The Best Programming Languages for Digital Marketers

In the world of data-driven marketing, more and more tasks require a bit of coding. You might need to add an extra parameter to your tracking code or pull raw data from Google Analytics. You may want to create a simple prediction—or automate a few repetitive tasks in your PPC campaigns. Or maybe you simply want to […]

The post The Best Programming Languages for Digital Marketers appeared first on CXL.

In the world of data-driven marketing, more and more tasks require a bit of coding. You might need to add an extra parameter to your tracking code or pull raw data from Google Analytics. You may want to create a simple prediction—or automate a few repetitive tasks in your PPC campaigns.

Or maybe you simply want to speak a common language with your developers so that you can brief them better and understand they may say, “This will take two weeks.”

Regardless, knowing one or two data languages—even at a beginner level—is a great help for an online professional’s daily job and, consequently, huge advantage over your competition.

Which programming language should marketers learn?

For a newcomer, there are four programming languages worth learning:

  1. SQL*
  2. JavaScript
  3. Python
  4. Bash

*Technically, SQL is a “declarative language,” not a programming language, but it has the “functionality of a mature programming language.”

Of course, you don’t have to learn all four at once. Even learning one will solve many, many problems that you never could’ve imagined solving before. Besides, after learning your first programming language, it’s usually much easier to learn the rest—you’ll already understand the underlying computer logic.

But which language is good for what? Who should choose which one? It depends on several things:

  • the tasks you work on;
  • the tools you work with; and
  • your goal with the given language.

In this article, I’ll answer all these questions and provide an overview of the four most popular languages for digital analytics professionals.

But first… Why do we have more than one programming language?

First, there are historical reasons. They’re similar to why we speak different languages in different countries. Python, SQL, JavaScript, and Bash were created by different computer scientists in different circumstances. Eventually, as other people started to use these languages, they found their own way to grow.

The second reason is more important. Python, SQL, JavaScript, and Bash are good at different things. You can solve some tasks in JavaScript that you would never be able to solve in SQL. For example, it’s easy to create predictions in Python but really difficult in Bash.

Every language will have its advantages and disadvantages, so you should choose wisely based on the questions you want them to help answer.

Let’s see the languages one by one!


Best at: data analysis; quick and simple data queries—even on multi-million-line datasets; joining data tables efficiently

Marketing goals: accessing in-depth user data from your company’s database; running advanced queries on Google Analytics data, etc.

Difficulty: 3/10

Get started with: SQL For Data Analysis Tutorial Series

SQL is a relatively simple language. It stands for “Structured Query Language,” but I like to call it “Excel on steroids”—it highlights its essence. SQL is great for running queries on really big tables (even tables with more than 10 million rows). It can get a massive job done, sometimes in seconds, while Excel often keels over with just a few hundred thousand rows.

Handling million-line datasets efficiently is one of the greatest powers of SQL, and it’s the main reason that many companies store a significant amount of their data in this format. The trade-off is that SQL doesn’t have a user-friendly point-and-click interface; we have to access our data by writing code-based queries.

Here’s the difference between a SUM() function in Excel (top) and SQL (bottom):

excel sum function

sql sum function

For me, SQL is the best “entry language” to the world of coding because it’s simple and straightforward.

How SQL solves real-world marketing problems

Create more accurate, detailed reports

One of my former colleagues, Rob, is an SEM guru—with zero coding background. Years ago, he endured continual data discrepancies among Google Adwords, Google Analytics, and internal company reports—sometimes greater than 10%.

Rob didn’t know which report was the best guide for his next move. (Everyone who works with PPC tools has probably experienced the same problem at least once.) He also didn’t understand how internal reports were built, and the developers—as usual—had a more important project than helping him debug them.

One day, he got so annoyed that he decided to learn SQL by himself. He picked up the knowledge he needed in less than two days. (In my experience, it’s quite feasible for almost anyone to reach an intermediate level in SQL in that amount of time.)

After that, he could easily query the company database and find the bugs. He was also able to build more accurate, more detailed reports since he had the business knowledge and the coding skills.

Create refined segments with Google Analytics data

Digital analysts should also know that BigQuery, a Google product, lets you use SQL to manage Google Analytics data.

For instance, you can join different Google Analytics tables to create reports that aren’t possible in the Google Analytics point-and-click interface. Other use cases include diverse, data-heavy tasks such as querying the metadata from 1 billion taxi rides or back-end monitoring of analytics for real-time fraud detection.

The standard Google Analytics user interface lets you use primary and secondary dimensions to create micro-segments by, for example, looking at the number of sessions by landing page and source.

But, if you know SQL, you can segment your Google Analytics data further by adding a third and fourth dimension (e.g. location and device type). Plus, you can add advanced filters (like counting only the top 20% most-active users) and run advanced calculations (like using the median instead of mean), and many, many more things.


Best at: implementing tracking codes; every kind of web-development solution

Marketing goals: improving accuracy or granularity of tracking for better data quality

Difficulty: 8/10

Get started with: Codecademy’s JavaScript course

JavaScript is one of the most frequently used programming languages in web development (besides HTML and CSS). If something pops up, animates, auto-scrolls, sparkles, or changes, it’s probably written in JavaScript.

But, for digital marketers and CRO professionals, it’s even more important to know that almost all tracking codes are written in JavaScript.

Here’s one you’ve almost certainly seen before:

gtm container code
The Google Tag Manager container—and most tracking snippets—use JavaScript.

That code snippet—used to implement Google Tag Manager—is in JavaScript. So are the code snippets for Google Analytics, Hotjar, Crazy Egg, Facebook, Reddit, Google Ads, DoubleClick, and many other tools. To insert these tracking codes onto your website, you don’t need to know much about JavaScript; most tools have simplified the process to a copy-and-paste action.

However, to set up advanced tracking—like scroll-depth or cross-domain tracking—a basic knowledge of JavaScript can ensure proper implementation. With more advanced applications of JavaScript, you can automate repetitive tasks in Google Ads, pass UTM parameters between different websites, and take advantage of countless other opportunities.

JavaScript is definitely more difficult than SQL. It can take a few weeks of 1–2 hours of daily learning to build a solid foundation. But that’s more than enough to understand and modify your tracking scripts, simplify automation, and, most importantly, improve communication with developers.

How JavaScript solves real-world marketing problems

Connect inventory with ad campaigns

One of my clients had an ecommerce company with more than 100,000 products. They constantly added new products, discontinued old products, ran out of stock, or filled up their storage house with best sellers.

The business was so complex that their ads couldn’t keep up— they should’ve started and stopped Google Ads campaigns every minute, but it was impossible to do that manually.

So, one of their marketing interns(!) started to learn JavaScript with the kind help of a mentor from the web team. With about one month of learning and practice, he was able to set up a JavaScript automation that passed information between the company database and Google Ads.

After that, the campaigns started and stopped automatically based on the company’s inventory data. It saved their team countless hours of work, reduced human errors, and saved ad spend—they no longer wasted money promoting out-of-stock or discontinued products.

Enable robust A/B testing

JavaScript has a critical role in conversion optimization, too. jQuery, a JavaScript library, powers the HTML/CSS changes for A/B testing.

Marketers without jQuery skills rely on testing tools’ visual editors to design and publish A/B tests, often with bad results.

Either the lack of coding skills limits their ability to test what matters or over-ambitious changes break the site.


Best at: predictive analytics; machine learning; APIs

Marketing goals: running advanced analytical methods on datasets; making predictions to guide future planning

Difficulty: 7/10

Get started with: Python for Data Science Tutorial Series

Python is a favorite language of data scientists. In many aspects, it’s similar to JavaScript. (The two languages have about the same level of complexity.) Even the syntax looks similar. For example, compare the “if” function in JavaScript (left) with the one in Python (right):

python vs javascript syntax comparison

The big difference, though, is that Python is primarily for “back-end” and analytics tasks, not “front-end” tasks (e.g. website development).

From an analytics professional’s point of view, the greatest advantage of Python is the many analytics extensions written specifically for data-science tasks. If you want to run predictive analytics or a machine learning project, you will probably work in Python.

Additionally, many popular digital analytics and marketing tools offer access to APIs. (A quick, non-technical explanation of APIs is available here.)

How Python solves real-world marketing problems

A few recent projects come to mind:

  • At many startups, we used Python to predict whether users would cancel their subscription in the near future—then reached out to them proactively.
  • At a larger company, we used Python to estimate the expected increase in support tickets and planned our quarterly hiring on those numbers.
  • At a very small company, we used Python to scrape huge walls of text automatically and analyze media mentions.
predictive analysis python
You can run advanced analytical projects, like this predictive analysis, in Python.

There’s another example, too, albeit one that’s a bit out there. In early 2018, I wrote code in Python to connect to a cryptocurrency service provider’s database using their API solution.

Still in Python, I queried the exchange rate of Bitcoin every minute, and, still in Python, I tried to fit a statistical model to predict the exchange rate change for the next 10 minutes. (Before you get excited: I tested over 10,000 models, and none gave an accurate-enough prediction.)

But you can run similar projects using your Google Analytics data, Google Ads data, or any internal company data—if you know Python. I’d recommend starting with a few weeks (1–2 hours per day) to learn and practice the syntax, and to get familiar with its marketing-oriented extensions.

It’s a realistic goal to run your first (simpler) predictive analytics project in Python at the end of a month. Of course, don’t expect to become a senior Python developer from one week to another, but at least you’ll be able to communicate with developers much efficiently. Additionally, if you’ve already learned JavaScript, learning Python will be easy. (It works the other way around, too.)

Note: You may have heard about another popular language similar to Python called R. It’s used mainly by mathematicians and statisticians, and its syntax is a bit more difficult to learn compared to Python. Because of that and other reasons, I recommend Python over R for digital analysts and CRO professionals. However, some “hardcore” statistical packages are available only in R.


Best at: moving files, automating scripts, connecting other languages

Marketing goals: automate reporting

Difficulty: 5/10

Get started with: Data Science at the Command Line

Bash is a “bonus” language—and it’s not really famous, either. But you should know that it’s the built-in language of every computer that runs on an Ubuntu/Linux operating system and, thus, the language of most data servers, too.

Bash is not the first language you should learn, but it’s not difficult. The basic view is the classic “command line” design. Many programmers like it: It looks cool (a question of taste, of course), and it offers clear access to communicate with the computer on a core level.

bash command line
Bash offers clear, deep access to data servers’ operating systems.

How Bash solves real-world marketing problems

I use Bash for three specific tasks (mostly in data-science projects):

  1. Moving, copying, cleaning, and re-structuring data files on a remote data server.
  2. Automating scripts (e.g. pulling user data, formatting it, and saving it to a server automatically at midnight).
  3. Connecting Python and SQL scripts and running them together.

For instance, in the early days of Prezi (my first workplace), we built the entire data infrastructure using only Bash, and it handled the data of 3 million users—for the above tasks and reporting—very well.

You can learn the basics of Bash in one or two days. However, it makes sense only if you’ve already learned SQL or Python (or both).

What’s the best way to learn a programming language?

To learn these languages, there are plenty of free and paid resources, online and offline. (I’ve linked a resource at the top of the section for each language.) Whether you prefer books or workshops or even interactive courses, any will guide you effectively through the theoretical basics.

But the most important part of the learning process is practice. Find a small project to apply your freshly gained coding skills to demonstrate the real-world problems you can solve and boost your motivation. Practice will also deepen your knowledge.

Here are a few pet-project ideas:

  • Python: Build a bot that can play Blackjack.
  • SQL + Bash: Scrape articles on news portals and try to collect similar articles from different portals.
  • JavaScript: Try to implement click tracking on your website (or on a test website).

These are just my ideas. By all means: Find something that interests you and that you would enjoy building—then build it!

Even if you don’t plan to become a practitioner, programming knowledge may be essential to help find the right candidate for your marketing department.

How can you find and hire professionals with these skills?

It’s a tough question.

A digital marketer or CRO professional who has a business mindset and also knows how to code is rare. Why? Because they’re two different skill sets. More and more companies, however, have realized the importance of coding in digital analytics, and the demand for these talents is higher than ever.

The easiest way to get the right person on board is to find someone internally—either a programmer who’s interested in business and marketing or a digital marketer who’s open to learning how to code. With the right training programs, it’s not too hard to transition a motivated talent’s career path.

How to interview candidates for programming positions in marketing departments

If you find a good candidate from outside your company (it’s not impossible), make sure that the interview committee has at least one technical person and one business person. It’s not uncommon for these applicants to go through 6–8 well-designed interview rounds, including a take-home assignment and whiteboard technical screenings.

When you look to hire someone, your main goal should be to understand if she can apply her knowledge in practice. Asking theoretical questions like, “Please explain the role of indentation in Python,” or, “How would you calculate standard deviation?” won’t help.

As an applicant, this was a “tell” that the committee had no idea why they wanted to hire me. (And, as a result, I turned down their offer at the end of the process.) The best questions are 100% practical and make the applicant think! For example, “You have these two data tables with X and Y user information. Who are our power users? Why? Write a query for it!”


There’s nothing magical about SQL, JavaScript, Python, or Bash. Anyone can learn the basics of these languages, whether your goal is to become an entry-level practitioner, improve communication with your development team, or make it easier for you to hire the right people.

To get started, pick the one that has the best potential to answer your most important marketing questions, then start learning! After the first few hours, you’ll see that it’s not as complicated as you may think—and how useful it can be in your everyday job as a marketer.

The post The Best Programming Languages for Digital Marketers appeared first on CXL.

Mixpanel vs. Google Analytics: The 2018 Guide

This post is not a dry feature-by-feature comparison, nor does it include a winner-take-all verdict. Your business won’t benefit from either of those things. Instead, we’re comparing Mixpanel and Google Analytics in the terms that drive business growth—identifying the core use cases for each tool and the business problems they solve, while highlighting the features […]

The post Mixpanel vs. Google Analytics: The 2018 Guide appeared first on CXL.

This post is not a dry feature-by-feature comparison, nor does it include a winner-take-all verdict. Your business won’t benefit from either of those things.

Instead, we’re comparing Mixpanel and Google Analytics in the terms that drive business growth—identifying the core use cases for each tool and the business problems they solve, while highlighting the features that make it possible.

Anything else is merely a list of data points. That’s as useful as analytics without analysts: troves of data but no actionable insights.

The core use cases

  • Google Analytics is the standard for measuring acquisition—identifying the sources of traffic to your website or app. Google Analytics also tracks on-site behavior through events and goals. It does not, however, de-anonymize data. User ID tracking allows you to track the behavior of individual users, but their identities remain unknown.
  • Mixpanel, in many ways, picks up where Google Analytics leaves off. It has robust, user-centered tracking that connects company CRMs to the online behaviors of real people—and enables you to send targeted messages to them, at a group or individual level. Mixpanel’s event-based tracking is fundamentally different than the Google Analytics pageview model.
google analytics users vs sessions
Google Analytics switched its default metric from “sessions” to “users” in 2018, mirroring Mixpanel’s emphasis on users over pageviews.

Most businesses, even Mixpanel’s paying customers, retain Google Analytics. It’s free, after all, and, at the very least, offers the chance to corroborate data across two platforms.

  • When Google Analytics makes the most sense: If your business relies on its website solely for marketing purposes—to attract visitors and generate leads—Google Analytics provides most of your actionable data. You’ll be able to see which channels (paid, organic, social, etc.) have the highest conversion rate and identify the content that earns the most interest from your target audience.
  • When Mixpanel makes the most sense: If your website or app is your product, however, Mixpanel offers the granular detail that’s essential for monitoring user behavior. You’ll be able to see which acquisition channels are best for long-term retention or lifetime value, not simply those that drive initial conversions.
  • When it could go either way: Other companies—like ecommerce sites selling physical products—may straddle the use-case gap. The size of their business, the number of products, the length of the buying cycle, and other variables may determine whether Mixpanel can deliver a strong ROI.

In the end, the more user data you have to push into Mixpanel—and the more that data shapes your business decisions—the more value you’ll extract from it.

The business problems that Google Analytics and Mixpanel solve

g2 crowd reviews mixpanel google analytics
Hundreds of user reviews of both platforms on G2 Crowd reveal the key benefits—and shortcomings—of each. (Image source)


1. “We don’t know where our traffic is coming from.”

G2 Crowd, a business technology review site, has almost 3,000 reviews of Google Analytics. When we analyzed all of them, the patterns were easy to spot, one phrase above all others: “…traffic is coming from.”

That phrase alone—there were other, similar ones, too—appeared 53 times. Google Analytics excels at revealing your traffic sources. In the words of hundreds of reviewers, Google Analytics was essential to “identify channels,” “track campaigns,” and monitor “traffic flow.”

“Google Analytics is definitely better at measuring traffic,” noted Dan McGaw, the founder and CMO of Effin Amazing. Other digital marketing experts I asked seconded his opinion. Google’s dominance of the ad market has bolstered its primacy for acquisition metrics—integrations with Google Ads (search and display) are comprehensive and seamless.

Mixpanel also tracks acquisition sources but, in its tracking and reporting, emphasizes what those users do, not the raw visibility of web pages. As Mixpanel’s Aaron Krivitzky explained:

[Cost per Acquisition] and [Cost per Click] are important, but they tell you nothing about user retention, they tell you nothing about lifetime value, and they tell you nothing about the actual end-user sentiment, behaviors, or experience.

For high-traffic sites or short-lived apps, the need to understand real-time acquisition paths and performance offers a point of separation between the two tools.

2. “We need to know how our site (or app) is performing right now.”

A small subset of businesses may have unique acquisition strategies that depend on real-time adjustments to copy, design, or resource allocation. All Mixpanel reporting is real-time; Google Analytics has real-time reports that, after a few hours, filter into core reports.

Fully integrated, real-time access to data has potential benefits for large media sites:

  • The Huffington Post tests multiple headlines for articles. After a few minutes of data—a meaningful amount, given the site’s traffic volume—they discard the less-popular headline.
  • Real-time analytics help streaming video services shift resources based on demand. If, say, a season finale or critical moment in a sporting event risks overwhelming servers, real-time analytics can tip off technical teams to the need before issues affect users.

Real-time reporting can also help large ecommerce companies manage their products. “Mainly it’s an inventory issue,” according to Steve Kurniawan of Nine Peaks Media, “although it can also help other things like tracking product deliveries and negative reviews.”

There are other use cases, too, such as app developers seeking to maximize the value of users for a viral game with a lifespan of days or weeks. Real-time data can validate near-constant changes to keep users engaged.

Still, the use-cases for real-time reporting are limited. For most Mixpanel users, the platform’s most meaningful feature is its ability to tie actual prospect and customer data to online behavior, providing greater insight into the experiences that affect engagement, conversion, and retention.

Engagement, conversion, and retention

3. “We can’t connect analytics data to our real-life customers.”

“Mixpanel is a great tool for tracking user behavior and acting on it,” affirmed McGaw. Realizing those core benefits—tracking and acting—requires stitching together online actions at two levels:

  • From device to device. Google Analytics has cross-device tracking. However, its system relies on cookies and User IDs, and includes only those “who have opted in to personalized advertising.” For companies with a SaaS platform or app, that system may not be enough. In McGaw’s experience, “If you have a web app and mobile app, Google Analytics is pretty shitty at tying those users together.”
  • From user to human. Google Analytics’ User ID policy prevents this second connection, providing the sharpest break between its capabilities and those of Mixpanel. By connecting companies’ user data to its analytics, Mixpanel can build analytics cohorts based on CRM data and push messages to high-value segments.

What do these differences look like in the real world? Saas consultant Sid Bharath gave an example:

Let’s say Google Analytics shows that paid search has the highest conversion rate. With this data alone, it seems like you should double down on paid search.

However, if you had Mixpanel in there, you could see how those converters interacted with your software. So you could filter by paid traffic, and it may show that these customers actually churn at a much higher rate, or they’re not as engaged as organic customers.

At Mixpanel, Krivitzky has seen potential clients come to him with similar business needs: “How do my end-users use my app? Which features are sticky? How common is X use case versus Y?”

In addition to identifying high-value cohorts, Mixpanel also lets businesses send targeted messages to those same users—a benefit noted regularly across more than 200 product reviews from G2 Crowd. “With Mixpanel’s cross-device marketing automation suite,” noted McGaw, “You can email, text, push, and pop up across devices to provide a pretty amazing experience.”

Mixpanel custom messaging
Mixpanel allows businesses to send messages to real people based on online behaviors.

4. “We don’t understand our customer journey well.”

Both Google Analytics and Mixpanel offer conversion funnels. Mixpanel funnels—based on feedback from experts and hundreds of reviews—are more user-friendly to create and more granular in their segmentation. (Notably, Mixpanel funnels also apply retroactively to data.)

Google Analytics funnels and Behavior Flow visualizations lack the customization of Mixpanel funnels, which allow segmentation by user characteristics.

Creating segmented user groups based on broad demographic data is possible in Google Analytics; however, it requires applying a segment to the Funnel Visualization, rather than embedding that segment in the funnel itself.

mixpanel funnel
Mixpanel users can set up custom funnels by selecting any series of user behaviors.

The customization and comparative intuitiveness of Mixpanel funnels—which can be set up simply by defining a series of steps—were a common highlight in user reviews:

“Easy ability to create funnels”

“Easy to set up events and funnels”

“Easy to set up powerful user funnels”

“Easy to understand user interface for funneling”

In contrast, Funnel Visualizations and Behavior Flows in Google Analytics were a frequent frustration: “Trying to set up goal funnels is an exercise in mania.”

google analytics funnel visualization
Google Analytics funnels are notoriously difficult to set up and interpret.

The difference between technical capacity and ease of implementation is not trivial. It speaks to the importance of actually using analytics tools—not just having access to the data within them—and hints at some of the data-centric problems the C-suite often looks to solve.

C-suite level

5. “We don’t know the impact of each marketing channel.”

Cross-channel attribution extends beyond first and last clicks. For large organizations, stronger attribution can help demonstrate the ROI—or lack thereof—for multi-million-dollar television campaigns or in-store promotions.

The Google Analytics 360 Suite offers extensive attribution tracking. “Some clients ask us to design and implement marketing attribution models for them,” recounted Kurniawan. “In these cases, we use Google Analytics 360 Suite, which offers an amazing custom marketing attribution feature.”

In contrast to standard Google Analytics, which uses a last non-direct click attribution—thereby attributing 100% of the conversion to some online source—Google Analytics 360 accounts for television ads’ impact on online behavior and also folds in-store purchases into conversion data (crediting online sources for offline purchases).

That benefit comes at a cost—starting at $150,000 per year. For many organizations, no matter how accurate the insights, the ROI simply isn’t there. (Mixpanel, whose paid version starts at $999 per year, charges by the number of data points tracked; costs can quickly scale into a four-figure monthly fee, especially for SaaS or app companies that offer freemium models.)

6. “We need to make data collection more efficient.”

In some cases, the technical capacities of Google Analytics and Mixpanel are similar. The value of one tool over the other depends on the technical abilities of those using the platform.

Several Mixpanel users—in in-house and agency settings—viewed the user-friendly components of its analysis as true cost savings because they reduced the amount of human labor necessary to create and interpret reports.

Nishank Khanna, the CMO of Clarify Capital, explained why they added Mixpanel to their analytics suite:

We were using Google Analytics for years, until it became a chore to track custom event-driven data. As our business needs for analytics grew more sophisticated, we needed to easily define events to track, ensuring a flexibility that promotes focused and meaningful analytics.

We found that in Mixpanel, and the time saved was day and night.

Khanna’s experience wasn’t unique. As Kurniawan confirmed, the ease of implementation is, at times, more important than the technical differences:

Technically, for event tracking versus pageview tracking, both Mixpanel and Google Analytics can do them very well given enough effort. So it’s a matter of ease of use: Setting up event tracking is significantly easier on Mixpanel opposed to Google Analytics.

In our case, since we set up marketing systems including analytics for clients, some (well, most) of them are not tech savvy. Ease of implementation is very important so we don’t have to go back and forth often.

Throughout user reviews for both products, definitions of “easy” and “hard” varied. Complicating matters, an “easy” setup often meant easy data collection—not easy interpretation.

The paradox was clear. Some Mixpanel reviewers lauded its easy setup; others lamented its challenges. Google Analytics was “great if you have no experience” but also only useful “if you have experience.” Just because you can gather data quickly doesn’t mean you’ll know what to do with it.

No product feature or technical capability can replace human interpretation, at least not yet. Both Google Analytics and Mixpanel have rolled out systems to automatically push alerts or pull insights from their platforms.

Mixpanel (left) and Google Analytics (right) have both rolled out automated alerts to make data more accessible.


“Mixpanel is most valuable for companies who need to track revenue over a long period of time, or really know how their users use their product,” summarized McGaw.

In other words, What percentage of your business decisions are driven by things that happen after a website visit or form fill?

If the stopping point in Google Analytics is only the midpoint (or less) for your customers’ online experience, Mixpanel can extend and deepen the analytics portrait—tracing individual users or cohorts all the way back to their initial interaction.

Still, the question of which analytics platform to use doesn’t hinge on these two alone. Others, like Amplitude, Kissmetrics, and Heap, also gather and aggregate user and product data.

All require an investment beyond the tool itself—one in human capital—to translate data into the product improvements and user satisfaction that make analytics meaningful.

The post Mixpanel vs. Google Analytics: The 2018 Guide appeared first on CXL.

Five Strategies for Slaying the Data Puking Dragon.

If you bring sharp focus, you increase chances of attention being diverted to the right places. That in turn will drive smarter questions, which will elicit thoughtful answers from available data. The result will be data-influenced actions that result in a long-term strategic advantage. It all starts with sharp focus. Consider these three scenarios… Your […]

The post Five Strategies for Slaying the Data Puking Dragon. appeared first on Occam’s Razor by Avinash Kaushik.

If you bring sharp focus, you increase chances of attention being diverted to the right places. That in turn will drive smarter questions, which will elicit thoughtful answers from available data. The result will be data-influenced actions that result in a long-term strategic advantage.

It all starts with sharp focus.

Consider these three scenarios…

Your boss is waiting for you to present results on quarterly marketing performance, and you have 75 dense slides. In your heart you know this is crazy; she won’t understand a fraction of it. What do you do?

Your recent audit of the output of your analytics organization found that 160 analytics reports are delivered every month. You know this is way too many, way too often. How do you cull?

Your digital performance dashboard has 16 metrics along 9 dimensions, and you know that the font-size 6 text and sparkline sized charts make them incomprehensible. What's the way forward?

If you find yourself in any of these scenarios, and your inner analysis ninja feels more like a reporting squirrel, it is ok. The first step is realizing that data is being used only to resolve the fear that not enough data is available. It’s not being selected strategically for the most meaningful and actionable insights.

As you accumulate more experience in your career, you’ll discover there are a cluster of simple strategies you can follow to pretty ruthlessly eliminate the riffraff and focus on the critical view. Here are are five that I tend to use a lot, they are easy to internalize, take sustained passion to execute, but always yield delightful results…

1. Focus only on KPIs, eliminate metrics.

Here are the definitions you'll find in my books:

Metric: A metric is a number.

KPI: A key performance indicator (KPI) is a metric most closely tied to overall business success.

Time on Page is a metric. As is Impressions. So are Followers and Footsteps, Reach and Awareness, and Clicks and Gross Ratings Points.

Each hits the bar of being “interesting,” in a tactical oh that’s what’s happening in that silo soft of way. None, passes the simple closely tied to overall business success standard. In fact, hold on to your hats, a movement up or down 25% in any of those metrics may or may not have any impact on your core business outcomes.

Profit is obviously a KPI, as is Likelihood to Recommend. So too are Installs and Monthly Active Users, Orders and Loyalty, Assisted Conversions and Call Center Revenue.

Each KPI is of value in a strategic oh so that is why we are not making money or oh so that is why we had a fabulous quarter sort of way. A 25% movement in any of those KPIs could be the difference between everyone up and down getting a bonus or a part of the company facing layoffs. Often, even a 5% movement might be immensely material. What metric can say that?

When you find yourself experiencing data overload, don an assassin's garb, identify the metrics and kill them. They are not tied to business success, and no senior leader will miss them. On the ground, people will use metrics as micro diagnostic instruments, but they already do that.

A sharp focus on KPIs requires concentrating on what matters most. Every business will have approximately six KPIs for a CEO. Those six will tie to another six supplied to the CMO.

After you go through the assassin’s garb process above, if it turns out that you have 28 KPIs… You need help. Hire a super-smart consultant immediately!

2. Focus only on KPIs that have pre-assigned targets.

This is a clever strategy, I think you are going to love it.

Targets are numerical values you have pre-determined as indicators success or failure.

Turns out, creating targets is insanely hard.

You have to be great at forecasting, competitive intelligence, investment planning, understanding past performance, organization changes and magic pixie dust (trust me on that one).

Hence, most companies will establish targets only for the KPIs deemed worthy of that hard work.

Guess what you should do with your time? Focus on analysis that is worth your hard work!

Start by looking at your slides/report/dashboard and identify the KPIs with established targets. Kill the rest.

Sure, there will be howls of protest. It'll be John. Tell him that without targets you can’t identify if the performance is good or bad, a view every CEO deserves.

John will go away and do one of two things:

1. He will agree with you and focus on the KPIs that matter.

2. He will figure out how to get targets for all 32 metrics along all 18 dimensions.

You win either way. :)

An added benefit will be that with this sharp focus on targets, your company will get better at forecasting, competitive intelligence, investment planning, org changes, magic pixie dust and all the other things that over time become key assets. Oh, your Finance team will love you!

Special caution: Don't ever forget your common sense, and strive for the Global Maxima. It is not uncommon for people to sandbag targets to ensure they earn a higher bonus. If your common sense suggests that the targets are far too low, show industry benchmarks. For example, the quarterly target may be 400,000 units sold. Common sense (and company love) tell you this seems low, so you check actuals to find that in the second month, units sold are already 380,000. Suspicion confirmed. You then check industry benchmarks: It is 1,800,000. WTH! In your CMO dashboard, report Actuals, Target and Benchmark. Let him or her reach an independent, more informed, conclusion about the company’s performance.

3. Focus on the outliers.

Turns out, you are the analyst for a multi-billion dollar corporation, with 98 truly justifiable KPIs (you are right: I'm struggling to breathe on hearing that justification, but let's keep going). How do you focus on what matters most?

Focus your dashboards only on the KPIs where performance for that time period is three standard deviations away from the mean.

A small statistics detour.

If a data distribution is approximately normal then about 68 percent of the data values are within one standard deviation of the mean, about 95 percent are within two standard deviations, and about 99.7 percent lie within three standard deviations. [Wikipedia]

By saying focus on only reporting on KPIs whose performance is three standard deviations from the mean, I’m saying ignore the normal and the expected. Instead, focus on the non-normal and the unexpected.

If your performance does not vary much, consider two standard deviations away from the mean. If the variation is quite significant, use six (only partly kidding!).

The point is, if performance is in the territory you expect, how important is it to tell our leaders: The performance is as it always is.

Look for the outliers, deeply analyze the causal factors that lead to them, and take that to the executives. They will give you a giant hug (and more importantly, a raise).

There are many ways to do approach this. Take this image from my January 2007 post: Analytics Tip #9: Leverage Statistical Control Limits

Having an upper control limit and a lower control limit makes it easy to identify when performance is worth digger deeper into. When you should freak out, and when you should chill.

Look for outliers. If you find them, dig deeper. If not, move on permanently, or at least for the current reporting cycle.

Use whichever statistical strategies you prefer to find your outliers. Focus sharply.

4. Cascade the analysis and responsibility for data.

In some instances you won't be able to convince the senior leader to allow you to narrow your focus. He or she will still want tons of data, perhaps because you are new or you are still earning credibility. Maybe it is just who they are. Or they lack trust in their own organization. No problem.

Take the 32 metrics and KPIs that are going to the CMO. Pick six critical KPIs for the senior leader.

Cluster the remaining 26 metrics.

You'll ask this question:

Which of these remaining 26 metrics have a direct line of sight to the CMO’s six, and might be KPIs for the VPs who report to the CMO?

You might end up with eight for the VPs. Great.

Now ask this question:

Which of these remaining 18 metrics have a direct line of sight to the eight being reported to the VPs, and might be KPIs for the directors who report to the VPs?

You might end up with 14 for the directors.


Repeat it for managers, then marketers.

Typically, you'll have none remaining for the Marketers.

Here's your accomplishment: You've taken the 32 metrics that were being puked on the CMO and distributed them across the organization by level of responsibility. Furthermore, you've ensured everyone's rowing in the same direction by creating a direct line of sight to the CMO’s six KPIs.

Pat yourself on the back. This is hard to do. Mom is proud!

Print the cascading map (CMO: 6 > VPs: 8 > Directors: 14 > Managers: 4), show it to the CMO to earn her or his confidence that you are not throwing away any data. You've simply ensured that each layer reporting to the CMO is focused on its most appropriate best sub-set, thus facilitating optimal accountability (and data snacking).

I’ll admit, this is hard to do.

You have to be deeply analytically savvy. You have to have acquired a rich understanding of the layers of the organization and what makes them tick. You have to be a persuasive communicator. And, be able to execute this in a way that demonstrates to the company that there’s real value in this cascade, that you are freeing up strategic thinking time.

You’ll recognize the overlap between the qualities I mention above and skills that drive fantastic data careers. That’s not a coincidence.

Carpe diem!

5. Get them hooked on text (out-of-sights).

If everything else fails, try this one. It is the hardest one because it'll demand that you are truly an analysis ninja.

No senior executive wants data. It hurts me to write that, but it is true.

Every senior executive wants to be influenced by data and focus on solving problems that advance the business forward. The latter also happens to be their core competence, not the former.

Therefore, in the next iteration of the dashboard, add two more pieces of text for each metric:

1. Why did the metric perform this way?

Explain causal factors that influenced shifts. Basically, the out-of-sights (see TMAI #66 if you are a subscriber to my newsletter). Identifying the four attributes of an out-of-sight will require you to be an analysis ninja.

2. What actions should be taken?

Explain, based on causal factors, the recommended next step (or steps). This will require you to have deep relationships with the organization, and a solid understanding of its business strategy.

When you do this, you'll begin to showcase multiple factors.

For the pointless metrics, neither the Why nor the What will have impact. The CMO will kill these in the first meeting.

For the decent metrics, it might take a meeting or three, but she'll eventually acknowledge their lack of value and ask you to cascade them or kill them.

From those remaining, a handful will come to dominate the discussion, causing loads of arguments, and resulting in productive action. You'll have known these are your KPIs, but it might take the CMO and her team a little while to get there.

After a few months, you'll see that the data pukes have vanished. If you've done a really good job with the out-of-sights and actions, you'll notice notice that the focus has shifted from the numbers to the text.

Massive. Yuge. Victory.

If more examples will be of value, I have two posts with illuminating examples that dive deeper into this strategy…

Strategic Dashboards: Best Practices, Tips, Examples | Smart Dashboard Modules: Insightful Dimensions And Best Metrics

You don't want to be a reporting squirrel, because over time, that job will sap your soul.

If you find yourself in that spot, try one of the strategies above. If you are desperate, try them all. Some will be easier in your situation, while others might be a bit harder. Regardless, if you give them a shot, you'll turn the tide slowly. Even one month in, you’ll feel the warm glow in your heart that analysis ninjas feel all the time.

Oh, and your company will be data-influenced — and a lot more successful. Let's consider that a nice side effect. :)

Knock 'em dead!

As always, it is your turn now.

Have you used any of the above mentioned strategies in your analytics practice? What other strategies have been effective in your company? What is the hardest metric to get rid of, and the hardest KPI to compute for your clients? Why do you think companies keep hanging on to 28 metric dashboards?

Please share your ideas, wild theories, practical tips and examples via comments.

Thank you.

The post Five Strategies for Slaying the Data Puking Dragon. appeared first on Occam's Razor by Avinash Kaushik.