Data Science Consultant
The majority of high ROI data science projects fall into the predictive analytics camp. That’s why there’s so much interest and buzz around this field. I’ve implemented predictive analytics solutions for my clients around several business cases. What I want to do with this series of posts is present some of these uses. In later posts I’ll detail the available approaches to predictive analytics with pros/cons to each. No math or jargon…I promise.
The “why” behind predictive analytics falls into a few high level categories:
- Digital marketing that leverages predictive analytics reduces the cost of marketing while improving its effectiveness.
- Industrial maintenance that leverages predictive analytics reduces the unexpected downtime in factories.
- Predictive analytics helps build products with higher margins and adoption rates.
- Predictive models can show the long term impacts of business decisions making strategy more certain.
- Team building, talent management and hiring can leverage predictive analytics to improve productivity and employee satisfaction.
This field of data science presents business with an opportunity to gain competitive advantages that many are taking advantage of.
The term, digital marketing, encompasses a lot of different online marketing techniques. The tools available to marketers have expanded greatly. That’s brought to light some of the challenges which predictive analytics address.
The biggest challenge of digital marketing is in where, who and how:
- Where to market – what channels, what sites, what platform, etc.
- How to market – what message, what visuals, etc.
- Who to market to – which customers are most likely to buy or convert
Many marketing departments I’ve worked with are making the transition from descriptive analytics (reports that show how customers are acting now) to predictive analytics.
Descriptive analytics supports a try, measure, improve process flow. The marketing department tries a new campaign, measures effectiveness and then tries something new to make the campaign more effective. Predictive analytics supports a measure, improve, do process flow. The marketing department measures to create the predictive model, runs simulations to determine the most effective marketing campaigns and then executes the recommendations from the model.
The results I’ve seen from predictive models are typically 2 to 3 times as effective as digital campaigns using descriptive analytics alone. The cost savings are also significant. I’ve seen highly effecting marketing groups save 20% to 30% while still achieving their goals after making the switch. For groups that are new to digital marketing or are challenged by it the savings and improvements are significantly higher.
Customer Lifetime Value
Let’s talk about what that means in real numbers and metrics. Customer lifetime value (CLV) is the new yard stick for marketing. Converting and retaining customers with high CLVs and increasing or optimizing CLV are the two most profitable goals of any marketing team. The next metric is the customer journey. This is a customer’s path from discovery through their lifetime with the brand. Predictive analytics makes both of these metrics possible.
As the name implies, CLV is a forward looking metric and needs a predictive model to determine its true value. It’s also a personalized metric. Some customers are optimized for your brand meaning that they come to a brand with a high CLV. Others need to be optimized and come to a brand with a low CLV. This is also true of existing customers; some have high CLVs and are your targets for retention marketing while others are low CLVs and are targets for your optimization marketing. Then there are those customers in the middle of the pack who need a mix of optimization and retention.
The first job of a predictive model is to identify new customers and categorize their CLV. This is where programmatic and remarketing can be key tactics in your overall strategy. Using the predictive model, you can quickly categorize or predict a customer’s CLV which tells the business how much it should spend marketing to them. This is a big element in the savings and return equation. Marketing stops using the spray and pray approach and starts targeting customers with the appropriate spend based on their ROI. This answers the “who” question from above.
Customer Journey Map
Customer journey maps answer the how and where. A customer journey map (CJM) is a predictive model in and of itself. It predicts:
- Where customers will discover your brand
- How they want to be converted to a brand
- How and where they want to interact with your brand
- When they’re most at risk to leave your brand
- How a marketing campaign will impact their buying habits in the short term and long term
A mature CJM is highly personalized. However, most marketing teams I’ve worked with start out with a very general CJM. It describes where customers by large segment are most likely to be introduced to a brand organically or doing what they do without prompting. This includes content marketing, social media marketing and more traditional forms of advertising targeted by platform (mobile, TV, print, etc.) and location (websites, programs, time of day, etc.). The CJM changes the paradigm of discovery by focusing it on just that, new customers being introduced to the brand. Brands are able to put their marketing budgets to work for them by introducing themselves, organically to customers with the highest lifetime values.
The CJM also describes conversion. How do customers want to be converted to a brand? Is it via email or through a promotional offer? Do they want to be educated about the brand’s personality to form a connection? Creative marketers have always known that conversion is an art. CJMs give creative types the data to prove their case for highly innovative approaches to customer conversion as well as measure the ROI of those campaigns.
Finally, the CJM will help marketers manage the customer relationship throughout its lifetime. It helps to unify their brand experience throughout the journey. It helps the business identify gaps and understand how to create a seamless omnichannel experience. It shows a business how to navigate times when customers are most at risk of being converted by competitors. Marketers have a number of tools at their disposal to optimize and retain customers. The CJM shows the business which tools and what timing is most effective. This targeted approach reduces costs while improving the impact of marketing dollars.
A mature CJM is highly personalized. The predictive model running behind the scenes is able to create a 1 on 1 conversation with the customer. Content feels like it was written for them by talking about their needs and showing customers how much they have in common with the brand. Social media campaigns touch on their key emotions, encouraging sharing and engagement. Email personalization goes well beyond putting their name in the subject. It feels like someone at the company took the time to write them a personal note. Their experience across channels is familiar and makes them the center of their journey. While a high level CJM is a competitive necessity for marketing the mature CJM is a competitive advantage.
There are many more granular types of predictive models in marketing. The customer lifetime value and customer journey map are the two with the most strategic impact. When it comes to ROI these two predictive models bring the largest returns in my experience. There are others that can address specific challenges like shopping cart abandonment or high customer churn rates. For clients that haven’t embraced predictive analytics yet, these are a great starting point in the migration from descriptive to predictive.
Making The Transition From Descriptive To Predictive When The Business Is On the Fence
Many companies are on the fence when it comes to predictive analytics. There are individuals within the business who see the value but that view hasn’t been evangelized to reach the rest of the business. In many cases that I’ve been involved with, the marketing group is the driver for companywide adoption but they’re up against an entrenched mentality which leads to heavy resistance. Results are the only way, in my experience, to change that mentality.
Smaller, more granular predictive analytics initiatives are a good way to quickly get those results. Show the board of directors ROI from these initiatives and they’ll give you all the funding you need. These initiatives can usually fit into the existing budget so they don’t require special approval. They take between 3 and 6 months to implement and start yielding returns as soon as they’re applied.
The process of selecting which predictive analytics initiative to start with is all about pain points for the business. What’s your biggest headache as a marketing group? Is it churn, engagement, conversion or something else? Target the first initiative on that pain point. Don’t try to solve the problem and don’t let new problems creep their way into the scope of the initiative. Set a goal for improvement and stay focused. An initiative that goes over budget or doesn’t meet the goal doesn’t build the business case you need.
Remember why you’re doing this and evangelize the business need. Most companies are reacting to the predictive movement in digital marketing. They only do so after a business disruption like a loss of market share or failed product launch. Don’t let that be your business. If you’re reacting, you’re playing catch-up and that’s expensive. Your customers are expecting more personalized experiences. Canned doesn’t meet their expectations and the worst way for a business to learn that is by losing them to competitors.
The final piece of the funding puzzle is the presentation. Show the goal, the costs and the returns. In three slides you can get all the funding you need for the next round of initiatives. Don’t get caught up in the numbers or charts. Don’t get lost in the technology. The point is the ROI so don’t feel like you have to defend or present the approach.
My next post will cover the industrial internet of things and how the data from these devices is powering predictive analytics for manufacturing. Be sure to share your thoughts in the comments section.