With the drying up of Internet traffic sources, successful user payment conversion is more essential than ever to monetizing products.
This requires meticulously data-driven operations. Current business growth solutions mine historical user behavior for valuable clues, for example, performing attribution to grasp user payment trends. Still, it would be much better to be able to proactively predict user attributes and behavioral preferences in advance, and then use these predictions to make optimal decisions that could increase payment conversion. That's where HUAWEI Prediction enters the picture…
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
1. Which types of users tend to make payments?
User payments don't usually come out of nowhere, as they are usually preceded by a series of actions, such as viewing an ad, experiencing a product, or comparative shopping activities. Therefore, the goal of payment prediction is to find those users who demonstrate a high payment potential, among all recently active users. In-app purchase events are the direct result of payment behavior.
HUAWEI Prediction's payment prediction task trains a model that takes user payment data from the most recent two weeks into account, to predict the probability that app users from the previous week will make a payment over the next week. Naturally, since the basic data and model training are highly dependent on the in-app purchase events reported by your app, predictions are only accurate when there is sufficient, high-quality reported data.
2. What are the characteristics of users who demonstrate high payment potential?
Users can perform transactions for any number of reasons, for example, being drawn to a product description, or for believing that a product offers excellent value. To perform targeted marketing and operations strategies, it's important to be able to identify common attributes and behavioral characteristics of users with high payment potential.
For your app's prediction analysis, you can add metrics, such as user acquisition, total page views, and time of last use, to analyze your app's audience metrics in a thoroughly in-depth manner, as seen below.
In the example below, the high payment potential users clearly share common attributes and behavioral characteristics. Users in this audience tend to be those who have frequently used the app for a relatively lengthy period of time, and have also recently used the app on a frequent basis. From this information, we can surmise that such users have a high degree of familiarity with the product, and a strong desire to purchase, but are deterred for various reasons, such as that the price is slightly too high, or that the product is not their primary necessity. They are thus, in the process of waiting for a discount, or still comparing the product against other similar products.
In this case, a promotion such as a time-limited discount, or other form of incentive, will attract a large number of users who are still considering the purchase. However, if the promotion is pushed to all app users, the operations costs would also soar, and may even exceed the amount earned during the promotion. Next, we'll walk you through how HUAWEI Prediction can help you limit the promotion to target users alone, and keep costs under control.
3. How can I promote user payment conversions?
Now that established the high payment potential audience profile, we'll need to maximize the value of this group. This can be achieved by implementing a successful promotion.
Audiences identified by HUAWEI Prediction can be applied to a range of other AppGallery Connect services. For example, we can use Remote Configuration to carry out the promotion.
To do so, go to the Condition management page of Remote Configuration, and add the Prediction filter, as well as all corresponding parameters for the promotion. This ensures that you will send the promotion only to those users in the predicted audience, thereby minimizing operations costs.
After the promotion is complete, you can refer to the payment details for confirmation that it worked as intended, and proved more effective than previous promotions.
You can also apply the predicted audience to other services, for example, by utilizing A/B Testing to test how the predicted audience responds to different payment growth solutions.
To learn more about HUAWEI Prediction, feel free to check out this document.
Related
Making Mobility Safer, Smarter, and More Accurate with AppGallery
At the beginning of this year, the urban mobility industry was booming as more people looked to convenient and intelligent transport services to get around. Investment was accelerating, new trends and technologies such as better connectivity, automation, and artificial intelligence were driving innovation and disruption, and transport operators were diversifying their service portfolio and offerings.
Fast forward to September and the transport and navigation landscape paints a completely different picture against the backdrop of new urban norms. Global lockdowns, restrictions on movement, and more flexible working arrangements temporarily brought commuting and leisure travel to a halt around the world.
However, this didn’t spell the end for the urban mobility sector – transportation is still much needed by commuters and travellers – there has simply been a shift in consumer demands and behaviours as people get used to a new way of doing things. In this sense, it is crucial that transport providers keep pace with these changes by adapting their operating models.
Earlier this month, Huawei’s annual developer conference, HUAWEI DEVELOPER CONFERENCE 2020, looked at how AppGallery and Huawei are helping developers in the transport and navigation sector address these challenges and succeed in a continually changing industry, and we’ve summarized some of the key trends and insights from the conference.
Diversifying Service Offerings to Match Changing Demands
While urban mobility activity has dipped in recent months, there has been a spike in demand for other transport services such as food delivery and last-mile transportation. With demand for urban mobility not expecting to recover quickly, mobility-as-a-service (MaaS) providers are expanding their service offerings to meet these demands and improve their business resilience. Companies such as Uber and Cabify are offering these alternative services such as food and parcel deliveries, while providers such as Bolt have expanded further to include micro mobility solutions.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
While these new services represent exciting opportunities, MaaS providers will need to enhance their capabilities to address the pain points of the additional services. For instance, delivery solution providers need to introduce capabilities such as real-time location updates and trip planning to
ensure smoother operational efficiency.
Huawei’s developer toolkit, Huawei Mobile Services (HMS) Core, provides several map-based capabilities that are vital to transport operators looking to diversify. Capabilities such as Location Kit, Nearby Kit, Map Kit, Super GPS, ML (machine learning) Kit, Scan Kit, and Awareness Kit facilitate more functional services for operators and better experiences for users.
The Map Kit, for instance, comes with real-time global road condition reporting and extremely accurate route planning features that are crucial for parcel delivery solution providers. Huawei’s Super GPS supports roadside identification, making it twice as accurate as other providers. This
precise positioning enables people to meet their driver even when surrounded by high and dense buildings or faced with difficult road layouts. Meanwhile, Huawei’s machine learning capabilities create a more informative and interactive transport experience through its language translation
service in 19 different languages, facilitating communication between travellers and drivers andallowing operators to expand their service to more commuters.
Additionally, Huawei provides comprehensive one-stop operational support that developers can leverage to address integration as well as implementation issues, achieving a faster development process.
Addressing Travel Safety Concerns
Commuters now have a heightened travel safety consciousness and are actively minimising contact with others and avoiding crowded transit such as public transport. However, demand for shared mobility remains for people who still must travel such as on-site workers. This group of commuters are looking for transport providers that can provide reduced contact, as well as shorter waiting andtravel time.
While hygiene concerns can be tackled through precautionary measures, other concerns require a more technological approach through the optimisation of existing mobility solutions. With Huawei’s end-to-end travel solution, developers can incorporate a suite of HMS capabilities to
deliver more intelligent and accurate transport services, providing customers with peace of mind through safer and more convenient journeys.
For example, HMS Core’s location capabilities provide a precise integrated positioning method, enabling passengers to share their real-time location information and give timely warnings if the car’s position deviates from the planned route. Not only does this give passengers confidence and help build trust between operators and users, it also means travellers can automatically notify friends and family if there are any issues or delays. In addition, HMS Core enables the historical driving track to be recorded and retrieved later, so any issues that only become apparent after the journey can be highlighted.
Unlocking New Opportunities
With consumers trending towards mobile technology to find ways to move around, transport operators are increasingly prioritising mobile platforms and app marketplaces to acquire new users, unlock commercial opportunities, and secure a much-needed edge to beat growing industry competition. In this light, Huawei and AppGallery help global partners strengthen brand exposure to reach more users via multiple channels. It does this through a range of different measures such as providing various operation resources and advertising opportunities; offering exclusive online and offline promotional campaigns to strengthen brand exposure and unlock access to new markets; inclusion in recommendation cards and promotion on AppGallery to drive consumer traffic; access to conferences, roadshows and on-ground promotions; as well as industry events such as HUAWEI DEVELOPER CONFERENCE 2020 and Huawei Developer Webinar.
Bolt, the ride-sharing app with 30 million users in 35 markets across Europe and Africa, is a shining example of how partners can reap commercial benefits by utilising the extensive resources AppGallery provides – the app saw an increase of 136 times in European and African downloads
from week one through to week 13 after it offered exclusive gift packages to AppGallery users. Meanwhile, the Dutch navigation app developer TomTom partnered with Huawei for the launch of TomTom AmiGO on AppGallery. The joint marketing campaign included marketing resources such as social promotion and operation resources, and within two months European downloads of the
app increased 22-fold.
The transport and navigation industry is facing a series of unprecedented challenges as transport needs continue to change, meaning MaaS providers need to remain nimble through constant innovation. While these changes will not happen overnight, they can be accelerated by leveraging capabilities available through HMS Core and utilising the support offered by Huawei and AppGallery.
Smart, data-driven technologies are creating a world of frictionless communication, with huge benefits for business. Companies can provide tailored products that can reach the long-tail market thanks to a clearer, more precise understanding of customer habits, needs, and wants. Transparent, real-time information channels will eliminate errors and misunderstandings between consumers and service providers, ramping up customer satisfaction and loyalty. And AI-powered translation devices will power borderless business, helping companies go global as language barriers fall.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Understanding Customers
Companies are now using intelligent technologies to help them design innovative business models. Service providers have always had access to surface-level data, like customer buying habits, but soon they will be able to dig deeper into information like user emotions and personality. They can find online and offline information about a customer's career, interests, preferences, and social attitudes, and this vivid profile of the customer provides high-quality input data for AI algorithms, which can uncover the customer's real and hidden needs.
In the manufacturing industry, as vendors gain a deeper understanding of the real needs of end users. they can develop products that better satisfy the needs of the brand-name firms that sell their products and increase their own value within the supply chain by identifying and recommending new business opportunities.
Inclusive Communication
StorySign is an app that helps deaf children to read using AI technologies such as image recognition and optical character recognition. When a user scans a page from a story book, the StorySign app shows a cartoon signer who signs the words. Currently, StorySign can translate texts into 10 different sign languages: British (BSL), Irish (ISL), Dutch (NGT), Flemish (VGT), Italian (LSI), Spanish/Catalan (LSE & LSC), French (LSF), Portuguese (LGP), Swiss German (DSGS), and German (DGS). More languages will be added in the future. AI devices and software will allow the speech or hearing impaired to engage and contribute on an equal footing.
Understanding Product/Service Providers
Conflicts between patients and doctors eat up 6% of hospital efficiency every year, and 75% of doctors and nurses report that they have been subject to physical or verbal attacks caused by problems in communication. An AI assistant could interpret a doctor's prescription in layman's terms to make it easier for patients to understand. This will allow patients to feel more certainty about the health issues they face, their risks, treatment plans, and the expected effects. Doctor-patient interaction can be smoother and calmer, without the stress of complex terminology.
After receiving a prescription, AI tools can also help patients understand what the doctor has given them, confirm that it is the right drug for them, check for any alternative therapies, and calculate the correct dosage for their current condition. For the elderly and other patients who need support, these tools can make sure that drugs are taken on time, in the correct dose, without any confusion.
Borderless Communication
AI-enabled translation devices can help people speaking different languages communicate very effectively. The combination of human plus AI translators makes for a much more effective team: Together, they can achieve 95% accuracy with 0% omissions. The ability to communicate across language barriers is vital for big companies and organizations that now work around the world, with people speaking every different language on the planet. Translation devices with inbuilt specialist domain knowledge will be a boon that breaks down the barriers to communication in commerce, charity, government, and academic settings, enabling everyone to forget the friction and focus on their work.
For details about Huawei developers and HMS, visit the website.
HUAWEI Developer Forum | HUAWEI Developer
forums.developer.huawei.com
Very insightful.
"The purpose of business is to create and keep a customer." - Peter Drucker
As traffic sources dry up, enterprises have found it increasingly difficult to acquire new users, and face potentially enormous losses caused by user churn. Therefore, determining which users are at risk of churn in advance, and taking proactive steps to retain them, is critical to achieving success on the market.
Products that feature a high user retention rate, not only benefit from the increased revenue, but are also easier to promote, as long-term users are more likely to recommend the product to other users.
HUAWEI Prediction offers a groundbreaking new model for user retention that can give your app the fortitude to withstand even the most volatile market.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
With regard to predicting user churn, four questions immediately spring to mind:
How can I determine whether a user is at risk of churning?
Are the predictions accurate?
What are the characteristics of users who are at risk of churning?
What can the predictions be used for?
Let's address these questions one at a time.
1. How can I determine whether a user is at risk of churning?
First, we need to determine how long it takes for an inactive user to be regarded as a churned user to be won back. Determining the correct length of time is critical. If the time period is too short, the predicted audience will inevitably include a number of retained users as well, which can cause the cost of winning this group of users back to soar unnecessarily. But if the time period is too long, the user group will include those users who have been permanently lost, for whom there is no possibility of winning back.
The churn prediction task preset by HUAWEI Prediction obtains insight into industry-specific user lifecycle attributes, by taking a massive amount of experimental data into account. The task uses the active user data from the previous two weeks to train the model, which then predicts the probability that active users of the app from the previous week will be lost over the next week. Users who are inactive in the next week, or those who uninstall the app, are regarded as churned users.
2. Are the predictions accurate?
The audience of churned users will change with time, as users' behavior tends to vary on a daily basis. That's exactly why predictions are based on the most recent data. HUAWEI Prediction provides two indicators to evaluate the accuracy of its predictions, the true positive rate and false positive rate. In a churn prediction task, the true positive rate refers to the ratio between the number of users correctly predicted to have been churned, to the total number of churned users; the false positive rate refers to the ratio between the number of users falsely predicted to be churned, to the total number of users who are not churned. You've likely noticed that a higher true positive rate indicates a lower false positive rate, as well as a higher prediction accuracy.
You can also check the predictions from the most recent 7 days to evaluate the quality of the data reported by your app. Sufficient and quality data are essential for accurate predictions, as the prediction model is trained according to historical data.
3. What are the characteristics of users who are at risk of churning?
The service has preset three probability ranges: high, medium, and low, and also gives you the freedom to customize a probability range according to your needs. Thus, each prediction will generate four audiences. You can then perform in-depth analysis into user attributes, user behavior, and other audience characteristics.
Let's use an audience with a high churn probability as an example. The prediction details page displays the number of users in this audience. You can customize the metrics of interest by selecting user attributes and behavior cards. For example, for predicted high-probability churned users, you can select the User acquisition and Total sessions cards. According to the customized metrics, most users in this audience have a lengthy usage history, and a high number of total sessions.
In this case, we can infer that the users are likely to churn, not because they have not been targeted by your activities, or think that the product design is poor, but rather, because they are already in the inactive stage of the user lifecycle. Therefore, targeted operations should be planned with the goal of reactivating them.
This leads to the question of "how"?
4. What can the predictions be used for?
You're likely familiar with some commonly used user activation and winback activities, for example, discounts and message pushing. Most of the time these activities target users who have already churned, but the rate of success can be disheartening.
The following is a real case study. A game had been hampered by a low user retention rate, and had tried a number of marketing activities, including time-limited gift packages, with the goal of winning back users. However, the data revealed that almost all users who claimed the packages were active users, rather than the real target of the promotion, which was churned users. This imposed high costs with little reward.
The game's operations team then turned to HUAWEI Prediction and applied the predicted audience in Remote Configuration. They sent the gift package to only the identified audience, via Remote Configuration. A remarkable 80% of target users were re-activated, demonstrating the unmatched prowess of HUAWEI Prediction in helping apps retain users in the most cost-effective manner possible.
To reach users who will churn in the near future precisely and retain them in time is the best cost-effective operations strategy. That's why we should use predictions.
To learn more about HUAWEI Prediction, check the document.
Whether it's online or offline operations, a lack of a steady customer stream can cause you to rethink your marketing strategies. A number of solutions have been proposed, including more ads, better products, and improved user services, but sometimes, even the most diligent efforts aren't rewarded.
So wouldn't it be great, instead, to be able to determine the potential of existing users, and predict return users in advance? This would give you a high-level understanding of your users' habits and needs, enabling you to win them back and convert them into loyal, paying users.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
1. Method for evaluating user return potential
From registering an account to completing a second purchase, a user's return potential is highly correlated with their behavior data, which includes page views and active time. Therefore, analyzing user behavior data in your app is of immense help in predicting return users.
HUAWEI Prediction uses a massive amount of experimental data to train a machine learning model and uses this model to predict the probability that active app users will make a payment over the next week with their payment data from the most recent week as inputs. This service can predict return probabilities of users depending on such events as in-app purchases reported to HUAWEI Analytics by your app.
2. Method for extracting the most value from existing users
After determining how users behave within your app, you'll want to predict the audience with the highest return potential, and formulate targeted policies for these users, to maximize efficiency.
For example, some products have a membership system, and continue to attract return users by providing discounts and rewards point redemption schemes. Most current membership systems work in the following way: A registered member earns points by making purchases, which can then be exchanged for coupons and gifts. This type of framework does attract return users to some extent. However, due to a relatively high exchange threshold, limited gifts and complicated rules for using points, users will tend to lose interest in the membership scheme.
Our return prediction can optimize this type of incentive scheme, helping your app attract return users at a lower cost. It is found that the predicted return users tend to be more active, based on either their most recent app usage, or throughout the most recent week. From this information, we can surmise that such users would like to make a second purchase due to the need arising from a specific scenario, and will tend to take multiple products into consideration. Therefore, if your product is not competitive enough, it will likely lose many of its paid users.
3. Method for winning back and converting users
Predicting return users helps identify users with a high return potential. The next area of focus is boosting the probability of return. Some shopping apps will make use of their membership system to provide time-limited bonus points, coupons and cash incentives on weekends and holidays. After the promotion is complete, it is found that the activities attract return users at a higher cost, which is not ideal.
HUAWEI Prediction precisely segments return users, helping you avoid this problem. You can set predicted users with high return potential as the audience, and add the Prediction filter in Remote Configuration, as well as the corresponding parameters for the promotion to ensure that the promotion is only sent out to those users in the predicted audience.
The results from a remote configuration experiment in A/B Testing indicated that the treatment group which contains predicted return users yielded as high of a conversion rate as that from the control group, and at a 60% lower cost.
Predicting and analyzing behavioral characteristics of users with high return potential, through HUAWEI Prediction, enables you to formulate targeted marketing strategies, and achieve sustainably rapid growth, even in the wake of increasingly scarce traffic streams.
To learn more about HUAWEI Prediction, feel free to check out this document.
Financial security is the cornerstone for the development of the banking industry. A long-term challenge to this industry is financial fraud, which can be reduced with the help of innovative technologies.
China Merchants Bank, a leading retail bank in China, has three key requirements for financial security: controllable risks, moderate compliance, and leading experience. Since 2016, China Merchants Bank began deploying its Libra System, a next-generation risk control platform, which was formally integrated with HMS Core Safety Detect in April 2021.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
The Libra System is a unified risk control platform that can be applied to multiple online and offline channels and scenarios of China Merchants Bank, including swiping a bank card using a POS machine, withdrawing cash through an ATM, and operations such as sign-in, transfer, payment, and purchase of financial management solutions on its mobile app. In such scenarios, the Libra System will instantly perform risk detection and provide a risk control score. Wu Lei, director of the cyber security team at China Merchants Bank, said: "HMS Core Safety Detect supplemented risk detection dimensions that were previously unavailable, and also fulfilled our three key requirements, which are controllable risks, moderate compliance, and leading experience."
SysIntegrity (system integrity check) in HMS Core Safety Detect can quickly check whether the device running the China Merchants Bank app is secure, helping the app implement trusted execution environment check and make decisions regarding risk control. The app also used WifiDetect (malicious Wi-Fi detection) in Safety Detect to ensure that transactions made on insecure Wi-Fi networks are safe. WifiDetect helps the mobile banking app detect risks under the premise of regulatory compliance, improving transaction security for China Merchants Bank.
On top of that, Safety Detect also meets the strict low-delay response requirement of China Merchants Bank for great user experience. On April 8, 2021, China Merchants Bank app version 9.2.0 was released to major app stores, and in its release notes the app lists HMS Core Safety Detect support. The China Merchants Bank app calls Safety Detect during app launch, payment, and transfer, with its highest daily calls reaching 17 million, and maintained high app performance stability throughout. Wu Lei said: "We require that the response speed for transaction risk decision-making is within 30 milliseconds. This is achievable because Safety Detect is integrated into the Libra System with few external connections. Its high availability, security, and flexible development are ideal for us, without affecting user experience."
In the future, HMS Core Safety Detect expects to cooperate with more customers in the banking and finance industries to build financial security, safeguarding user property security. Click here to find more information of HMS Core Safety Detect!