Despite the hype we've been living in for the last decade, AI/ML is not the ultimate cure for key organizational pain points that data can solve.

The reality is that most companies struggle with having foundational descriptive and diagnostic analytics capabilities, let alone leveraging basic machine learning approaches to solving problems.

Instead of focusing on the algorithms and technology, the right approach is to deeply understand the business problems, walk in the shoes of your functional partners, and solve the problems with data in a manner that your company's analytical readiness will tolerate and that aligns well with the operating rhythm of the business.

A few years ago, I was talking to a relatively nascent data science team at a leading consumer durables company. At that time, the firm was undergoing a large-scale digital transformation, and by all measures, they have succeeded.

My conversation with the data science team then was a great example of the misalignment between DS expectations and the realities of the business environment.

The company needed help improving its Gross Profits through smarter Price Management.

The senior data scientist on the team proposed that they try building a Deep Reinforcement Learning model to optimize Prices across 10K SKUs and 100K Customers.

At that time, the Pricing team relied on dozens of Excel sheets and manual processes to set weekly prices.

Going from that to a Deep Reinforcement Learning Model based Pricing Optimization solution was like making the jump from 1st-grade basketball (where my six year-old-son currently plays) to the NBA (where he has a 0.05% chance of playing 12 years from now).

The data science team eventually chose a simpler approach than RL but still employed complex ML models coupled with Mixed Integer Optimization. Their Price Optimization project lasted about a year, but the company never used it.

Fast forward several years, and the Pricing team still uses random Excel sheets to manage their work.

This data science team was brilliant and highly capable technically. Still early on in their careers, over time, they learned that simple, pragmatic approaches that are co-created with business stakeholders are mostly the ones getting adoption and impact.

Instead of celebrating complexity, let's turn our attention to some of the most impactful Revenue Growth Analytics use cases.

For most organizations, below are the top data-informed capabilities and associated analytics approaches that will drive immediate, measurable Sales and Gross Profit impact for your company in a way that your Sales, Pricing, Finance, and Marketing teams can understand and embrace:

Promotional Analytics

Understand the ROI of your promotional investments, which are often 10-25% of Gross Revenues for Manufacturers (on the higher end for Consumer Product Goods). 

Optimize your spending by eliminating the most unproductive promotions, and reallocate the savings to higher return promos that will benefit you and your Retail partners. 

At most, you'll use some multiplicative regression modeling to estimate Price Elasticities and Merchandising coefficients, but the rest of the work is mostly data engineering and elementary school math. 

This capability can be implemented in 30 days using Excel, Tableau, Power BI, and the like. Year 1 impact can be substantial, often +3 to +7% in Gross Profit improvements.

Everyday Price Optimization

Align your prices with the perceived value your product delivers to Customers (vs. Competitor products). 

After you (or ideally a 3rd party) survey your Customers (could be B2B or B2C) about your products and those of the competition (along with their purchase intent scores), you will ask to get the raw response data. 

You will use either logistic regression or some essential ML to understand to what extent each product attribute drives purchase intent. From that point on, it's basic math to calculate Perceived Benefits and Relative Prices (i.e., Equivalized Prices) to understand whether your Products are Price or Value Advantaged or Disadvantaged. 

Based on that, you can quickly get an idea of how you should right-size your Prices, along with an estimated benefit to your Sales and Profits.

Dynamic, Automated Clearance Pricing

15-20% of Distributor inventory is gathering dust in warehouses and has been largely unsold in the last six months. Most Distributors in the US, especially small- to mid-cap companies, are struggling with manual clearance pricing to eliminate unproductive inventory. 

Using basic regression modeling to estimate historical Markdown elasticities and setting up a rules-based, automated Markdown Pricing engine will significantly impact Gross Profits and cash position. 

It takes about 90 days to build this capability using simple methods that everyone can understand and get behind, leveraging a popular tech stack that Pricing, Supply Chain, and Sales partners are familiar with.

Bulk Purchase Optimization

The primary reason Distributors end up with a large share of Unproductive inventory is that Merchants buy substantially more than what they can sell in a reasonable time.

Building a rapid decision tool for Merchants and Buyers that considers COGS, historical demand, price elasticities, inventory holding costs, and manufacturer vendor rebates to suggest which SKUs and what quantities to buy can bring substantial benefits to your Distribution business. 

The approach is simple, and beyond linear regression or basic ML to estimate Price Elasticities, you can build this analytics capability using simple math. 

Customer Segmentation & Predictive Insights

Knowing who your customers are is a big problem for traditional companies, especially in the Nonprofit space (and most Distribution companies). Please note: having emails, phone numbers, names, and addresses is far from knowing who your customers are and what makes them purchase (or donate) more or less from you. 

Aside from those with a robust Customer Data Platform and CRM, most Nonprofits need help understanding who their donors are and which segments they should target more surgically to drive outsized impact. 

Simple things like collecting emails from your donors, appending them with key demographic and affinity/interest metrics, and combining them with RFM analysis will have a meaningful impact on your Marketing efforts and overall Donation revenues. 

Comparing your donor profiles with an entire geographic market will help you understand where you are over or under-indexed and help you quickly size the incremental Revenue (and Customer Lifetime Value) opportunity by targeting specific segments more aggressively. 

Adding essential ML (think Random Forest) will enable you to build predictive models to understand which donors will donate again within 3/6/12 months and how much so that you can align your Marketing efforts accordingly. 

Customer Churn & Cross-Sell Modeling

Many sales teams can be more efficient and productive with the help of (basic forms of) ML. 

One key area that Sales teams need help with is identifying high-likelihood-to-churn Customers with some causal insights (i.e., why do we think they will churn). It can be accomplished using simple techniques like logistic regression or ML algorithms like a Random Forest or GBM. Serving up a targeted customer list each month to Sales teams will enable them to engage proactively, often months before a potential customer churn occurs. 

A second related area is modeling high-potential customers for cross-sell opportunities based on historical transactional patterns, product affinity analyses, and customer behavioral clustering. 

These capabilities are short-burst analytics exercises that can be done in less than 30 days, giving the sales team a targeted hit list of customers to engage with to increase Sales and Profits.

Summary

While AI/ML has been a buzzword in the business world for a while, it's not always the ultimate solution to data-related problems. Most companies struggle with foundational descriptive and diagnostic analytics or simple ML approaches. To succeed with data analytics, analytics teams must first deeply understand the business pain points and work closely with functional partners to solve them in ways that align with their analytical readiness and operating rhythm.

The key to success lies in practical and co-created approaches with business stakeholders rather than in celebrating complexity for its own sake. The Revenue Growth Analytics use cases mentioned in this post are examples of impactful capabilities that drive measurable sales and gross profit improvements. We can implement them using simple math and essential ML with popular tech stacks with which pricing, supply chain, and sales partners are familiar.

Ultimately, traditional (i.e., non-tech) businesses need to prioritize practical and achievable goals for Analytics and Data Science initiatives rather than focusing on lofty, complex goals that may not be attainable or impactful in the short- and medium-term. By adopting a pragmatic approach, businesses can ensure they're taking full advantage of their data and driving measurable impact for the organization.