Kris

Enabling Mobile App Tracking for Analytics

By Kris on March 21, 2015

Enabling tracking

One of exciting project I’m working on involves enabling mobile analytics capabilities on mobile native apps. Generally speaking, good news is that many tools that we’ve been using in web analytics world support mobile analytics tracking. I’m sure through recent years of mobile growth, I’m sure these companies have worked really hard to bridge the gaps to make marketers data needs reflected in the complex work of mobile app development.

The concept of slapping a tag or pixel on mobile app is quite different in mobile app. You’ll here a lot of the word ‘SDK’ or software development kit. That is because pretty much all the 3rd party analtyics tracker would need to be deployed into the mobile app’s code.

Looks like this.

Google Analytics Mobile SDK example

Source: Tetsuya Shiraishi – Google Analytics Framework

This google’s article on best practices for mobile app analytics is actually informative. I believe it could apply to most of the 3rd party analytics tracking, too.

Track different apps in separate properties.
Track different platforms of an app in different properties.
Track app editions based on feature similarities.
Track different app versions in the same property.

So if you have two different apps and have each in both Google Play and Apple App Store, then it’ll look like this in Google Analytics:

GA property 1: App A & iOS
GA property 2: App A & Android
GA property 3: App B & iOS
GA property 4: App B & Android

Great news for those who have been using Tag Manager. Yes, many tag manager systems support mobile app deployment, hence you can use tag manager to deploy varoius 3rd party trackers without having engineers to code gazillion SDKs. Make sure to check with your vendor and read through the documentations.

You also want to make sure that the tags you plan to install into tag manager can talk with the data objects defined within tag manager, sometimes called as container tag and associated universal data objects.

Campaign Tracking

Event though you might see tag manager saving many of the pain points, that is really not the end of the mobile analytics tracking saga. Especially, some tools are optimized to worked in certain ways, so that the app store data could be integrated with the analytics tools.

Google Analytics for example, in order to get the Google Play or Apple App Store data to talk with Google Analytics, you’ll need to make sure that Google Analytics is set up with proper store identifier ID. That’ll give you valuable information like app installs per campaign sources, where without this set up you’ll lose the source attribution data when users land on the app store landing page.

Here is a good video that describes this process.

Considerations

Although I highlighted Google Analytics in examples here, it is important to understand your business needs and priorities. At the end of the day, most marketers want to know which and how each marketing stategies or tactics are working, so they can make better decisions to improve the marketing spend. That said, various tools are positioned to do well in different areas:

Tracking campaign performance measuring against the installs
In app usage and engagement. (Typically event tracking based tools do well in this)
A/B testing
Campaign attribution analysis
Mobile app platform performance (or server side performance)

Think through and discuss with your team on the needs. You’ll soon learn that to get all the needs out, there are probably going to require many SDKs implemented than what you may have initially thought.

A lot is happening in the mobile app tracking for marketing, so feel free to share your experience here.

Great writings/discussions on mobile analytics

http://www.kaushik.net/avinash/mobile-site-app-analytics-reports-metrics-how-to/
http://www.quora.com/What-does-Mixpanel-do-that-Google-Analytics-is-incapable-of-doing
http://www.analyticshero.com/2013/07/24/web-analytics-vs-mobile-analytics-whats-the-difference/
http://www.kaushik.net/avinash/mobile-marketing-customer-data-acquisition-behavior-targeting/

Key behaviors in marketing analytics career I learned

By Kris on March 8, 2015

I’d like to share some of my lessons learned from my career working with data in marketing, focusing around the things that I tried which I believe added value to the business.

Share what you have learned

In this field of analytics and marketing, it is pretty common to find people who have learned things keep those knowledge and insights for themselves as if that belongs to that person. This is not unique to people in analytics, but that is a pretty lame move in general. Your team and people working in the company are trying to perform, and analytics are important function of this entire business process. The biggest part of analytics people’s job is to support solving business problems through data, evangelize data and insights, support the business move forward achieving it’s goals. Without sharing the data and insights, there is no evangelizing.

Also when you assume what you have shared is being translated to others, don’t assume they have shared with others. Many companies have great internal resources and capabilities that allows employees to share what’s going on. If not, then I’m sorry to hear that. In tools like Jive, it is important to create your own area and start publishing your learnings. You can always reference that source, and communicate to the wider audience. This is obviously not the only way, but the idea is try and find ways to share more than that one point of contact, boss, CMO, etc. What good it is to not share your awesome work and let it sit and decay.

Be curious, build a habit to find answers and apply

Every job, great jobs come with opportunities to learn. Sometimes it may come in a form of unexpected things you have to deal with, or things encountered through new projects, or even come by terminologies encountered through work place, experts, school, and peers. Never be shy to ask, there are no dumb questions. There is Google for us to search wealth of knowledge online as well, and we should take every advantage of it. Up until this point, I hope people do search for their own answers when encountered with questions that you want answers to. Curiosity in a dictionary states the meaning ‘a strong desire to know or learn something’, and for me, this is one of the must have behavior in analytics expert.

Another key thing is to ‘apply’. One of the common characteristics I see in successful people in any industry is that they don’t just consume things, but they apply their learnings and practice what is in their mind to help drive tangible or real impact. Example could be… don’t just talk about tools and wish it get’s purchased or implemented, find ways to bring in a proof of concept deal, and prove the value of the tool.

Communicate and be clear

Communicate your findings, but reserve for the insights. Don’t communicate for the sake of communication. When you communicate, you want to make sure you’re respecting every moment of people’s valuable time. What that means is, be clear as much as possible. Whatever you find through data, you don’t want that to be miss interpreted or taken action in a wrong way from what you originally recommended. If you have found something interesting, but not necessarily gives an answer to the business question, steer the communication to drive testing, or excite others to ‘test it out’. A/B testing tools are so cheap now a days and it is no longer something we should be wait for. Clear communication needs to have a direction, and expected action or outcome that is lead by you.

Always be clear on the problem you’re trying to help solve.

Unless you have some spare time and just playing around with data, don’t waste your time solving and analyzing problems your team, department, and company is not solving for. If a data practice you’re involved isn’t involved in the problem business is not trying to solve, that is a super low priority thing to work on. When analyzing data and concluding your findings, make sure to go back, and ask if the finding helps solve the core business problem or not.

It doesn’t help to be confused on the problem business is trying to solve either. If you’re trying to analyze data for some problem that is way off from the real problem, then that’ll be a huge waste of time. Make sure the problem is clear and you really understand it.

In a knut shell…

I might have learned the hard way to understand the key behavior points listed here (which is not limited to), but hope you find this list of key acumens required in marketing analytics helpful.

My apologies if this came out to be me puking out things many people already know, but I thought it was important for me to put this out and see if this helps for others. Being in this field of data for over 12 years, there is definitely more to it, but I guess I’ll stop here and write more in near future.

Design thinking process and data driven optimization

By Kris on November 9, 2014

As a data person who looks at data to help marketers understand what’s happening on their website or digital marketing programs, it is part of our work to recommend the business on possible next steps and actions to improve the business outcomes.

Going to analytics related conferences and seeing many charts and slides from various experts, it is very common for us to come across data analytics and optimization process diagram.

Usually, those work out really well. It helps us data experts gain a sense of structured process on how to make our data work for our business.

Optimization Process 2013 Nov

In reality, in the Optimization phase, analytics folks would need to work with web designers, UX expert, engineers, etc. These experts are likely going to work or have different thinking process than analytics experts. I won’t be able to cover the entire subject around design thinking process, but I think it would be very interesting to understand how designers think around solving problems.

I won’t be able to cover the entire subject around design thinking process, but I think it would be very interesting to understand how designers think around solving problems.

A typical scenario of data person handing over insights and recommendation to a designer would end up in some ideation or brainstorming session between analytics person and designer to come up with the test version within the A/B testing practice. If the designer has some knowledge around data and how to optimize, then that is great. It makes many conversations very easy, and great designer would be able to digest the analytical findings and run their designer’s thinking process to come up with some great test design.

If the designer has some knowledge around data and how to optimize, then that is great. It makes many conversations very easy, and great designer would be able to digest the analytical findings and run their designer’s thinking process to come up with some great test design.

This is just my limited view on things, but when I attend A/B testing type meet up or conferences around CRO, I feel like the key designers aren’t there participating at the event. Usually, it is some data expert from marketing division or someone who manages the Optimization team, but not so much of actual producer type of designer.

I feel like in modern marketing, all roles should be involved with data and be part of the optimization process. They should be proactively engaged and participate.

I always wondered how the test designs are generated and introduced as something that may work better than the control. A lot of designers are very opinionated and with full of ideas. It is always interesting to hear what designers articulate and explain their design.

I think the first area of things we analytics people should start to look into is the studies behind the design thinking process.

I’m sure there are many kinds of methodologies, so I’ll just introduce what I found from Stanford University’s d school site and good old Wikipedia. I think by checking out some of these thinking, and be informed would be a great starting point to make collaboration between right brain people and left brain people even better.

Some references to understand design thinking process

Design thinking process links:
http://dschool.stanford.edu/redesigningtheater/the-design-thinking-process/
http://dschool.stanford.edu/our-point-of-view/#design-thinking
http://en.wikipedia.org/wiki/Design_thinking

After checking out some of these sources, let me try to understand the what’s being quoted from one of the sites.

“The Design Thinking process first defines the problem and then implements the solutions, always with the needs of the user demographic at the core of concept development. This process focuses on need finding, understanding, creating, thinking, and doing. At the core of this process is a bias towards action and creation: by creating and testing something, you can continue to learn and improve upon your initial ideas.”

Let’s digest this a bit…

“The Design Thinking process first defines the problem and then implements the solutions, always with the needs of the user demographic at the core of concept development.”

Ok, this is great!! Because data people can help prioritize what the problem is from the data that marketers and designers need to focus on.

“At the core of this process is a bias towards action and creation: by creating and testing something, you can continue to learn and improve upon your initial ideas.”

Awesome!! Isn’t this what the CRO (conversion rate optimization), and data people are trying to do as well? To look at data, create a hypothesis to test, and TEST TEST TEST !!

So it looks like in Stanford’s design thinking process, it consists of these 5 steps:

EMPATHIZE: Work to fully understand the experience of the user for whom you are designing. Do this through observation, interaction, and immersing yourself in their experiences.

DEFINE: Process and synthesize the findings from your empathy work in order to form a user point of view that you will address with your design.

IDEATE: Explore a wide variety of possible solutions through generating a large quantity of diverse possible solutions, allowing you to step beyond the obvious and explore a range of ideas.

PROTOTYPE: Transform your ideas into a physical form so that you can experience and interact with them and, in the process, learn and develop more empathy.

TEST: Try out high-resolution products and use observations and feedback to refine prototypes, learn more about the user, and refine your original point of view.

This is great to know. It looks like to me, the connection with data people and design people need to happen a lot around the ‘Empathize’ and ‘Define’ phase. That is where we may different processes in coming up with a hypothesis.

No wonder designers don’t understand what data people say because designer’s ‘Empathize’ phase could totally be in ‘Optimize’ phase in my earlier diagram, where we’ve defined the hypothesis already. When designers seem to create hypothesis in the ‘Define’ phase.

I may be wrong on some details as many people and processes are different from the various business environment but hopefully, you get the idea.

For me, as a data person, who wants to be a leader in my space, would need to understand how your peers think and solve the problem. This is key to successful collaboration and solving problems together.

How to Analyze and Find High-Value Test Ideas

By Kris on October 4, 2014

Recently I ran several tests for a specific team where we’ve seen double-digit improvements in a goal we were measuring and testing against. The method I used to find the “What to test?” question was not that difficult. Over time, I believe it may be more difficult to find more wins, but here are some ideas which I think you can apply in your work to better understand “How to Analyze and Find High-Value Test Ideas”.

A lot of times, when you talk to marketers, they will tell you things like “use analytics”, but I know you folks want to know a little more than that.

Here we go.

1. Find the top potential areas of issues or opportunity for high impact results

The best way to start the conversation with your stakeholders is to understand their business and what is important to them. They are the experts in their field. As an analytics expert, you have to be a great listener.

I’m sure this applies to many of you, a lot of times the analytics may not be set up perfectly to track the key outcomes for the business users. So the first typical thing to check is, rather the analytics tools are capturing the key data that ties to your business outcomes.

Example, PDF downloads, contact sales form submit, order checkout complete, etc. That said, you have to understand what your business’s key end action you want the website to accomplish. Track and analyze all your micro conversions (or end actions)!!

Once your analytics tools are set up to track those end actions, work from the bottom and understand what drives the end action that really matters to the business. The goal is to understand and find key pages that are leading to that end action. So if you’re using Google Analytics, use ‘segments’ to create segments around people who converted or not converted, and analyze the pages people land on, bounce, or leave if they don’t convert.

You’ll soon find out particular key pages that aren’t home page, and start to give you questions that make you want to know what is happening to those pages that showed up on your radar.

2. Understand your users, and capture qualitative data if you can

Never underestimate the voice of the customers. Tools like Qualaroo make it so easy for marketers and analysts to gather user’s voice from the site, and you have to start leveraging the data to better understand customer’s intent, pain points, opportunities, ideas, etc.

User feedbacks in a context of this topic could focus on the pain points or understand if users were able to find the key elements on the page. Like asking “Did you come to the site for XYZ reason?”, and if they answer ‘yes’, then ask them “Were you able to accomplish XYZ” and focus generating questions to gain further insights on why customers could not accomplish.

All these qualitative insights can help you gain a better understanding of the size of the problem, or potentially gain ideas on what to test. Remember that the quantitative data from tools like Google Analytics will only give you what is happening to the site, but doesn’t give you the “Why” things are happening in such way.

Heatmap and session recordings could be a great tool to use to further dive into the details of usability issues as well. Since heatmap and recordings are qualitative data, those are great tools to have in your marketing technology stack.

3. Prioritize your tests based on what is important

By now, you have a list of pages where you think you can improve and test some ideas to make it better. Now, the question to ask is which pages are more important. You could base that importance on several things like (not limited to): traffic volume, close to the end action, high-value touch point, etc.

You could base that importance on several things like (not limited to): traffic volume, close to the end action, high-value touch point, etc.

For example, look for high traffic page with high bounce rate for returning visitors. That could be the first indicator for something to look at and dig deeper. You can also look into surveys to see if there is anything site visitors are complaining about.

Another idea is to look at your valuable traffic segments and the key landing pages that differ to that of the low-value visitors. Give higher priority to the ones where visitors who convert are touching.

These priorities will help you set where the engineering or design resources go.

4. Focus on what you can control

Why work on something that you can not influence. You find an opportunity on a site that requires an entire layout change that requires some kind of engineering effort. That is probably not a good test to start with if there is an alternative, like optimizing a bad content on a landing page. You want to be able to go after the wins you can really win and build up.

In one of my experience, I came across a microsite where the site was owned by another group. However, VOC (voice of the customer) survey indicated some issues on that microsite. Since my team couldn’t immediately impact and take control of that microsite, we then had to focus on what we could immediately take control of. Such as messaging from the website we have authority to change, etc.

5. Look for easy setup, easy changes

Take into consideration of resource, budget, and time factoring it into your assessment. In order to operate and test fast, and to try many things, you’ll have to focus on things that are easy to set up or execute on.

One important factor to think about is the quality of the test, too. I may come to sound like doing a lot of test fast is the way to go, but if you’ve done the steps 1 to 4 really well, then it doesn’t have to be about numbers of tests, given the cost of quality of the test may outweigh the quantity. The main point is doing these steps or homework well so that you identify the are of high-value with potential to give you better bang for the buck, so the test proves your great ideas and ultimately contribute to the business.

Hopefully, these five steps were able to give you some good ideas on where to start and leverage data to build high-value tests. Enjoy analyzing and testing!!

Using Audience Data to Optimize Website Experience

By Kris on August 31, 2014

Integrating audience data is a very powerful way information a marketer could have to plan and reach the right segment of the audience a brand has to go after. As audience data becomes available or able to integrate into tools like web analytics software, companies can now optimize the site experience by understanding three key points here:

1. Type of audience the site is attracting.
2. Type of audience is the e-commerce site converting.
3. Type of audiences the site should attract more due to a better opportunity to convert.

Google Analytics allows Ad Words integration with retargeting enablement as an option. In that process, you can determine if you want to pull in audience data to Google Analytics. Make sure to check with your privacy policy so that these type of data enablement aren’t violating your privacy policy.

1. Type of audience the site is attracting.

First thing marketers typically do is plan their marketing effort around reaching the audience they need to go after. So the 1st point here is really looking at the traffic to your site by the audience type. I’ll be using data from ‘Affinity Category’ as the main examples.

I love using the advanced segmentation. If you have a site with different sections that potentially attract or reach different audiences, then it is worth segmenting your data to a particular section so audience data is applied more specifically to a subset of website’s data. My examples here are actually using particular segments of sessions based on users who have visited a certain part of a site. The reason I did that is I wanted to be specific to a particular category of business the site is offering. (For example, if your site offers watches and necklace, you could be segmenting your data to sessions that include any pages that belong to watches.)

Here is an example. This data shows a segment based on sessions that hit a certain section of the site. Notice the kinds of affinity categories this site section if attracting. You’ll notice that this data is really different when we sort with different data from sessions.

2014-08-31-audience-traffic
2. Type of audience the site is converting. (e-commerce site)

If you sort the data by the transaction, you may see that the site is attracting and converting audiences that are different from above. Hence people who are buying could be different from the audience you’re getting the most traffic from. If you’re paying and targeting for an audience who aren’t buying your service or product, then you might want to rethink your paid media strategy.

2014-08-31-audience-transaction
3. Type of audiences the site should attract more due to a better opportunity to convert.

Again, if you re-sort the data by conversion rate let’s see what we have. It is very interesting to see that we have another different set of audience type on top of the list. Cat lovers…

This tells you where you need to acquire more traffic and the kind of audience you could be targeting more of to drive higher sales or business outcomes. I guess we’ll have to acquire more cat lovers who are local news junkies.

2014-08-31-audience-convrate

One can argue that the audience data may not be accurate. However, it is what Google AdWords are likely going to be using to help you target your ad, and if it is this data that is being recorded saying the cat lovers are likely going to convert, why not take advantage of it.

What marketers need to be aware is that chasing after an audience that does not fall in the target audience the brand is chasing after is a debate you need to have internally within your company/team. Performance is one thing, but if you’re in a business where audience type matters to your brand and the overall impact of how your brand is perceived, then it is a very important discussion that the entire marketing team needs to be aligned on. So that one department is not chasing after an audience for the sake of conversion, but the debate about how you can strategically attract and convert the right kind of audience for the brand and measure it. In that case, perhaps your goal is to make the all that list we reviewed to have certain audience dominate the top 3.0, but think about it. First, let’s analyze your data.

Digital marketing is changing really fast. What digital marketing had a hard time to measure and integrating data such as the demographic or psychographic data is becoming something of a common practice. Looking at A/B testing tools and where it is going at integrating 3rd party audience data is something to look out for, but it won’t be surprising when you see more of this with web analytics as well.

Bonus

Other interesting ways to use this data is to segment affinity category data and see where this audience land on the site and behave. You might have thought you were reaching that audience on a certain section of your site, but it may turn out you have more work to do to place the right message for the right audience on places you never put many efforts on. Maybe this something I can write more about next time.

Finding Customer Pain Points on your Website

By Kris on April 29, 2014

One of the common questions optimization (CRO) experts get is, how do you find customer pain points on your website so that it could be fixed or tested to improve site experience?

Common answer you may get back is finding customer’s pain points on the website and then go fix it by using analytics tool like Google Analytics, Site Catalyst, Webtrends, etc. However, the art of finding the customer’s pain points is usually deeper than what you may expect. Here are some ideas, but would definitely love to hear from you if you have any.

Look deeper into error page visitors

The 404 Error Page is something all brands should worry about. There is an unavoidable situation like when I entered some random URL in Google search which gives me the following result. However, if your visitors are hitting this page through existing link on your production site, or your marketing campaign is driving to a 404 page, then that could be a super lame experience for the people visiting your website.

google404

Things to consider in reviewing your error page visitors are by taking a deeper look (or segmenting) the 404 visitors:

What URLs people used to arrive at 404
What traffic sources are leading to this error page
What landing page is driving traffic to this error page? Not all people land on 404 page directly, could be a particular page with an expired link. Could be many other things.
Details around the technical environment. Certain browser, device type, OS, time of day 404 error page increases?

Use VOC and segment data on people who could not Task Complete

What website owners think as an error or failure point may be different from how customers see it. If you have some kind of survey or feedback capture mechanism on your site, then you might want to cater questions that could help you understand customer pain points.

VOC Task Completion for Mobile

Usually, the survey tool allows you to serve the survey on any page you want, or if you have it served across many pages, the data export gives the URL where the survey was taken.

So in this example, you might be able to find pages with high responses from people who said ‘No’ to ‘Were you able to find the information you were looking for?’. Now you have a good starting point for trying to formulate a good hypothesis on why a particular page sucks, and what needs to get optimized.

One of my favorite analysis and observation points is Clicktale and iPerceptions integration, where I get to see session recordings of people who left negative feedbacks. For example, survey respondent to the iPerceptions survey who have rated the site experience at 0 (0 out of 10, 10 being great), Clicktale will allow you to see user behavior from those who rated 0 on the customer satisfaction survey. You may find something interesting about the site that you did not know about. How awesome is that, to be able to understand what customer exactly went through?

Drop off between form submit and actual form submit completed

It is very common to tackle optimizing some kind of behavior funnel, or form submission funnel to reduce people dropping off in a certain step. Example, trying to increase shopping cart closure rate.

What don’t we hear much from this type of analysis is that what if people click on that last ‘Submit’ button, but the data did not go through? In eCommerce example, it is that situation when you’ve reviewed all the information you put in and ready to checkout, the site freezes, or you hit that final ‘confirmation’ button but not sure if you’ve hit the thank you page. This is quite different from page to page drop off because the user’s intent to close the sale is ‘real’, not one of those fuzzy last minute change of decision.

This is a good point to check your data because you can use your analytics tool to track such drop-off. It is pretty common in CRO world to optimize websites from the bottom of the funnel, which is likely to generate quick great CRO results.

Site speed or loading time

It shouldn’t be a surprise that the faster the site serving content to users, the lower the bounce rate would be, or maybe even the site conversions. Here is an example of average document interactive time (DOM readiness) versus bounce rate. You’ll definitely see a causation between the speed and bounces. There are things you can’t do much as a brand, such as improving one country’s network infrastructure. But most brands should be able to control common technical things on the site, ranging from minimizing code, reducing the weight of the image assets, etc.

site-load-time-bounce-ga

Monitoring your site’s performance and constantly thinking about avoiding customer having to wait for the content is definitely something to optimize. Assessing this data per site section, by platform, or page level could give you a better perspective where customers are feeling the pain.

Top landing page with high bounce rate

Usually, this is the first thing most analysts look at to understand which landing pages are the popular entry point for the site visitors but yet have high bounces. Typically, most analysts would dig deeper in potentially the problematic page. Common data to dig deeper to understand the drivers of the high bounces are:

– Traffic sources that are driving traffic and high bounces (example sources… traffic from social media, traffic from banner ad, etc.)

– Device type. Is it mobile or tablets that are driving high bounce?

– Timing. Could it be due to some timing with the press release, product announcements, promotional discounts?

– Country and network speed. Like some emerging country with expected slow network speed may have high bounces

– The specific segment that bounces. You may have different segments built out in your analytics tool that could be based on a combination of things that defines some kind of group. Example… Support user from those who purchased before and came back to support page, campaign ad visitors who have returned to site for the 3rd times, or returning visitors who came from email, etc.

– Technical environment. Like browser and the version of the browser. Don’t be surprised if you get few particular browsers on a certain version that gets high bounces. Talking with web engineer may help you understand why certain browsers can lead to higher bounce rate.

In summary, there are so many things you can do with your analytics tool, survey tool, and Heatmap tool to understand customer pain points. It is not just about finding out what is working and obsessing about improving conversion rates. In many situations, it is important to think of the inverse scenario. Like looking at data on those who did not convert, asking and finding what’s broken, a collection of pages not adding value, error page visits and why people visit that page. It is these unsexy things that marketers/analysts should be thinking about that could help build a better understanding of customer pain points. So that you can build a great hypothesis to test/experiment or just fixing the website immediately.

Kris

Register to get latest updates on Marketing Technology & Analytics!