Predictive Analytics...Predictive analytics helps you find patterns in the data using historical...
Transcript of Predictive Analytics...Predictive analytics helps you find patterns in the data using historical...
How much time do marketers spend planning, talking, thinking, dreaming, wondering - and not actually doing? Too much time. It can be overwhelming to know what the right move is, especially when there is money and metrics involved. You don’t want to waste money, you don’t want to waste your time, and you don’t want to fail to meet your goals. Universally, the goals that every marketer is trying to reach is more revenue, more awareness. What if you could reach your goals - because you knew what to do. This is where predictive analytics comes into play. In this paper, we’ll walk through how to use predictive forecasts through every stage of the customer experience and align with your goals.
The customer experience is the most important part of your business. Without customers, you aren’t able to meet any of your business goals. Without a solid customer experience, you will have a hard time attracting and keeping new business.
Let’s tackle the customer experience and understand what our audience wants one piece at a time. We’ll explore each step of the customer journey and how you can use predictive analytics to create more effective marketing plans for your customer experience.
First, let’s get on the same page about what a predictive model is and what steps are included in the customer experience.
Predictive Analytics
Predictive analytics helps you find patterns in the data using historical information and projecting it forward. This is accomplished using statistics and machine learning. You use the forecast as a guide to figure out where to go next and not be blindsided by something you weren’t paying attention to. It’s like driving a car. In order to drive safely, you have to look forward. Knowing what is behind you is helpful but not as important as seeing what lies in front of you so that you don’t crash.
Prediction falls into two general buckets: scoring and forecasting
● Scoring builds a model to understand, as best as possible, why something happened
o The most common model is multiple linear regression, which is literally ancient
o The most well-known type is credit scoring ● The goal is to answer the question, “What caused this?” and secondarily, “Is this
likely to happen?”
● Forecasting attempts to predict when something will happen ▪ The most common model is ARIMA, a 1976 algorithm
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
▪ The most well-known type is financial forecasting o The goal is to answer the question, “When will this happen?”
Virtually every major predictive algorithm is available for free in open-source software like R and Python.
To learn more about predictive models, watch Co-Founder Chris Penn's explainer video here: https://www.christopherspenn.com/2018/05/you-ask-i-answer-what-predictive-models-do-you-work-with/
Customer Experience
A customer experience map is a visual representation of the steps someone takes to find out about you and your product to purchase your goods. It is sometimes called the path to purchase, buyer’s journey, owners’ journey, sales funnel, and a whole host of other things. A true customer experience map covers both the buyer's and owner's journey in order to get a full 360 picture.
The goal of understanding the customer experience is to figure out how to reach your audience, have them purchase your product or services, and keep them coming back for more. Ideally, someone would just see your product, buy it, love it, and tell everyone about it. Sometimes it’s that easy but most of the time people need more information and time to make a decision and stay engaged. You can keep your mapping to four simple steps: Awareness, Feet in the door, Opportunities, Customers. This is a great place to start and might be enough for you to map your marketing activities to converting someone through each stage. Google has a really good breakdown of building your own customer journey map. Check it out here. However, to really understand your customer you'll want to look at a more detailed customer experience map, like this example:
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
This version of a journey covers the buyers and owners journey, comprising a very detailed customer experience. The purpose of outlining your customer's journey is to figure out what methods work best to reach your desired audience at each stage and to help alleviate their pain points. You have digital channels such as email marketing, social, search, and paid as well as different tools and techniques at your disposal. Well, so does every other marketer. You're charged with figuring out how to rise above the noise that exists online and make sure that you're reaching the right people at the right time.
Thinking about planning each step of the customer experience can feel overwhelming. That’s where using a predictive forecast can be helpful. In the next post, we'll walk through the planning process for executing a predictive model.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
PLANNING
Let's walk through the steps to put a predictive forecast together effectively.
● Project - Set the strategy, goal, and desired outcome ● Pull - Extract the data from where it lives ● Prepare - Clean, refine, and prepare the data set ● Pick - Identify which variable to predict ● Predict - Create the prediction ● Plan - Build a plan of action from the forecast
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
Project
You could just start running the predictive algorithm on a data set to see what happens, but it’s more efficient to have a plan and know what you're after. Begin by building out the project. This includes setting your goals and understanding your desired outcome. What stage of the customer experience are you trying to get more insight into? If you don’t know that information, start more simply with “what’s the problem I’m trying to solve” or “what’s the question I’m trying to answer?” - once you have that information you can start to build out your requirements. This includes making sure you have the right predictive algorithms and you have access to the right data sets. Your project plan will also help you build out your timeline from start to finish. When is your analysis due (if to a client or stakeholder)? Working backward, do you have enough time for insights and analysis, enough time to clean the dataset and run the algorithm? You should be able to answer all of these questions before getting starting on the data set.
Pull
The next step is to pull the data you'll use for your analysis once you’ve got a good sense of the overall project. There are a lot of good options when it comes to running a predictive forecast. CRM, Marketing Automation, Email Marketing, and social data are all good places to start. You can also use your revenue or accounting data. If you don’t have access to any of that data you can go with data that is publicly available. Some good quality options include, but are not limited to, Google Trends, Data.gov, and Statista.
Prepare
Once you have your data set, you need to clean and prepare it. You’ll want to be aware of any missing data, any anomalies, or anything that could make the algorithm get stuck. You also want to make sure it’s in the correct format. For instance, if the date column came out as a long string, you’ll want to convert that to something more standard like YYYY-MM-DD, Make sure the format matches what lives within the algorithm script. If not, you could spend a lot of time redoing the analysis. You might also find that special characters or emojis (depending on the source of the data) got into the file. You’ll want to make sure that kind of data gets cleaned up and cleared out.
Pick
When you have a clean dataset you need to determine which is the most important variable to predict on. This is where having a handle on the different methods of predictive is helpful. First, you can run a driver analysis to determine the variables that are the most important to your audience. Once you know what is most important, you can run the time-series analysis to determine the likelihood of an event happening.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
Watch this video to learn more about the predictive models: https://www.christopherspenn.com/2018/05/you-ask-i-answer-what-predictive-models-do-you-work-with/
Predict
Now it's time to run your algorithm and get your prediction. There are a couple of options for running a predictive algorithm. You can purchase something off the shelf. At this time, there aren't a lot of great off the shelf options unless you're an enterprise company. An off the shelf predictive model can cost you upwards of six figures. You can build the algorithm yourself. There are free coding tools, such as Python or R-Studio, that can accomplish this. The investment would be the time to learn how to code in those languages. The third option is to hire a consultancy, like Trust Insights, to run the model for you.
Plan
You have your results - you're ready to dive into the insights and build a plan to execute and measure. Are the results what you expected? Are there any surprises? The plan that you build will align with the overall goals you outlined originally in the project. The part of the customer experience that you're focused on will dictate how you build out and execute your plan.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
AWARENESS
Now that you're oriented to the customer experience and gone through the steps to plan out your predictive project, we'll start walking through how to use predictive at each stage of the journey starting with awareness.
Awareness is the first step of the customer experience. People need to find out about you and your services/products. When people are doing their research they are looking to search engines, such as Google and Bing, and various social media platforms. Ideally, they would find you when they start searching for more information. How do you increase your odds? Having content that answers their questions and explains who you are and what you do. How do you know what people are searching for and when to have that content ready? Predictive Analytics using search data.
To demonstrate this. we used cheese. Yes, cheese. Using 5 years of historical Google Trends data and about 150 different kinds of cheese, we ran a time series analysis and
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
projected the search volume forward 365 days from when the data was pulled. This was the result:
What do you do once you have this information? You can start creating your content marketing calendar. Take a look ahead 30 or 60 days, perhaps even longer depending on how you structure campaigns. What are the topics people will care about and then when do you need to start creating content? You'll want to create content that answers their questions. Use a keyword planning tool to get variations on the keywords and dig into the questions people are asking. Let's use "cheddar" as an example. Your audience might be searching for topics such as: what is cheddar, how do I use cheddar, cheddar recipes, how to store cheddar, what are the different kinds of cheddar, where can I get cheddar, etc. You would want to have answers to all of those questions in various forms like blogs, videos, podcast, case studies, white papers, and website content. You'll also want to be sure that all of your content is SEO optimized so it ranks higher in search engines. Once you have that content you can share it across your social channels for people to find.
Think about it in terms of your own services. What do you want to be known for? If you're a shop that specializes in running PPC campaigns for your clients, you'd want to know when people are looking for services like yours. If you run the keyword "PPC" through a keyword planner you might see suggestions such as SEO, Google Adwords, pay per click, and others. You would then extract Google Trends data related to those terms and use that data to forecast when people are likely to be searching week over week. That forecast becomes the content calendar that you'll now work against. No more guessing as to when you should have content ready to go. This calendar will also
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
help you understand when to boost content and pull back on spending in other places that likely won't get as much traffic.
If you have clean, consistent data for your own company, consider using your own site search data to create a predictive forecast. When are people likely to be searching your website for specific content? If you use a dataset, such as your own website traffic, to create a predictive forecast you'll start to understand when traffic is likely to be higher or lower in volume, and when you need to create those awareness campaigns to bring more people in.
Using a predictive content calendar to boost your awareness is both a short-term and long-term solution. You know in the next few weeks what people are looking for but it also gives you the opportunity to plan farther ahead for larger scale campaigns. Timing is everything, why not get it right?
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
CONSIDERATION You've made the audience aware of you, now you need to get them to come back to you and get more information to make a decision. This is known as the consideration phase.
At this point, your audience knows a little bit about you. Perhaps they stopped by your site, they found your social profiles or they signed up for your newsletter. Your next step would be to create an email drip campaign to start sending them more of that smart, targeted content you've been working on in the awareness phase. If you knew when people were not going to open their email you'd probably not send it. Email Opens How do you get access to email open data? You use search data or your own email marketing data. We took a look at when people would be searching for how to set up their out of office message in their email and projected out those trends in a time series forecast. This gives us an idea of when people are most and least likely to check their email at the week level. If you have clean, consistent data from your own email marketing software - even better. That will be more accurate to your particular audience.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
Search Trends for Retargeting Retargeting campaigns are your friend. Make sure you have a retargeting pixel set up on your website, specifically on your product and services pages. When people are doing their research in the awareness stage you have an opportunity to start building out your digital audience. You can use the same search data forecast from the awareness stage to determine how much you should be spending and when. When you build out your ad strategy with your topics and keywords, a predictive forecast can help you figure out when to spend more money. During the consideration phase, retargeting campaigns are the most efficient route because you're targeting people who have already shown interest in you. Your predictive calendar will help you determine what kinds of ads to show and when. “When you build out your ad strategy with your topics and keywords, a predictive forecast can help you figure out when to spend more money.”
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
EVALUATION Now that you've served up a lot of timely, useful information. your audience is ready to make a purchase decision by evaluating your services.
Lead Generation Data You probably have a contact form that gives an option to indicate the reason someone is reaching out. The options are probably to request more information, to sign up for your newsletter, or most importantly, that they want to purchase your services. You can take that form data and run your predictive algorithm on the contacts that have raised their hand to work with you. By doing this you can forecast the peaks and valleys in your lead generation efforts. If you know when you're going to come upon a lull in activity, you can create plans to even out those "down" times.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
Drip Campaigns In the consideration phase, you ran a predictive analysis to understand when people were likely to open their email. You can re-use that data and map it against your lead generation analysis to create drip campaigns. Email drip campaigns are a great way to keep your audience engaged. Drip campaigns can be tailored to the different segments of your audience and what stage they are at. If someone has visited your website a handful of times, downloaded some content and signed up for your newsletter you can create a drip campaign that outlines more specific services that align with their profile. Emails are a really powerful conversion tool within the customer journey. Retargeting Ads Retargeting ads can be used at just about each stage of the customer journey, it's just a matter of changing the context of the ad. At the awareness stage, you want to give people a little more information to encourage them to keep exploring your company. At the consideration phase, you want to point them to your blog, newsletter and other content that showcases your skills and experience. At the consideration phase, you can create ad campaigns that really highlight your services and solutions. By running retargeting ads, you're targeting people who have demonstrated a level of interest in what you have to offer. You can set up rules within your campaigns so that once someone has filled out your services contact form they no longer see those ads. It's a super useful tactic for any digital marketer.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
PURCHASE
Congratulations - you now have a paying customer! Once someone is a customer they have a different set of needs and you have a different set of responsibilities, especially if you want to keep them long term. Let's dig into the second half of the customer experience, the owner's journey.
E-Commerce Data
With customers comes revenue. Revenue powers your business. You want to be able to predict seasonality with your revenue so you can do your financial planning, tune-up spending or pull it back. Using the same predictive algorithm, you can run a forecast against your e-commerce data, getting a better sense of when you can expect the highs and lows of incoming money. You can think about comparing this analysis to your leads forecast from the evaluation stage. This will help you understand the lag between leads and customers, and if the peaks and valleys trend the same way. Having this data will
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
help you stabilize your incoming revenue and make it more predictable for your financial planning.
Customer Profile Data
Converting one customer is great. Converting more customers is even better. Once you have a database of customers and their profiles you can use predictive analytics to figure out how to win more customers. You should be capturing data that tells you when you've won or lost customers, including the "why", "what" and "how". If you pull all of your customer data from your marketing tech stack into one massive spreadsheet it will be nearly impossible to analyze using manual methods. Instead, you can use the driver-analysis algorithm to determine the factors that comprise the profile of a customer that you are more likely to win. This analysis will help you understand the things that are the most important that you need to be doing and offering up. Once you have the variables that outline the most important things to your target audience you can run a time-series analysis to figure out when to have that information in front of other potential customers - increasing your likelihood of conversions.
As we move through the owner's journey, the next phase we'll explore is ownership and dig into your customer support data.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
OWNERSHIP You have steady, predictable revenue coming in. Your customers will have expectations about service and support. Let's dig into the ownership phase of the owner's journey.
Customer Support Data Depending on what kind of product or service you sell, you likely have a customer support system. This might be a ticketing system where a customer can report bugs or issues, or maybe you have a call center where a customer can call and talk to someone when they have questions or want to renew or upgrade their product.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
Call Volume The number of customer support calls your team gets may have a pattern or seasonality. You can assume that some of the spikes in volume will be tied to product releases and major commercial holidays. When you don't know the anecdotal patterns, you'll want to get a sense of when your phones will be ringing off the hook. When the phones will be non-stop you'll want to make sure that you have proper staff coverage. When you're looking at a lull in call volume you can use that time to make sure staff is trained up on the latest product features, policies, and procedures. Ticket Volume Another critical data point to look at is your ticketing system - this is where a customer can report issues. Using the predictive algorithm and your time-stamped ticketing data you can predict down to the 5-minute interval to find out what time of day people are contacting you the most. Maybe all of your tickets are coming in overnight but you don't have staff looking or responding until the morning. This might be a good business case for a chatbot or other automated response system which will allow you to gather more data and keep your customers engaged. This might also be an indication that you need to stagger your support staff by time zones to make sure all of your customers have access to what they need.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
LOYALTY Your customers are well supported, so let's figure out how to keep them long term. This is the loyalty phase of the owner's journey.
Building Brand Loyalists You best customers are the ones that bring in the most revenue for the longest amount of time. The chart below is an example of driver analysis. As you'll recall, driver analysis takes the variables related to your customer (or whatever you're analyzing) to determine which variables matter the most.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
In this example, the variables that matter the most to you are high value and repeat customers. The cluster graph shows your customers broken up into four quadrants, with your high value, frequent customers in the top right corner. Each quadrant has its own set of opportunities. High Value; Frequent Customers This is your most important segment of customers and the ones that need the least convincing to become loyalists. These are the folks that you want to be sure to check in on and keep happy so that they stick around long term. This is the segment that should be getting the "insider information" and special VIP deals. High Value; One-time Customers This segment has made a large purchase and has not returned. They might need some nudging from your marketing efforts. Keep this segment informed of special deals, new products, and features, and make sure they don't have any complaints that your
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
customer support team needs to address. The goal is to turn this segment of your audience into repeat customers. Low Value; Frequent Customers Frequency is important because it means predictable revenue. This segment is making a lot of low-value purchases, which is a good thing. They are loyal but without the budget. Your strategy with this group is to make sure they are shown products and features that match their price points, with the occasional nudge to higher value services. Low Value; One-time Customers This will by far be your toughest group to convert into loyalists. You would want to start by asking them what brought them to your company in the first place and try to build a relationship before doing any hard sells of new things. They might need more information before making any additional purchases so be sure that it's easy to find and understand.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
Customer Loyalty Cycle Smaller cycles will exist within the larger customer experience cycle - namely, the customer loyalty cycle. Using driver analysis, you would look at variables, such as lead scores, returning visitors, tracked customer engagements, and repeat purchases - helping you understand what will keep a customer coming back to purchase more, strengthening your relationship with them as they continually cycle through purchase, ownership, loyalty, and then back around again and again.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
EVANGELISM Whew! We made it to the last phase of the customer experience: Evangelism. Once you've made someone aware of your brand, they have decided to purchase something, you've resolved any issues they have and they become a repeat customer - you want them to spread the word about their experience with you so that you can attract more customer just like them.
Customer Review Data We live in an age where people always check reviews before making a purchase. Customers tend to be hesitant to make a purchase if there is no information about the company or if the company has mostly negative reviews. You have sites like Yelp that allow you to review pretty much any kind of business. Facebook has a feature that allows people to ask for recommendations of services and companies. Products live and die by Amazon reviews. You can use sites like homebuilder to get reviews of contractors from real customers. Using predictive analytics, you can determine when you're likely to get an increase of customer reviews, as well as the general sentiment around the review itself. Customers who are happy and satisfied with their products and
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
service are more likely to leave positive reviews on your behalf, becoming your evangelists.
Review Volume Similar to customer support call volume, you'd want to know when to expect reviews - beyond the expected seasonality. Knowing the peaks and valleys of expected reviews can help you plan out your messaging and marketing strategies. Yelp is one of the most popular review sites where people can leave their thoughts and ratings. You might also look at Google reviews, Facebook reviews if you have a business page or even some of the technical review sites such as G2 Crowd or Capterra. Review Sentiment Machine learning can categorize content into positive, negative, and neutral sentiment. In addition to volume, you'll want to get an understanding of the sentiment with each review. In the example above we kept it simple and took a look at the number of Yelp stars associated with each review. The scale is between zero and five, five being the best rating. if you're looking at a high volume of reviews that are mostly negative you have some research to do, trying to understand the reasoning behind unhappy
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
customers. Perhaps there were issues with the latest version of your product or your customer support staff wasn't as responsive as your customers wanted.
CONCLUSION The thing that most people get wrong about the customer experience is confusing the buyer's and owner's journey. Each part should be treated separately with its individual phases. With proper planning, you can apply predictive analytics to each phase of the customer experience. The end result will be a stronger and more tailored marketing strategy. Predictive analytics is a guideline for your plans. You already have a good sense of the patterns that happen in your industry and how your audience behaves. Using a forecast to inform your planning will help enhance it, make it data-driven, and potentially help you uncover things you didn’t think were there.
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights
Contact Information: For questions about this paper or to engage Trust Insights: email: [email protected] website: www.trustinsights.ai twitter: @trustinsights
Trust Insights : A Data Science Consulting Firm for Marketers https://www.trustinsights.ai | @TrustInsights