Build a Lowest Cost MarTech Stack

There are several ways to define the term technology stack. From an IT perspective, a technology stack is the set of software that makes up a computer’s infrastructure. 

From a business perspective, a technology stack is all the different software each line of business uses. For example, a manufacturer might use one technology stack to stay on top of their supply chain and another for HR, and so on.

Marketing is a broad field, and every aspect is aided by technology, so the stack that provides the software infrastructure for marketing activities can be very large. In 2018 there were 7,000 different vendors offering MarTech. Unlike what Adobe Creative Suite does for creatives and Microsoft Office does for many office functions, there is no such thing as a MarTech bundle that provides all the tools you need.

According to Gartner, marketing technology consumed 29 percent of the average marketing budget in 2018 (up from 22 percent in 2017) and was the #1 marketing expense. But companies don’t have to spend that great a portion of their AI-related marketing dollars on technology. That’s because some of the best marketing tools are available as free open source software.

Open source refers to source code, the foundational code on which software is built. Many software companies make their original source code available as open source. According to opensource.org, open source software is free under a license that does not restrict how anyone uses the code, as long as they don’t discriminate against people, groups or fields of endeavor. Giving away the source code makes good software popular and creates opportunities for other developers to add features and resell the code as part of a commercial software product. 

Licensors of open source code package their own version of the software and license it for a fee. So, open source software can be offered as both non-commercial (in the original open source version) and commercial (upon adding other elements). 

Linux is an example of a company that made its source code available at no charge and also created products with the source code as the foundation of those products. Eventually, both the non-commercial and commercial sides of Linux were sold for more than a billion dollars each.

Companies can pay for a commercial license of the modified software or use the free non-commercial version of the open source code. But the open source code is generally not as user friendly. It usually requires a business worker who  is going to be able to load and work with it, as is, to have some technical inclination. Alternatively, they need the help of someone experienced with the software to configure and adapt it to the particular needs of their business.

For some companies, licensing the commercial version might make the most sense, while for those that are going to bring in experts to work with the software anyway, licensing the free open source code saves money and time. 

SMBs and mid-market companies that are building a marketing technology stack often find that free open source software operated by data engineers and data scientists under contract is the most cost-effective way to go.

These seven types of technology all have a place in the AI-related portion of a marketing technology stack. That doesn’t mean that you need all seven of them. Each company’s needs are different. Also, some solutions provide more than one technology. The following descriptions are to help you understand what the technology does. This is meant to be a helpful guide, not a shopping list.

01 Data Integration/ETL/Data Ingestion Software

Data integration, ETL and data ingestion are three ways to describe processes that all lead to the same desired outcome: data integration. ETL is the most common (but not the only) way to perform data integration. It involves:

• Extracting data from multiple source files

• Transforming it into the same format

• Loading it into a new target file for fast and efficient use

The takeaway here is that before you can analyze data from different sources efficiently, you have to bring it together in one file. Setup and implementation of all stages of the integration are done by data wranglers, so integration is also known as data wrangling.

Once the data is transformed it is loaded to the target file. At that point it is ready for use in Business Intelligence and Analysis software that incorporates Artificial Intelligence and Machine Learning.

Some of the solutions that perform data integration also provide data visualization and/or data analytics tools.

Xperra recommended: 

• KNIME
• Alteryx
• HPCC Systems

02 Automated Email Marketing Platform

An automated email marketing platform monitors individual customer behavior from a variety of sources, including:

• Whether they open or click in emails you send

• What they browse or buy on your website

• Forms they complete

The platform uses rules you create to score the value of every customer for which you have an email address. The platform automatically sends the correct email campaigns to the customer based on their score, what products and offers interest them most and the email preferences they have selected. 

If you want to send out a new email offer, the platform will use your criteria to determine which customers will be most receptive to the offer and send the email to them. B2B companies can rely on the platform to determine which nurturing campaign each of their leads should be in, if any.

Automated email marketing does the grunt work and allows marketing managers to concentrate on more strategic issues.

Xperra recommended:

• Mautic

03 Customer Data Platform

The customer data platform (CDP) is the single most important tool for developing a single customer view (SCV) of each customer. With an SCV you don’t have to search through your CRM, online purchases, in-store purchases, credit card records and other databases to see everything you need to see about a customer. Instead, the CDP brings all that information together and creates the SCV as a single source of truth for each customer — also known as a Golden Record — which can be customized and shared with each marketing unit. 

With a CDP, you can make sure your customer data is clean and that all duplicates have been removed. Poor data quality is a huge barrier for marketers. It costs US companies billions of dollars a year in marketing ineffectiveness. Sending multiple offers to the same customer due to duplicate records creates a poor customer experience and hurts the brand. 

The CDP makes it possible for businesses to efficiently and effectively market directly to customers.

Xperra recommended: 

• Pimcore

04 Stream Processing aka Real-Time Analytics

The easiest way to explain stream processing is to contrast it to its counterpart, batch processing. Batch processing requires each of the following:

1. Collect and store data

2. Stop data collection 

3. Process the collected data

Once the processing of a single batch is complete you can process the next batch. Eventually, you end up having to aggregate data across multiple batches. You also have to store all the batches to have access to the data you need. If you are batch processing a lot of data you end up with a lot of data in storage, much of which will never be useful.

Stream processing analyzes data in real time, as it is streamed. Because no batch needs to be created, any irrelevant data can be discarded, putting a lot less stress on the system.

Online activity and internet of things (IoT) are two sources of data that are handled far more effectively with stream processing because they have no natural stopping points. 

Stream processing allows businesses to analyze activities in progress rather than waiting for batch reports, which are always historical in nature. Real-time analytics helps uncover potential problems while there is still time to do something about them, Similarly, it uncovers opportunities at the earliest possible point of discovery, enabling businesses to get a head start on new products and processes.

Xperra recommended:

• Dremio
• WSO2 Stream Processor and Analytics Server

05 Business Intelligence & Analytics Platform

Strictly speaking Business Intelligence (BI) and Business Analytics (BA) are two different operations. 

BI came along first, converting data into dashboards and visualizations that present an operational view that is both comprehensive and easy to understand. Dashboards can be customized for different lines of business. Permissions determine what information is available to each worker/manager/executive. A salesperson’s dashboard might give that salesperson an accurate view of how they are doing, while the sales manager’s dashboard will report on all salespersons and allow the manager to dig deep into the data for more granular reporting.

Some BI platforms include the ETL tools needed to integrate data.

BA doesn’t just report the data, but uses Artificial Intelligence and Machine Learning to analyze it, generate data models that predict future outcomes and surface the models with the best predicted outcomes. BA is a system for proactively using data to improve decision making.

Purchasing a legacy BI platform that does not include BA is a half measure that will need to be corrected in very short time — as soon as you realize what you’re missing. A comprehensive platform will include ETL, BI and BA.

Xperra recommended:

• KNIME
• Tableau

06 Machine Learning Platform

Machine Learning (ML) is a subset of Artificial Intelligence (AI). 

ML takes action on specific tasks without needing to be told explicitly to do so. It does that by studying algorithms and statistical models ML recognizes patterns and inferences in sample data and uses them to teach the machine how to handle new data.

Machine learning has become a part of everyday life in so many ways that we tend to not even recognize ML when we make use of it. Here are three examples.

• Google handles more than 63,000 searches per second. As you begin to type your search query, Google gives you multiple options for completing your query. Thanks to ML, Google can predict what you want to search for at a very high rate of success.

• Fifteen years ago, the antivirus software on your PC would identify known virus signatures and alert you to the presence of a virus. Today’s computer viruses and other malware are devised without a telltale signature. To compensate for the lack of signatures, your antivirus/antimalware software uses ML to recognize threatening behavior in your computer and quarantines it before it can harm your machine.

•  Every transaction you put on your credit card creates data on what you buy, where you buy it, how much you spend and more. That data trains the credit provider’s fraud protection system to call out transactions in real time that are very different from your typical shopping pattern. The result can be a denial of credit for a particular purchase until you contact the credit provider to verify the purchase is, in fact, yours.

For companies with a sufficient number of customer records, ML generates and analyzes hundreds of models based on existing data to tell direct marketers which customers:

• … are most likely to respond to a particular offer

• … are least likely to respond to the same offer

• … should be added from a new list because they match current likely customers

Using ML for direct marketing campaigns raises response rates, reduces costs and adds new likely buyers.

Xperra recommended:

• H2O & Driverless AI
• DataRobot

07 Website Analytics Software

Google Analytics is the most popular website analytics tool in use today, but it’s not the only one. Google Analytics has two big things going for it.

1. It’s free.

2. Because it is so popular, it is easy to find people who can use it for you or train someone else on your team to do so.

But Google Analytics isn’t perfect. Here are three reasons why it might not be your best choice.

1. For the non-expert, it’s difficult to use.

2. Google Analytics reports based on a sampling of your data, not all of the data.

3. It doesn’t have all the features provided by other web analytics software.

Those first two items may not be an obstacle for your organization. But some companies benefit from leveraging the additional features that other web analytics software provides. For example:

• See which businesses visit your site

• Conduct A/B tests

• View heat maps to see what content engages visitors

• Access individual visitor profiles

• Monitor competitors

• See data in real time

• Record sessions

• Create markers for each stage of your marketing funnel

• Analyze how visitors interact with your form 

• Enhanced SEO reporting

In addition to advanced features, some Google Analytics alternatives provide the means to integrate your web analytics data into your organizations CRM, forums, photo galleries, webmail and more.

Xperra recommended:

• Matomo

Approaching your marketing technology stack strategically saves a lot of headaches in the long run. Even if you can’t implement it all at once, a strategy that envisions the technologies you will need and how they will work together enables you to develop a phase-in plan for all the different elements. 

By strategically customizing your marketing technology stack you eliminate the roadblocks that can set you back if you take a piecemeal approach. include:

• Saving money by incorporating the right open source software for your company’s circumstances.

• Designing a stack that aligns with your IT plans. If you are moving your servers and data to the cloud, it probably makes sense to use marketing technology that functions in the cloud.

• Ensuring that the data consultants you start with have the expertise needed to help you through the development of the entire stack.

Designing your customized MarTech stack from the beginning is the only way to go.