Katie Darren – Kochava

Avoid an Overcrowded Media Mix

Wed, 15 May 2019 23:47:16 +0000

Today’s marketer is data-driven. More than simply connecting our clients with their data, Kochava seeks to empower strategic decisions through marketing intelligence. Our growing Client Analytics team is comprised of data analysts that work hand-in-hand with clients, delivering face-to-face business value assessments (BVAs) and quarterly business reviews (QBRs) that inform, educate and empower UA managers, CMOs and other decision makers to position their brands for success.

In this series, members of our Client Analytics team will explore marketing intelligence insights available to our customers from the tools Kochava provides.

Go for unique traffic

Here’s food for thought—how many of a partner’s installs are unique vs. influenced? In essence, does a media partner have unique reach, or are they stepping over the toes of another partner and bidding against each other for the same placements? With multi-touch attribution, Kochava can show the path a user took across multiple impressions and clicks to finally reach the point of conversion.

Not only can Kochava show you influencers that overlap across partners, but we can also give insight into self-influencers, or partners who repeatedly hit the same user again and again. Too many self-influencers from the same partner or DSP may suggest poor frequency capping, translating to a bad user experience. When two partners have significant influencer overlap, offering minimal uniqueness in reach, the marketer should consider trimming one of those partners. Ideally, you want unique traffic, with partners delivering valuable users that no other partner could have delivered.

Improve quality

Further insights can be gained by adding a qualitative layer that assesses performance by downstream completion of key performance indicators (KPIs). Kochava offers flexibility to customize and refine the analysis against vertical-specific KPIs, such as: free trial starts for a video streaming app, first, second and/or third order placements for a QSR app, and level completes or gameplay for gaming. Whether it’s a single KPI action or a combination/sequence of multiple KPIs, the time window within which these activities must be completed can also be refined.

Combining the influencer layer with the qualitative layer can offer a unique intersection of insights. For instance, compare the quality of the unique vs. influenced installs. Here, “quality” would mean installs with the completion of KPIs downstream within the optimal time frame. Ideally, you want to see higher quality for your unique traffic than your heavily-influenced traffic. That being said, in certain cases, trends may show that the combination of two or more partners influencing a user consistently results in higher downstream engagement with KPIs. This may suggest that the confluence of these media partners and/or the combination of marketing channels they traverse delivers a winning combo for engaging quality users.

Also, compare the percentage of unique traffic for all attributed media partners to the average unique traffic for the app. Make note of media partners with a below-average percentage of unique installs and look deeper into these media partners. Kochava can even decomp performance and quality at a much more granular level, such as by site or creative ID.

Unattributed (a.k.a. organic) traffic also offers a helpful quality baseline. Organic users are those who seek out and install an app on their own, without clicking on any ads. They often index higher on downstream performance and engagement with the app. Measuring paid media partners against organic quality trends allows you to see those under- or over-indexing on quality. Consider trimming partners that significantly underperform. At the same time, be watchful of partners that consistently parallel organic trend lines, as this may be a proxy for clever organic sniping tactics. Implementation of fraud tools to prevent click flooding and click injection will help mitigate organic sniping.

Use case

A gaming company ran ads with 10 media partners during April 2019, and they want to understand the uniqueness and quality of the installs driven by each partner to determine if there are opportunities to cut/expand marketing budgets next year. “Purchase” is their key KPI and they typically see purchases happen within seven days of the install so they want to see what percentage of installers had a purchase within seven days of the install.

The Uniqueness & Quality By Media Partner chart below shows the results of the analysis that the gaming company completed. After running this analysis they compare which media partners had unique and quality installs and determine that five of the 10 media partners they are running media with have lower than the overall average of 69% unique and also have unique quality that is lower than the overall quality.

The client decides to dive into the five media partners (media partners C, F, G, H, and J from the below chart) that had a very small amount of unique installs and those unique installs had lower quality than overall quality as a first step. After looking deeper at the traffic these media partners provided and the cost associated, the gaming company decides to cut back dollars put toward these media partners and invest deeper into higher quality unique sources. By cutting back dollars put toward these media partners, they can become more efficient with how they spend their marketing dollars to drive more high-quality, unique installs.

Here is an example of the results used to make their media partner cuts:

The takeaway

We typically recommend that marketers focus on the quality of a media partner’s unique traffic—all else being equal—those are the installs that the marketer would not have received without the partner in question.

In the table above, we see that media partner F was only 25% unique—meaning three-fourths of their installs would have attributed to other partners if partner F wasn’t in the mix. And, the influenced installs were of higher quality than what the partner uniquely touched. Overall, partner F is not contributing to better installs.

There are a number of ways to optimize your ad spend, and thoroughly evaluating your media partners is an important part. Contact Kochava to learn how you can leverage our turnkey partner analysis methods to see which partners are delivering unique, high-quality users.

On the lookout for your next media partner? Download the latest Kochava Traffic Index to see the top 20 ranked partners for Q1 2019.

Ready to start looking into your media partner mix? Contact your client success manager or support@kochava.com. With Kochava, you have a support team at the ready to meet your needs.

Not a Kochava customer? Contact Us Today.

Katie Darren – Client Insights Analyst

Kochava

The post Avoid an Overcrowded Media Mix appeared first on Kochava.

]]>

New Ways to Grow Your Audience

Tue, 16 Apr 2019 21:44:51 +0000

Today’s marketer is data-driven. More than simply connecting our clients with their data, Kochava seeks to empower strategic decisions through marketing intelligence. Our growing Client Analytics team is comprised of data analysts who work hand-in-hand with clients, delivering face-to-face business value assessments (BVAs) and quarterly business reviews (QBRs) that inform, educate and empower UA managers, CMOs and other decision makers to position their brands for success.

In this series, members of our Client Analytics team will explore marketing intelligence insights available to our customers from the tools Kochava provides.

Targeting new users effectively is every marketer’s daily challenge. User acquisition or growth managers are tasked with reaching new high-quality users and that job doesn’t end when users install the app. Although people are spending more time on mobile devices, because of its overstimulated environment, retention drastically decreases with time, regardless of the vertical.

Without knowing, you may be mistargeting with irrelevant messaging or expending valuable ad spend on a less optimal segment. To reach high-quality users, revisiting your audience data and considering different ways to analyze it will shed light on better ways to target, and assessing app engagement is a great place to start.

How can you target more effectively? The One-and-Done analysis of user engagement

How a marketer determines engagement differs widely and is based on a variety of factors and nuances, such as the app vertical, how the app monetizes, and other variables. To help distinguish the different types of users in an audience, the Kochava Client Analytics team has a simple, customizable query called the One-and-Done analysis.

What is the One-and-Done analysis?

The One-and-Done analysis is a versatile SQL script that helps measure the quality of users where quality is the percentage of users who complete an event. The analysis can help you segment users under three classifications based on actions performed during a specified time frame:

The Non-Performer
- User installs the app and then disappears without completing a KPI

The One-and-Done
- User installs the app and performs one post-install action but never returns
The Performer
- User installs the app and performs consistently through regular engagement with KPIs

You can determine which KPIs to evaluate, the specific time frame of installs, and when the post-install events occurred. Then, use the Query tool to write SQL scripts to analyze, cohort, group, slice, dice, and pivot your data in unique, customized ways. Or, the Kochava team can help you customize the query for your specific brand’s needs.

In addition, measuring ad partners by how many users are classified as “one-and-done” is beneficial in comparing user quality by media partner. Partners that heavily over-index in one-and-done users can be trimmed or eliminated, whereas partners with the lowest percentages can be prioritized.

The potency of this analysis can be further increased by making sure the app is measuring a healthy mix of KPIs throughout the user funnel. See Post Install Event Examples for a list of recommended events to track by app vertical.

Below is an example of how you can visualize the data from the query to help distinguish the one-and-done vs. performer segments by media partner, providing insight into which partners drove quality user engagement.

Sample use case

A financial service provider specializing in money transfers has a goal to increase this conversion event and are exploring ways to grow their user base. In seeking assistance from the Client Analytics team at Kochava, they learned that a percentage of their converted users were “one-and-done.” Additionally, they learned which media partners repeatedly delivered this user type. They can now apply their learnings toward their next campaign by removing certain media partners and creating a lookalike campaign of their newly segmented performing users.

How can you apply the One-and-Done Analysis results?

There are a couple of ways in which you can apply your learnings from the One-and-Done analysis to grow your audience:

Audience Targeting: Use your performers as a seed for a lookalike model. Activate a campaign with the lookalike audience to acquire more users with attributes that commonly define performers, getting stronger ROI on your acquisition efforts.

Push Engagement: If your goal is to retain the one-and done user, load this audience segment into your push provider platform for a push, in-app message or SMS campaign to reengage them with a contextually relevant offer.

Marketers can also export a list of device IDs by segment and activate them with a preferred media partner.

Grow your audience with the One-and-Done analysis

While you may already be segmenting your audience, the One-and-Done analysis brings highly customizable and flexible turnkey segmenting. If you haven’t identified and segmented one-and-done users, you are probably mistakenly considering them active performers.

If you are looking for greater insight into the quality of installs, the One-and-Done analysis may be a solution to explore. The output is easy to understand and can be automated on an ongoing basis to provide actionable audience segments.

Interested in refined and customizable user segmentation? Contact your client success manager or support@kochava.com.

Not a Kochava customer? Contact Us Today.

Katie Darren – Client Insights Analyst

Kochava

The post New Ways to Grow Your Audience appeared first on Kochava.

]]>

Beating the User Attrition Curve with Predictive Behavior Modeling

Wed, 03 Apr 2019 21:32:24 +0000

Today’s marketer is data-driven. More than simply connecting our clients with their data, Kochava seeks to empower strategic decisions through marketing intelligence. Our growing Client Analytics team is comprised of data analysts who work hand-in-hand with clients, delivering face-to-face business value assessments (BVAs) and quarterly business reviews (QBRs) that inform, educate and empower UA managers, CMOs and other decision makers to position their brands for success.

In this series, members of our Client Analytics team will explore marketing intelligence insights available to our customers from the tools Kochava provides.

Nearly 70% of mobile users churn within 90 days. With this in mind, how do you keep the users you find? Predictive behavior modeling was developed by Kochava data science and engineering teams to help clients predict the churn of a user before it happens. This data empowers marketers to improve retention rates and navigate strategy toward better return on investment (ROI).

How predictive behavior modeling works

After a new install, our machine learning algorithms go to work using a form of decision tree modeling to analyze recency, frequency, trend metrics, and other data variables during the first 7 days of a user’s interactions with the app.

On day 8, the user is assigned a churn score. “Churn” in this case means how likely is the device to not have activity in the app between day 8 and day 38 after install.

A user with a score of “Low” indicates that it’s likely they will have further engagement with the app between day 8 and day 38. A user with a score of “High” indicates that it’s likely they will churn, or have no further activity in the app between day 8 and day 38.

Determining prediction accuracy

Predictions are only as good as their ability to accurately forecast. Assessing the accuracy of the predicted churn scores from the machine-learning model required analysis of the downstream activity for each churn group to see how closely the predictions aligned with the outcome.

Through extensive examination of multiple data sets and applications across various verticals, a common trend line emerged. With overwhelming consistency, users adorned with “Low” churn likelihood scores displayed the highest percentage of session and in-app event activity between day 8 and day 38 post install. Users scored at “Medium Low,” “Medium High,” and “High” exhibited an exponential decline in engagement over that period. With prediction accuracy confirmed and aligned with expectations, these churn scores provide valuable, actionable intel.

Applying predictive behavior modeling

Providing advanced insight into a user’s likelihood for attrition allows marketers to strategically intercept that user with targeted reengagement efforts. Using analytics, marketers can segment audiences according to churn-likelihood scores and syndicate “Medium High” and/or “High” likelihood segments to reengagement partners like Kochava for focused targeting campaigns to drive retention.

Marketers can build dynamic audiences around churn scores and set triggered push campaigns with contextually relevant, dynamic content to promote user retention and growth.

User audiences with a “Low” churn likelihood can be valuable seed audiences from which to build lookalike models to attract quality users with similar characteristics that promote user longevity and loyalty.

Predicted churn data also informs insights into media partner quality. Brands are leveraging audience breakdown by partner based on churn score distributions. They can deprioritize partners that over-index in users with high likelihood for churn, and focus on partners who consistently deliver users with low churn likelihood and a higher ROI.

Maintain user retention

Predicted behavior modeling offers turnkey marketing intelligence with meaningful actionability. Reengage users most likely to churn with custom messaging via push notifications. Apply audience segments with a low likelihood for churn to create lookalike audiences of high-value users. Or, compare media partners and identify which ones deliver quality users least likely to churn.

To learn more about Kochava Predictive Behavior Modeling and how it can help your marketing efforts, contact your client success manager or support@kochava.com.

Not a Kochava customer? Contact Us today.

Katie Darren – Client Insights Analyst

Kochava

The post Beating the User Attrition Curve with Predictive Behavior Modeling appeared first on Kochava.

]]>