These updates to Airship Journeys make this customer journey tool even more insightful to use and some of our customers have already seen double to even triple digit percentage performance improvements!

Measure Your Messaging Impact with Control Groups

The best way to understand how big of an impact your messaging has on your customers is to compare your messaging against a control group. With this update, you can set control groups and use integrated metrics to easily visualize the impact each journey has on your business goals and continually optimize each step in your user journey.

What kind of results might you get by using control groups with Airship Journeys? Here’s what three of our customers saw when they measured the performance of their customer journeys by using control groups:

• Radisson Hotel Group created a cart abandonment series for their mobile app that spanned three messages over four days with the first message guaranteed a room’s rate. They were able to drive an 11% lift in completed reservations over the control group.
• Music social network Vampr launched a reactivation journey that targets inactive app users with a series of three push notification and a final email within two weeks. They were able to reactivate users at a 277% higher rate than those in the control group!
• On Black Friday, J. C. Penney Company used Airship Journeys to transform a one-message cart abandonment automation into a two-message journey and it made a big difference. They saw an all-time increase of 40% higher purchase completion rates over the control group baseline.

Find the Winning Message with A/B Testing

With A/B testing, you can update the content of any message in the journey and test that update against the original content. In this way, marketers can track their impact and make continuous and measurable improvements over time. You’ll be able to see which messages perform better and improve your overall conversion rate. 

According to Gartner, organizations that prioritize testing are twice as likely to outperform their peers. Putting control groups and A/B testing together gives marketers the power to provide an exceptional journey experience for the customers. With Airship Journeys, this is all easy to use and simple to modify.

You can learn more about Airship Journeys (including a demo of using these new features) and building a culture of experimentation in our webinar, “Creating Extraordinary Customer Journeys with a Culture of Experimentation.” If you missed the live showing, you can still register to get a recording! 

The post How Brands Are Improving Conversion Rates by Using Airship Journeys Optimization appeared first on Airship.

• 66% of consumers say instant, on-demand engagement is very important to winning their business 

• 78% say that they would be more loyal to a brand that understood them and what they are looking for

• 52% say they’d switch brands if they didn’t feel they were receiving a personalized experience 

That’s why we created Airship Journeys, a revolutionary new solution that simplifies how marketers create, measure and perfect customer journeys. We’ve published a playbook highlighting nine powerful strategies inspired by leading brands who are already using Airship Journeys to craft highly effective customer journeys at each stage of the customer lifecycle. 

Check out our key takeaways below, and download the full Airship Journeys Playbook: 9 Powerful Strategies to Acquire, Build and Grow Your Audience for more insights, tips and inspiration!

Customer Journeys: Acquire

Attracting new customers is expensive and time consuming. That’s why, during the acquisition phase, you need to give them a warm welcome and deliver an engaging customer experience that drives action and adoption. 

New customer activities, like an opt-in on your website, provide the perfect opportunity to launch a multi-touch campaign highlighting key features of your brand, and show them just how much value you have to offer. Use the series to encourage them to download your app, and demonstrate some of the ways it can make their lives easier. This is also a great time to engage them in your loyalty program, so they can start realizing the benefits of deeper engagement with your products or services. 

Customer Journeys: Retain

Most marketers know it’s far less expensive to retain existing customers than attract new ones. But customer expectations are high, and rising. That’s why, during the retain phase, it’s essential to deliver value in the moments that matter most to your customers. 

For example, you can trigger a series of messages around actions like a ticket purchase or reservation, and send SMS or emails with important information, upgrades or changes. Entice them with an incentive based on app or web activity. Plus, you can use Airship Journey’s orchestration features to focus on the channels that matter most to your customers, while carefully controlling frequency to avoid overmessaging.

Customer Journeys: Grow

Now that you’ve established rapport with your customers, it’s time to grow those relationships — and their lifetime value. To do that, you need to show them you know them by engaging on their preferred channels, in real-time with contextual content that predicts their needs.

When a customer opens your app, you can bet your brand is top of mind. Leverage this moment to trigger a multi-channel series of offers or incentives based on past purchases or activity. Use the series to remind customers to visit your store or website. And send follow-up notifications offering post-purchase surveys, encouraging social sharing or recommending other items they might like.

This is also your opportunity to re-engage customers who are likely to churn, and take action to keep them. Use Airship’s machine learning algorithm to identify users at high risk and send an email with an offer or incentive to win them back, before it’s too late. 

Every stage in the customer lifecycle presents opportunities to deliver great CX, build loyalty and grow revenue and lifetime value. Download the e-book today to learn more about how you can use Airship Journeys to deliver meaningful, personalized messages at every juncture.

Let’s Connect

We can help you put together a winning customer journey strategy

Contact us today!

The post How to Create Seamless Customer Journeys for Each Stage of the Customer Lifecycle appeared first on Airship.

The Problem with Other Customer Journey Solutions

For our customer journey research, marketing teams from different brands shared their internal planning documents, including diagrams and spreadsheets for how they planned to run and orchestrate campaigns. These documents work as both an organizational tool for what needs to be deployed on what platform and how those different touch points can be orchestrated together.

We noticed a common theme with these planning documents: marketers were mapping out the progression of customers across milestones and organizing messaging around each milestone. Unfortunately, this method can lead to performance issues or poorly targeted messaging. It’s more difficult to determine important metrics like how many customers advance at each stage or know where customers fall out when mixing milestones and messages together.

How We Made Journeys Easier with Maps

We observed this workflow, and others like this, and saw an opportunity to make building, deploying and tracking messaging across milestones a whole lot easier. We developed a “branching” like system where marketing teams can quickly map the lifecycle as interconnecting journeys and view them together in a map. 

This map view shows the relationship between journeys as well as high level performance metrics, including user flow and conversion rates by milestone. You can also quickly access each underlying Journey for a deeper view of performance metrics by message. 

Journey Maps Examples

We’re confident that marketers will find Journeys easy and helpful to use with the new mapping functionality. We already had great adoption of Journeys from our customer base with over 150 Journeys deployed across 27 customers, so it was a quick start to find marketing teams willing to participate in product discovery. 

Here are some examples:

A health app wants a way to remind users to reschedule their appointment so that they continue to experience the value of the health service and app. This is easy to do with Maps by adding in a cancellation Journey, and leading the customer back into the appointment cycle.

A food and beverage company wants to onboard their users so that they become rewards members as quickly as possible. They need to capture users who were not completing registration steps so they can quickly re-engage them with deals and discounts to prevent them from churning. Maps makes it easy to identify and recapture the users who are falling out of onboarding.

Airship Journeys and Maps can help you create and connect Journeys, while giving you the data and tools you need to better engage with your customers. You can learn more about Airship Journeys here.

Let’s Connect

Let us know how Airship Journeys can help you reach and engage your customers

The post How We Put Together Journey Maps appeared first on Airship.

A few months ago, the team at Airship launched Airship Journeys, a revolutionary solution for creating, managing and perfecting customer journeys. 

The solution was the result of a year’s worth of research, asking marketers what works and what doesn’t in their existing customer journey technology. The verdict: most customer journey solutions are too complicated, tedious and confusing. 

With that in mind, we reimagined the process of building customer journeys, empowering marketers to connect with customers in a whole new way: across all channels, in the moments that matter most.

That’s why I am very excited to announce that I will be hosting a webinar with David Cook, Lead Solutions Engineer, to take you on a guided tour of Journeys in our webinar on February 18th. 

We’ll show you how you can build journeys that create massive value across all your channels and turn data into insights and insights into action.

Make sure to register and don’t miss out on this opportunity to get an in-depth look into Airship Journeys.

The post [Upcoming Webinar] Reinventing the Customer Journey: A Guided Tour of Airship Journeys appeared first on Airship.

To help marketers get a clearer idea on what AI can do for them today, Airship’s Lead Data Scientist, Lisa Orr, and Strategic Account Manager, Phil West, hosted a webinar on how to use data and our machine learning solutions, Predictive Churn and Predictive Send Time Optimization. We shared best practices and strategies from top brands who are using predictive data to reduce customer churn, improve engagement by 10% or more and drive increased customer lifetime value.

Here are the highlights from the webinar “How Top Brands Use Machine Learning Data to Reduce Churn & Boost Engagement (And How You Can Too)” — to get all of our insights, watch the full replay anytime.

1) The Differences Between AI, Machine Learning and Predictive Analytics

AI, Machine Learning and Predictive Learning are often put together and sometimes mistaken for each other, but they are three different tools that can help marketers make decisions based on data. To help differentiate between the three, we used the example of how a Roomba works.

AI is a branch of computer science dealing with the simulation of intelligent behavior in computers; the capability of a machine to imitate intelligent human behavior. A Roomba can appear intelligent, as if it’s making its own decisions wandering around, cleaning the dirt on your floor, but it’s actually following a set of rules (like turning after hitting a wall).

Machine Learning is a branch of artificial intelligence in which a computer generates rules based on raw data that has been fed into it, or put simply: a machine that learns from data. For a Roomba, beyond a simple set of rules (like turning after hitting a wall), the vacuum robot could be mapping the space, figuring out where it’s been so it doesn’t revisit the same places again.

Predictive Analytics is a variety of statistical techniques that analyze current and historical facts to make predictions about the future or otherwise unknown events and allow assessment of risk or potential associated with a particular set of conditions. This is a catchall phrase that ultimately means there is some smart data processing going on under the hood (like with the Roomba) providing marketers with a tool for making data-driven decisions fast.  

2) Segment Audiences Based on Churn Risk to Boost Results

Airship built a powerful machine learning model that uses app data to predict churn, and simplifies the concept into three groups: High, Medium, and Low. Here’s our advice on how to approach each risk group:

High Risk: You’ve spent the money to acquire and engage, but now they’re at risk: how do you bring them back? Don’t be afraid to engage with them on a different channel. Everyone likes a deal, so sending a deal or offer to high-risk users to bring them back works well with email, but it also delivers results with the app. One Airship customer set up an automation pipeline for their high-risk group and that worked so well, it became a top-three revenue driver.

Medium Risk: Keep these users educated and engaged. Using educational or inspirational messages can remind them why they downloaded the app in the first place. You can also highlight features that low-churn users are using, to drive that same behavior with them. This is a strategy that also works with High-Risk users.

Low Risk: These are your really active folks and your best users, so how do you reward them or drive additional value? Since they are super active with the app, they are a great group to test different messages. You can try an additional push and see if they opt-out or delete the app with that extra message, but more often than not, the additional push actually helps the users engage with the app, not push them away.

3) Segment Your Audience to Send Test Notifications

No one wants to spam their customer, but you also want to make sure that you’re sending the right amount of notifications to get the best engagement. Segmenting your audience by risk can also separate the right audience to test those additional pushes.

A good test is to target low-risk, and possibly medium-risk, with an additional push, and see how the messages engage with that audience. What we see is that the additional push increases engagement, rather than push customers away into opting-out. One of our clients targeted a low-risk audience for an additional message every week and it became their best performing engagement.

It’s also important to think about consistency, rather than simply frequency. Another customer noticed a dramatic fall in their app opens and they weren’t sure why, until they realized they forgot to send their one weekly push. It can be easy to take notifications for granted!

4) The Best Time to Send a Notification May Surprise You

When is the best time to send a notification? It depends not only on the brand, but also by user. Sending a message during a time an individual is most likely to engage has a positive impact on direct open rates, and that window of time may be outside daytime hours. For one brand, Airship’s Send Time Optimization data showed that the highest engagement hours were actually the late night hours, with high activity happening even between 1 a.m. and 4 a.m.

Another example is a high-end retailer whose data showed a huge spike at night with opens, between 10 p.m. and midnight. Thinking about what kind of customers were using the app at that time (perhaps opening the app at the end of a long work day) they decided to try out inspirational messages, with the option of sending it out to everyone or targetting just those using the app that that time.

What works for some, may not work for all. Lisa worked with a data analyst that had a global audience to take into consideration. Before Send Time Optimization, he had to do a lot of research and make a lot of guesses, thinking about the different geographical regions and activity patterns of his customers. The Optimization button took away all that decision making, instead using data to send messages at the best times for the specific groups of customers. It worked so well, he sent a thank you note to the Airship team!

Is your brand using machine learning to reduce churn and boost engagement? For more insight and use cases, watch the webinar here and make sure to subscribe to our blog newsletter to get the best of Airship straight to your inbox.

Are you making smart data moves? Airship can help you develop strategies around Predictive Data that maximize your audience potential and engagement. Connect with us today and let’s talk!

The post How to Use Machine Learning Data to Reduce Churn & Boost Engagement – Webinar Recap appeared first on Airship.

Figuring out when to send notifications is a tricky problem. Personalizing the experience to each user is challenging — let alone sending individually-timed messages across an entire audience base.

We solved both problems with our Predictive Send Time Optimization Solution. The model decides the best time based on individual engagement patterns — and provides an easy way to send to each user’s best time in a single step.

The (Mostly Manual) Ways Brands Have Been Trying To Predict Best Time To Send

While building this model we polled customers about their general engagement strategies to determine how they decide the best time to message their users.

Some marketers send notifications the moment they are finished composing them, regardless of user timezone.

Others use their intuition to identify the best time to send a campaign: a marketer of a dating app scheduled messages at 6pm local time as it represents the window when people get home from work and start thinking about their potential dates for the night.

On the more data-driven side, we learned that a telecom company selected best time for each individual based on a few rules: send at the hour in which they last saw an open from a user, unless that time was outside day hours; then send to an early morning hour the following day.

Our conversations revealed that marketers want to be more data-driven with their send time decisions, but are still relying mostly on intuition and convenience.

Machine Learning To The Rescue

We created a model that:

• Selects optimal engagement time personalized to each individual user

• Is localized to each user’s timezone

• Is extensible to dormant users, and

• Is both actionable when building campaigns and explorable in our data products

Our models are based on a user’s recent engagement history.

To start, app opens are localized to the user’s timezone and aggregated to the hour over the last 60 days of app activity. Best hour is determined by striking a balance between the user’s engagement patterns and a generalized model of engagement patterns across an app’s total audience.

The model includes a preference for daytime hours to avoid assigning late night best times to users who may infrequently engage with apps late at night. (However, if a user truly spends most of their time engaging at night then their personalized best time will fall into night time hours.)

Our model also creates a generalized “best hour to send” recommendation, which brands can use as a rule for sending messages to dormant or low-activity users. The general best hour aggregates opens across app users and selects the best hour based on the most frequent opening time for each app and app operating system.

Testing & Results

To test the model we compared predicted best times to true open times for open events that occured after model generation. We compared the best time model to both a baseline model (e.g. everyone’s best hour is set to 12pm) and to the general best hour (e.g. everyone’s best hour is set to the general model best hour).

We found that personalized best time more accurately predicts the hour a user opens than either the general model or the baseline model. It performed better than both the generalized best time (52% higher match rate on average) and the baseline model (100% higher match rate on average).

“We found that personalized best time…performed better than both the generalized best time (52% higher match rate on average) and the baseline model.”

Making It Easy To Put the Model To Work for You

We want to make deriving insights and taking action with the Predictive Send Time Optimization model simple for the marketer and their analytics teams:  

• As with our Predictive Churn model, we ship weekly predictions as user tags to facilitate integration between predictive model output and our data solutions.

• We make it easy to schedule messages with Predictive Send Time Optimization

• We offer simple ways to visualize the user-level predictions

• Analytics teams can ingest predictive tags into their own systems with real-time data streaming.

• Through our customer analytics solution Insight we provide a deeper dive into the model, surfacing the cross-audience best hour as it relates to each day of the week and for each app operating system.

Since launch we’ve seen major retailers, a major research firm and an international media company adopt optimized notifications into their daily workflow. We’ve seen higher direct open rates with optimized sends and have heard positive feedback on how the tool is saving teams time and effort. One of our customers is even using Predictive Send Time Optimization to meter traffic into their app.

We are excited to hear about the success of the model in the wild and look forward to building more predictive tools that fit our customers’ needs.

Ready to get started with Predictive Send Time Optimization? Let us show you how it can work for you in a personalized demo! Contact us today!

The post How We Built Our Machine Learning Model for Predictive Send Time Optimization appeared first on Airship.

With the cost of acquiring new app installs skyrocketing, keeping users engaged who have already installed is critical for maximizing acquisition spend and customer lifetime value. Urban Airship’s Data Science team has spent the last year developing a way to identify and target users who are likely to stop using your app. We are calling this Predictive Churn.

Here, I provide insight into the process of building a scalable predictive machine learning model over billions of events and address how these predictive capabilities lead to new insights into user behavior, fuel new engagement strategies and impact user retention.

Churn Prediction: Developing the Machine Learning Model

Churn prediction is a straightforward classification problem: go back in time, look at user activity, check to see who remains active after some time point, then come up with a model that separates users who remain active from those who do not. With tons of data, what are the best indicators of a user’s likelihood to keep opening an app?

For starters, we can look at a user’s general app activity. How often does the individual open the app? How recently have they opened it? If we are trying to predict who will be opening in the near future, a good indicator might be whether a user is already opening the app or not. What about the impact of receiving push notifications? Does choosing to receive them in the first place impact user app activity?

Looking back 60 days into customers’ data, we rolled up message sends and app opens into graduated windows with additional information such as device platform and app id. We then labeled each user as having churned or not churned based on whether they open the app in the next 30 days.

>> Related Benchmark Report: How Push Notifications Impact Mobile App Retention Rates

With the feature data rolled up for each user, we trained a model using the gradient boosted decision trees machine learning algorithm. We performed a six month historical study of churn prediction training the model over dozens of features (i.e. the observable user and app behaviors). Our goal was to get a high level of accuracy in predicting churn — as well as insight into what factors influence it.

Looking into the model and which features had the biggest impact, we found some interesting patterns:

• Open activity. By far the biggest predictor of future activity was how long it had been since their most recent open. Which makes sense as it’s the thing we’re trying to predict just in the opposite direction with regard to time. Open counts within recent time windows and recency of opens both play a large part in predicting who is going to churn.

• Send activity. Another interesting finding was that receiving push notifications had a positive impact on user retention. This again makes sense —  if you choose to opt in to push notifications then you’re signaling an active interest in the app and an openness to discovering more of its value.
  
  A recent data study from Urban Airship’s data science and marketing teams illustrates this point. Users who received at least one push notification had a 66 percent higher retention rate as compared to users who received no push messages.
  
  We also found a correlation between number of sends received and days retained where the more messages a user received the longer they continued using the app. As with open counts, both the recency and frequency of sends played a role in modeling churn activity.

Scaling the Model for Mobile

Now that we created a working model, the next step was to test its ability to scale to thousands of apps and billions of users.

Adding more apps quickly exposed a weak spot: the re-processing of data from a csv (the output from a MapReduce job to create our feature data) to a sparse matrix (format required by the boosted trees model). This processing step was causing the job to fail due to memory issues. Adding compute resources would solve the issue temporarily, but as we added more apps it became clear we needed to rethink our strategy. Writing to a csv was useful during the development phase so we could double check our work. But besides being human-readable, there was no real benefit for using the intermediary format.

We instead re-formatted the feature data into sparse matrices directly within the MapReduce job itself. So far further memory pressure has been solved by adding more machines in the MapReduce phase or by upping the size of the single machine used during the modeling phase. With the formatting change in place, we are able to train our model over thousands of apps simultaneously.

Productizing the Model

Once we had a working model at scale, the next step was figuring out how to best provide these predictions to our customers. For each user we feed into our model we get back a probability to churn score ranging from zero to one. The higher the score, the more confident we are that that a user will churn. Likewise the lower the score, the more confident we are that that user will stick around.

If you look at the outer edges of churn probability, you’ll find your app’s super fans (those who are definitely coming back) and your app’s more difficult patrons. And in the middle resides a mixed bag of individuals where the model is less confident about which way they will go.

This conception of the model led us to segmenting users into three groups: high-risk, medium-risk, and low-risk groups. Sending up predictions this way gives our customers the ability to adjust their audience segmentation by how aggressive their engagement strategy is.

For example, if you’re attempting to re-engage at-risk users you can scale up the reward offered if you include only the high-risk audience or scale it down if you include both the high-risk and medium-risk audience.

Once a user has been categorized as high, medium or low risk-to-churn, the data is immediately available through our real-time mobile data stream for analysis or action in other systems, dashboards to view five-week performance, and visualizations to show how effective your efforts are in moving users from high-risk to lower risk states.

Churn Prediction in the Wild

To illustrate the how Predictive Churn can influence engagement strategy (and vice versa) here are a few anonymized Urban Airship customers and their respective churn score distributions:

Above visualizations: Histograms of user churn prediction scores for three example Urban Airship apps. Each bar represents the percentage of users whose probability to churn score fell within the specified score range. Colors represent risk category: green is low-risk, yellow is medium-risk, red is high-risk. Predictions were made on February 3rd, 2017.

Comparing the three apps represented by the histograms above,  we can see very different user distributions for churn prediction:

• For App A most users are in the low-risk group

• For App B most users are in the high-risk group

• For App C we see a bimodal distribution where a large group of users are in the high-risk group and another large group of users are in the low-risk group

How does churn prediction relate to app engagement strategy? All three of these apps have large audiences (greater than 3 million unique devices) and all three use push messaging to engage their users. However, there is a large difference in how these apps approach engagement.

• App A (low-risk app) has a very sophisticated engagement strategy, using advanced features extensively (i.e. tags, lifecycle lists, in-app notification, message center etc.), are targeting most of their audience with segmented pushes, and are getting high levels of engagement with those messages (through direct or influenced app opens).
• App B (high-risk app) is using a very basic messaging strategy with very simple segmentation, is messaging infrequently to a limited segment of their audience, and is gaining almost zero audience engagement.
• App C (the app with a bimodal distribution for churn scores) has a middle-ground strategy. They are utilizing a few advanced engagement features such as aliases, badges and deep links, are sending almost exclusively broadcast (non-personalized) messages, and are seeing a good amount of audience engagement with those messages.

By comparing churn prediction and app engagement strategy our customers can identify areas for improvement, effect those changes and compare how churn scores change week over week.

For App B, we would recommend targeting more of their audience, messaging more frequently and expanding their use of messaging strategy to alternative communication such as in-app messaging and message center as well as improve targeting via tags and named user.

For App C, we would suggest moving away from broadcast pushes as well as targeting the high-risk audience explicitly.

What’s Next

Predictive churn enables companies to reach customers at the right time on the right channel and with the right content to turn them from a customer than churns to one that stays. But this is just the start of data science and machine learning capabilities. Not long from now, it will be standard for companies to use data to send automated and proactive notifications and improve customer engagement.

The post How We Built Our Machine Learning Model for Churn Prediction appeared first on Airship.