30-60 minutes, 1 hour or more. Divide and calculate the time distribution of each visit.

Similar to event analysis, session analysis also supports multiple indicators, multiple dimensions, and multiple filter conditions, as well as horizontal comparison across multiple user groups. At the same time, when analyzing a session, it also supports statistical analysis according to three different levels of detail: day, week, and month. Users can select the appropriate granularity for analysis according to the time interval of the query data.

Retention Analysis

Retention means that users have used apps, websites, and other applications and continue to use them after some time has passed.

The Retention Analysis Model is a method to measure user status/engagement beyond downloads, DAUs and other metrics to gain a deep understanding of user retention and churn, identify key factors influencing sustainable product growth and make market decisions, improve product, increase user count . cost, etc.

So what problems can a retention analysis model solve? For example:

1. Was there an iteration of the product last month, how to evaluate its effect? Is the behavior expected by the product manager?
2. As a social app, is there a difference between users who don't add friends after signing up and those who add 10 friends?
3. Short term retention is low, long term retention must be bad?
4. Two promotion channels attract different users. Which channel users are more likely to be valuable users?
5. What is the proportion of users who signed up in the last 30 days who didn't return within a half month?

The retention analysis model supports conditional filtering and benchmarking across multiple populations, as well as random sampling and full data calculation. At the same time, we can also use retention analysis to determine whether new users are ready to return to use your feature in a few days, weeks, and months, and we can also customize the start and end behavior for feature retention analysis.

Retention is calculated based on the start time of the behavior of a group of users and describes whether the desired behavior occurs after a certain period of time in the cohort where the particular behavior occurred. Both the initial behavior and the subsequent behavior can be any event or a specific event.

Different storage conditions can be set for analysis in different analysis scenarios:

1. The initial behavior and subsequent behavior are set to be the same, and repeated occurrences of different features are compared to ascertain the user's commitment to using the different features;
2. The initial behavior is the same, and different subsequent behaviors are given to compare whether the same optimization affects other features differently;
3. The subsequent steps are the same, specify different initial behaviors, compare, and find the impact of different product practices and features on key business goals.

Attribution analysis

When performing operational activities, we may place event content at multiple operational positions within the product in an attempt to capture user attention, direct traffic flow and user behavior, and facilitate interaction between users and products, resulting in conversion. In addition, users themselves can also receive information through touchpoints such as search and content recommendations. These touchpoints also play an important role in whether users can achieve conversions.

In other words, in the user conversion journey, many touchpoints on the site are involved in persuading and guiding users, influencing their final decision. Then, comparing different user touchpoints to see how well they contributed to the key metrics, do they all have the great conversion opportunities that operators expect, or are they undervalued? In the subsequent operation, how to adjust the resource input weight distribution of each work item?

For the above problems, attribution analysis provides an intuitive measure, Conversion Contribution, which is primarily used to measure and evaluate the contribution of user touch points on a site to the overall conversion goal (e.g. total order value), which can be extremely straightforward. Measure the conversion effect and value contribution of each workplace and touchpoint. There are five common attribution analysis models:

1. First touch attribution: 100% a conversion value is assigned to an event that should be attributed for the first interaction;
2. Last touch attribution: 100 % conversion value refers to the last interaction event to be attributed;
3. Linear attribution: evenly distributes the value of the conversion to all attribute events along the conversion path;
4. Location Attribution: Allocate conversion value according to the position of the event that needs to be attributed along the conversion path. As a rule, the events of the first and last interaction are 40 % each, and the events of intermediate interaction points are equally divided into the remaining 20%;
5. Temporal Decay Attribution: Distribute conversion value in chronological order of the events you want to attribute. The closer to the target event, the greater the contribution and the more will be assigned to the attributed event.

With the Analysys Ark Attribution Analysis Model, you only need to set up five simple steps (defining target events, touchpoint events, choosing an attribution model, defining a window period, and choosing a query time range), you can intuitively see how each touchpoint contributes to the overall metric conversion.

Heat map analysis

The heat map analysis model can use the heat map to visually display the user's click and scroll behavior on the website, H5 page and app, to help the product and operation staff understand the user's click preferences, help in page design optimization, content customization, etc. d.

There are four common types of heat maps:

1. The click location heat map is used to display the location of all clicks on the users website. The more clicks aggregated, the brighter the color. Often used for landing page analysis: is the CTA content being clicked on? Are there important buttons or elements that are frequently clicked and placed in places that only a few users can access? Are there any images or text that users click on but don't actually have a link?
2. Element click map to show clicks on interacting elements. For analysis: which specific elements generate how many clicks? What percentage of total page clicks are taken? Are there bugs that don't meet our expectations?
3. View a depth line showing the retention rate of users arriving in a specific area. The lower the percentage, the fewer users will be able to see the location. Often used to find the best CTA placement and monitor content marketing conversions.
4. An attention heatmap showing how long users stay in a certain area, the longer the stay time, the brighter the color of the area. Usually used for analysis: to understand what content on the page attracts visitors, and what content is considered important, but ignored by users? Is there any content that has been carefully read by users and placed too far down?

Different types of heat maps have their own advantages and disadvantages, such as the click location heat map. The disadvantage is that the amount of reported data will increase, but it can be very intuitive to qualitatively analyze the research needs of users and find a large number of unexpected clicks on non-interactive elements. Clicking on an element's heatmap filters out some of the content that can't be clicked. Clickable elements can be analyzed quantitatively, but this is not intuitive enough.

We can choose different suitable types in different scenarios. Currently, Analysys Ark already supports click position heatmap, click element heat map, web side view depth line, click position heat map and app side click element heat map.

Distribution Analysis

Distribution analysis can basically enable data decomposition after “dimension indexing”, split the original dimension according to a certain numerical range, and then analyze the distribution of each measurement range, which is very common in the following analysis scenarios: sum distribution, analysis of the distribution of a certain time period type of special event, analysis of the number of occurrences of a certain type of special event, and analysis of the age distribution of users who initiated a certain type of event.

Distribution analysis is mainly focused on two types of attributes, numeric type and date type, such as count, age, time, and frequency. Therefore, when the data uploaded by the user includes these two types of attributes, then in daily analysis, distributional analysis can be used to solve some specific tasks. Common indicators include: frequency distribution of X events, distribution of active periods of X events, distribution of active days of X events, and distribution of sum/average/capita values of X events and Y attributes.

Interval Analysis

Interval analysis is primarily used to calculate the amount of time between when a user fires a specified start event and completes a specified target event. That is, it basically provides statistics on related metrics in terms of time and step length from the start event to the conversion goal, so that people who pay attention to the conversion can observe the situation of the conversion process on these metrics.

There are many use cases for interval analysis: it can be used to calculate the login time interval and product repurchase cycle, as an analysis tool to measure user activity and stickiness, and as an adjunct to conversion funnel analysis. The Duration indicator is used to measure the conversion performance of a specific conversion path over time.

Interval analysis is an addition to the conversion funnel when used to measure conversion performance. However, both have their own idiosyncrasies: interval analysis focuses on the effectiveness of users' time to complete a conversion, while conversion funnels focus on conversion outcome metrics as well as conversions and the loss of each link in the conversion process.

While paying attention to the results of the conversion and the parameters that affect the results, we also need to pay attention to the performance indicators in the conversion process. For example, for financial and wealth management apps, from the landing page to the first deposit, many conversion links are involved, in addition to the final conversion. In addition to speed, you need to pay attention to the conversion efficiency between the main stages, especially registration, card linking and other links.

With interval analysis, we can observe the time interval of the distribution of users executing two given events. Combined with other analysis models, we can gain insight into the laws behind user behavior in order to study and improve user experience, activity, product conversion rate, and product value.

Path Analysis

Paths are the behavioral trajectories that users use in an application. In the process of working with a product, whether it is a product, an operations team, or a marketing team, it is hoped that the user behavior path can be clearly understood in order to test the operation idea, guide the iterative product optimization, and achieve the ultimate goal of user growth and conversion.

When there is a clear conversion path, it is easier to track conversion rates by building a funnel ahead of time. However, in many cases, despite having an ultimate conversion goal, users have multiple paths to achieve that goal. In this case, an intelligent path analysis model is required.

With the Analysys Ark Path Mining Model, you can open the black box of user behavior, you can explore the conversion goal source path, and you can visualize all user paths and proportions.

The path mining model can solve the following problems:

1. From which path do users mainly form a payment conversion?
2. What is the actual direction after the user leaves the expected path?
3. What are the differences in user behavior paths for different characteristics?

Funnel Analysis

Funnel analysis is a method of analyzing the conversion effect of users through a series of steps when they use a business. The Analysys Ark funnel analysis model can flexibly customize the conversion process across multiple steps, find key loss relationships and influencing factors, and then analyze user behavior for targeted optimization actions.

So what exactly can funnel analysis solve? For example:

1. The official site has a lot of traffic, but few registered users. What part of the process went wrong?
2. What is the overall conversion of users from "register - link a card - place an order - pay for the order"?
3. What are the differences in custom payment conversion rates across regions?
4. Two promotion channels bring different users, which channel has the highest registration conversion?
5. Last week, the registration link was optimized, did the conversion dynamics improve?

Ideally, users should follow the product design path to the end goal event, but the reality is that user behavior paths are diverse. By setting up key business paths through hidden events, we can analyze the situation with conversions and losses in various business scenarios. Not only do we locate potential product issues, but we also find the lost users in each link and then target them to drive the conversion.

Reading this article:

• What is the difference between sem and seo?
• Behavioral website ranking factors - how to increase?

Thanks for reading: SEO HELPER | NICOLA.TOP