Part one in all a comprehensive, practical guide to CLV techniques and real-world use-cases

Whether you’re an information scientist, a marketer or an information leader, chances are high that in case you’ve Googled “Customer Lifetime Value”, you’ve been disillusioned. I felt that too, back after I was leading CLV research in an information science team within the e-commerce domain. We went in search of state-of-the-art methods, but Google returned only basic tutorials with unrealistically manicured datasets, and marketing ‘fluff’ posts describing vague and unimaginative uses for CLV. There was nothing in regards to the pros and cons of obtainable methods when applied in on real world data, and with real world clients. We learned all that on our own, and now I need to share it.

Presenting: all of the stuff the CLV tutorials not noted.

On this post, I’ll cover:

• What’s CLV? (I’ll be transient, as this part you almost certainly already know)
• Do you really want CLV prediction? Or are you able to start with historic CLV calculation?
• What can your organization already gain from historic CLV information, especially once you mix it with other business data?

In the remaining of the series, I’ll present:

• Uses for CLV prediction
• Methods for calculating and predicting CLV, and their benefits and drawbacks
• Lessons learned on easy methods to use them appropriately.

And I’ll sprinkle some data science best-practices throughout. Sound like a plan? Great, let’s go!

Customer Lifetime Value is the worth generated by a customer over their ‘lifetime’ with a retailer: that’s, between their first and last purchase there. ‘Value’ will be defined as pure revenue: how much the client spent. But in my e-commerce experience, I discovered that more mature retailers care less about short-term revenue than they do about long-term profit. Hence, they’re more likely to contemplate ‘value’ as revenue minus costs. As we’ll see partially two though, knowing which costs to subtract is simpler said than done…

Experienced R&D teams know that for brand spanking new data science projects, it’s best to begin easy. For CLV, this will be as ‘easy’ as using historic transactions to calculate lifetime value up to now. You may:

• calculate a straightforward average over all of your customers, or
• calculate a mean based on logical segments, similar to per demographic group.

Even this rearward-facing view has many uses for a retailer’s marketing and buying (that’s, inventory management) teams. In actual fact, depending on the corporate’s data literacy level and available resources, this might even be enough (not less than to start). Plus, data scientists can get a feel for the corporate’s customers’ typical spending habits, and this will be invaluable if the corporate does later need to predict future CLV, on a per customer basis.

To allow you to and the corporate resolve whether you wish historic CLV insights or future predictions, let’s view some use-cases for every. In any case, you wish the marketing, management, and data science teams to be aligned from the start on how the project’s outputs are going for use. That’s the very best technique to avoid constructing the mistaken thing, and having to begin again later.

Many tutorials only discuss uses for CLV prediction, on a per-customer basis. They list obvious use-cases, like ‘attempt to re-engage the expected low-spenders to get them shopping more.’ But the probabilities go a lot further than that.

Whether you get you CLV information via calculation or prediction, you possibly can amplify its business value by combining it with other data. All you wish is a CLV value, or some type of CLV level rating (e.g. High, Medium, Low), per customer ID. Then you definitely can join this with other information sources, similar to:

• the products customers are buying
• the sales channels (in-store, online, etc) they’re using
• returns information
• shipping times
• and so forth.

I’ve illustrated this, below. Each box shows an information table and its column names. See how each table incorporates a Customer_ID? That’s what allows all of them to be joined. I’ll explain the columns of the CLV_Info table partially three; First, I promised you use-cases.

Let’s say you’ve ranked all of your customers by total spending up to now, and segmented them one way or the other. For instance, your marketing team asked you to separate the info into the Top 10% of Spenders, the Middle 20%, and the Bottom 70%. Perhaps you’ve even done this multiple times on different subgroups of your customer base, similar to per country, if you might have online shops all over the world. And now, imagine you’ve combined this with other business data, as described above. What can your organization can do with this information?

Truthfully, there are such a lot of questions you possibly can ask of your data, and a lot you possibly can do with the answers, and I could never cover all of it. I don’t have the domain knowledge you do, and that’s a massively vital, massively undervalued thing in data science. But in the subsequent few sections, I’ll provide you some ideas to get you pondering like a data-driven marketer. It’s as much as you to take this further…:

Explore CLV segments and their needs

• What makes a top-tier customer? Are they extremely regular, modest spenders? Or do they shop less often, but spend more per transaction? Knowing this helps your marketing and inventory teams discover what kind of shoppers they really need to accumulate — and retain! Then they will plan marketing and customer support efforts, and even inventory and product promotions, accordingly.
• Why are costs high and/or revenue low to your bottom-tier shoppers? Are they only ever purchasing items at extreme discounts? All the time returning things? Or buying on credit and never paying on time? Apparently there’s a poor product-customer fit — could you improve it by showing them different products? Or here’s one other query: are your bottom-tier customers at all times buying one product after which never shopping with you again? Possibly it’s a ‘poison product’, which needs to be faraway from your inventory.
• Are your high CLV customers more satisfied? Why? Imagine you’re a clothing retailer and your customers have an option to save lots of their sizing information to their account. This permits your online store to make sizing recommendations when a logged-in customer is about so as to add an item to their basket. You furthermore mght notice that almost all of your high CLV customers have saved their sizes, they usually have fewer returns. Hence, you believe you studied that recommendations: Reduce return rates > improve customer satisfaction > and keep shoppers loyal.
• How will you motion this information? Here’s only one idea: the web site team could add prompts reminding users so as to add their size information. Ideally this can increase revenue, decrease costs, and improve customer satisfaction, but in case you’re truly data-driven then you definately’ll need to A/B test the change. This manner you possibly can measure the impact, controlling for outdoor effects, and keeping track of ‘guardrail’ metrics. These are metrics you’d not need to see change during an A/B test, similar to the variety of account deletions.

Explore your demographics

The last section was about CLV tiers; now I’m referring to different customer subgroups, similar to those based on age range, gender, or location. There are two ways you can do that.

1. Perform the above CLV evaluation in your whole customer base, after which see how your subgroups are distributed amongst CLV tiers, like this:

2. Split into subgroups first, and then do a CLV evaluation for every.

Or, you possibly can try each approaches! It relies on the business needs and resources available. But again, there are many interesting questions:

• Which subgroups do you might have? Forget the plain ones I just listed; let’s get creative. For instance, you can split customers by their original acquisition channel, or the channel they now use most: online v.s. instore, app v.s. website. You can split by membership level, in case you offer it. Using tracking cookies out of your webstore, you possibly can even split by preferred shopping device: desktop computer versus tablet versus mobile. Why? Well, perhaps your mobile-phone-based shoppers have lower basket values, because people prefer to make big purchases on a desktop. The more domain knowledge you possibly can construct up, the higher your evaluation and — if it involves it — machine learning efforts will probably be.
• How does buying behaviour differ by customer subgroup? When do they shop? How often? For a way much? Do they respond well to promotions and cross-sells? How long are they loyal? Do they spend often to start with of their lifetime after which tailor off, or is it another pattern? This type of information can allow you to plan marketing activities and even estimate future revenue, and I shouldn’t must inform you how useful that’s…
• What’s a ‘typical’ customer journey? Are you acquiring most of your recent customers in physical stores? Does that mean your stores are great but your website sucks? Or are your in-store staff higher at getting people to enroll in membership than your website is? Either way, you can try to enhance the web site, or not less than, be smarter about which channels you advertise on. And what about recent customer offers, newsletter sign-up discounts, or friend referrals: are they attracting solid numbers of high CLV customers? If not, time to reevaluate those campaigns.

Get clever about your offering, and the way you promote it

• When you understand your customers higher, you possibly can serve them higher. For a retailer, that might include stocking up on the varieties of products their best customers appear to favour. A cell phone provider could improve the services that its high CLV customers are using, like adding features to their mobile app. After all, you’ll need to A/B test any changes, to be sure that you don’t introduce changes that customers hate. And don’t abandon your low CLV customers — as an alternative, try to search out out what’s going mistaken, and the way you possibly can improve it.
• Similarly, in case you understand your customers, you possibly can speak their language. By showing the fitting ads, at the fitting time, on the fitting channels, you possibly can acquire customers you wish, and who need to shop with you.

Know what to spend on customer acquisition

• Ever wondered why corporations start emailing you once you haven’t shopped there for some time? It’s because it’s expensive to accumulate a customer, they usually don’t need to lose you. That’s also why, once you browse one e-commerce site, those products follow you across the web. Those are -called ‘programmatic ads’, they usually appear because the corporate paid for that first click, they usually’re not willing to provide you up, yet.
• As a retailer, you don’t just want throw money at acquiring any old customer. You wish to gain and retain the high value ones: those that’ll stay loyal and generate good revenues over a protracted lifetime. Calculating historic CLV lets you also calculate your break-even points: how long it took each customer to ‘repay’ their acquisition cost. What’s the typical, and which CLV tiers and customer demographic groups pay themselves off fastest? Knowing this can help marketing teams budget their customer acquisition campaigns and improve their new-customer welcome flows (i.e. those emails you get after the primary purchase at a brand new shop), to extend early engagement and thus improve break-even times.

Track performance over time

• Re-evaluate to discover trends. Businesses and markets change, beyond the control of any retailer. By periodically re-calculating your historic CLV, you possibly can repeatedly construct your understanding of your customers and their needs, and whether you’re meeting them. How often must you re-run your evaluation? That relies on your typical sales and customer acquisition velocity: a supermarket might re-evaluate more often than a furniture dealer, for instance. It also relies on how often the business can actually handle getting recent CLV information and using it to make data-driven decisions.
• Re-evaluate to enhance. Periodically re-calculating CLV will allow you to make sure you’re gaining ever-more-valuable customers. And don’t forget to run extra evaluations after introducing an enormous strategy change, to make sure you’re not sending numbers within the mistaken direction.

I do know, I do know… you desire to talk Machine Learning, and what you should utilize CLV predictions for. But this post is long enough because it is, so I’ll reserve it for next time, together with the teachings my team learned on easy methods to model historic CLV and predict future CLV using real-world data. Then partially three, we’ll cover the professionals and cons of the available modelling and prediction methods. When you’d like a reminder of that, then don’t forget to subscribe. See you next time!