Scalr.ai

3 Best Ways To Use Ai in Business intelligence for Marketing Success In Your Organization

Matt Payne
·
August 24, 2021

Business intelligence is the most powerful way to understand key business critical metrics around your business, both in the past and in present.  Understanding your business intelligence focus points as well as what data analytics you collect that make up your BI is incredibly important, as these metrics should allow you to take massive leaps forward in your business. With that being said, marketers in 2021 are entirely focused on business data and analytics to understand everything about their marketing campaigns, and how to optimize them to produce results. Artificial intelligence is being used right now to power these same business intelligence analytics that allow marketers to achieve better results in campaigns, as well as collect valuable data along the way that can drive a business's overall metrics. Understanding how to use these Ai tools to create and optimize business intelligence for your marketing is key in today's algorithmic world.

What Is Business Intelligence For Marketing?

Layout of a business intelligence dashboard for marketers

Business intelligence tools for marketing are data driven information points that help marketing professionals run higher ROI and more efficient campaigns. Tools like data visualizations and ai models can help marketers predict future results, optimize current social media campaigns, visualize the results of marketing campaigns and much more. These bi tools allow marketing campaigns to understand deeper metrics about their target audience, and how to carefully create a campaign that returns the highest ROI for the organization. Like other forms of business intelligence, these data analytics and visuals are driven from raw data that is captured and becomes quantifiable through marketing data driven processes. We're going to look at using ai and machine learning to further the capabilities of our data analytics and marketing efforts to produce even higher ROI results than previous attempts at business intelligence in marketing.

Recommendation Systems As An Email Marketing Campaign Game Changer  

Popular email marketing metrics

In a world where 59% of buyers saying that marketing emails influenced their purchase decisions, and every dollar spent on email marketing returns an average of $42, email marketing is an industry you can't afford to neglect. Recommendation systems allow us to hyper personalize our emails to each user based on predictive analytics that allows us to target the right potential buyer with exact products. Companies like Amazon have implemented these same strategies (seriously, check it out) and credit them with a huge portion of their revenue.  These machine learning models are a great way to turn raw business and customer data into powerful business intelligence dashboards that can drive entire email marketing strategies.

How Do We Create BI Solutions With Data Insights and Recommendation Systems?

Deep learning based recommendation systems for marketing (like one we built *here*) require a few inputs to make the high ROI insights come to life. We're going to use ecommece and SaaS as our example niches for this. Here's an overview of the fields we need.

  1. Company data surrounding customers' interactions with the company. This can be past purchases, pages viewed, time spent on site, following on social media etc. The closer the data interaction is to our final goal (getting sales!) the better results it will give us towards our final goal.
  2. Data information that makes up the specific products. This allows our model to learn deep trends in the way products are similar, and the way users interact with them. Data from multiple sources that make up a product (customer experience reviews, product title, product description) work the best at reaching our end goal.


Once we've established our feature points we want to use to reach our marketing analysis goals, we build and train our recommendation system to focus on learning the relationship between specific customers' actions and what items they eventually purchase. This model is plugged into a business intelligence dashboard and input pipeline and gives marketers predictions for what products past and potential customers want to buy. Using this business intelligence, email marketing campaigns can send personalized marketing emails just like Amazon!

3 Real Examples Of Recommendation Systems That Are Used For Business Intelligence


These three recommendation systems are perfect for incorporating into your business intelligence strategy when focusing on email marketing initiatives.

Neural Collaborative Filtering For Marketing

Read More About Our NCF Model


This machine learning architecture allows you to make product and service recommendations based on only two data points that we know you already have! Using just customers and their past purchases we build BI models that predict future purchases for any user in your email list. This model let's companies quickly and easily get involved in business intelligence tools that ease email marketing decision making and integrate Ai into their business.


Deep Interest Evolution Network

Deep Learning Architecture for Click Through Prediction

For companies that have a higher level of business data, we can use a deeper level model that was built just for click through prediction. DIEN is an ecommerce and advertising focused model that fits well into our business intelligence for marketing focus. Although this model requires many more data points and a wider variety of sources, its accuracy in understanding the click through rate of an advertisement piece is top level. We've built this exact model for multiple marketing niches like email marketing, landing page optimization, social media ad conversion rates and much more. You can learn more about here.

Item to Item Recommendations For Marketing

The item to item recommendation model was originally laid out by Amazon as a way to recommend products based solely on an understanding of other products. Using your existing business intelligence tools you can easily integrate this model into your dashboard and use the recommendations all over your sales funnel. Its one of the original marketing and sales based recommendation systems and has driven Amazon billions of dollars in revenue. I highly suggest reading through the original paper released by them.

Predicting Future Trends In Sales With Forecasting Models

forecasting sales in power bi
Forecasting Sales In Power BI

Forecasting in marketing business intelligence is the use of predictive analytics to forecast future data values for many different business critical metrics:

  1. Ad Spend
  2. Add To Cart Leads
  3. Ad Campaign Performance
  4. Leads From Social Media Platforms
  5. Product Margins (If you're using dynamic price optimization)
  6. Number of Sales

Using the understanding of what the outcome of these data points looks like, marketers can create a strategy to improve these metrics. The power of using these models for BI dashboards is we can adjust the datapoints and see the forecasted future trend change in real time. This empowers businesses to tweak dashboard inputs like product price or ad spend and see how the product margins change, or the number of leads based on ad spend. We can incorporate real time customer data as our campaign goes along to see how we compare to the forecasted data, and make adjustments to improve our campaign.

Forecasting Customer Leads With Predictive Analytics

By understanding the actionable insights that go into a given lead generation funnel, we can use a machine learning based forecasting model to optimize our lead generation channels by tweaking the different parameters that make up our funnel. To build and use this model for business intelligence we need to break down the components.

Identify The Inputs

We first need to understand which parameters are important to the funnel we want to forecast. This is a simple task for a marketer who understands what analytics matter in a certain funnel.

  1. Web analytics like Google Audience
  2. Number of landing pages that we are using
  3. Any customer data collected
  4. What defines a successful lead
  5. Adjustable variables such as landing page size, CTA

Collect Funnel Process Data

We need to collect or retrieve these data points that make up successful customer behavior in our funnel. We'll use this historical data as a backbone for the forecasting model to predict future results. Once we have this data we can move to training a forecasting model to predict our lead generation.

2 Powerful Machine Learning Models For Forecasting

Two different machine learning architectures that we can use for forecasting in our business intelligence dashboards.

CNN Models For Time Series Forecasting

forecasting future data with a cnn model

Using the popular image recognition model we can define a 1 dimensional model for multivariant time series forecasting. These deep learning models work perfectly as marketing campaign models as they are built to handle a wide range of inputs. This software can be packaged up and used as its own pipeline tool or added to a dashboard that let's you visualize your predicted leads.

Sequence To Sequence Models For Forecasting

sequence to sequence encoder

These text and speech domain models can be easily converted to a time series based model that works much better than the slightly outside LSTM for forecasting. These NLP models originally focus on predicting the next word or character in a sentence based on the previous words, so you can see how they would be able to translate to predicting the next data value. They also take advantage of an important algorithm called beam search which I highly recommend you read more about to understand the conditional probability that makes this a great model for predictions.

Clustering Algorithms As Business Intelligence Tools To Segment Customers

Many marketing initiatives look to split wide groups of customers and their attached data into smaller groups based on clustering similar data points. Clustering campaigns (also called marketing segmentation) is a great way to understand what makes up a target audience in terms of various customer data points. Multivariate clustering allows marketing departments to create dashboards and reporting channels in BI tools like Power Bi and Tableau that let you visualize these segmented customer groups.

Popular Ways Marketers Can Use Clustering in Business Intelligence

  1. Customer browsing activity (using web analytics and tools like hotjar to track on page behavior)
  2. Customer insight based on specific items purchased
  3. Keyword rankings info to analyze SEO campaign benefits
  4. Comparing Offline vs Online company sales analytics
  5. Clustering customer purchase data to learn more about the customer experience


power bi dashboard with clustering
Use Business Intelligence software like Power BI to Visualize

Popular Clustering Algorithms Businesses Can Use Right Now


Free to use machine learning models companies (or marketers) can use right now to improve decision making processes based on real analytics.

Affinity Propagation: A personal favorite of Scalr.ai, is a complex unsupervised clustering algorithm that works best for our marketing domain. The algorithm contains many key benefits with the main one being the ability for the algorithm to choose the number of clusters based on the data provided, without you needed to specify how many clusters you want. This allows us to let the algorithm work its magic and show us how the data should be split up. We've used this algorithm across different marketing domains and are experts in integrating it into BI visuals and reporting.

K-Means: The most popular general purpose algorithm is a classic in the world of machine learning. Its fast runtime performance and simple use of the distance between points to decide clusters make it a favorite when starting out in the world of Ai insights. K-Means does require you to provide a number of clusters but does scale very well with lots of samples. This is the most popular method of customer segmentation, but does require an understanding of your goal final output in terms of # of clusters.

BIRCH: BIRCH is a higher level clustering algorithm that is built to scale in size and is perfect for large company insights tasks like that of social media ad performance, large email marketing campaigns, long term value analysis, and general market research tasks. BIRCH algorithm holds accuracy and runtime performance for increases in both # of clusters and samples.

Looking To Integrate Ai in your business intelligence? Talk to Scalr.ai

We build custom ML/Ai models and business intelligence software tools for businesses looking to gain deep enterprise insight into the business critical data that makes up your organization. These custom software tools allow you to easily reduce costs, boost profit margins, and streamline decision making processes. Let's talk today about transforming your raw business insight into custom business intelligence and powerful analytics.