Project 1 – Lifetime value (LTV) introduction

Welcome! As the first official post, I want to take a moment to set the stage for the upcoming collection of work I will be posting. As this site is all about analytics, with a focus on business – I want to discuss some things that are very near and dear to my heart: analytics, the customer relationship and sustainable profitability. These series of articles should be generalized enough for the educated business owner to derive meaning while insightful enough to aid the seasoned analyst in efficiently contributing. But please note: this is geared for someone with a business, web, online marketing, engagement and analytics background. If you’re looking for some resources to learn statistics, Google Analytics, Microsoft Excel, or just re-acquaint yourself – here are some I think are good (and concise).

The underlying fundamental in any business  the understanding of the Lifetime value (LTV) of each user. In its clearest definition: this is the gross revenues we expect to see from each visitor/user/etc over a given lifetime (term). This number is crucial to forecasting trends (and expected revenues), quantifying the impact of bad behavior (such as sending out too many emails or ad abuses) and showcasing opportunities. I’ve personally seen many companies spend a lot of money on acquisition tactics like Search Engine Marketing (SEM) without understanding what the return (ROI) is. I’ve also worked with companies who were hammering customers with emails (one was sending out 9 a day) and experiencing massive unsubscribes. They had no idea what the value of these subscribers was or that the unsubscribes caused collateral damage to other business units in the company. I can’t overstate the need to spend time developing a good, multi-cohort LTV calculation. 

Part 1. Choosing the business/platform you want to review

The first step (obviously) is in determining what platform or business you want to review. Be careful on scope here as you may quickly end up with 10 different customer data sets, leaving you mired in “analysis paralysis.” Work with the business owners to understand what measurements are meaningful and listen, listen, listen. The second step is to make sure you can get the data to continue work. If you don’t have the data you should investigate ways on instituting these measurements.

Part 2. – Defining the Cohorts (segments of customers)

\sum_{i}^{n} \frac{n!}{i!(n-i)!}
Calculating the total combinations of a n-set Venn diagram

746px-Symmetrical_5-set_Venn_diagram.svg

In order to effectively calculate the LTV values, it makes sense to quantify each customer segment in as granular of terms as possible (e.g. casual web user, heavy email reader, uses mobile devices and buys a lot of t-shirts from your company store). For this exercise, I love the VENN diagram and you can see how many combinations (31) there are with just 5 base elements (A,B,C,D,E). Now, a word of caution – please try and keep the elements to 6 or less – otherwise the work gets complicated and you will have issues succinctly communicating with the business owners “analysis paralysis.” In my experience – 4 or 5 is ideal. Anything less looks half-hearted and anything more becomes confusing to the reader. Have you seen a 6 element Venn diagram? Its cool visually but worthless from a storytelling perspective.

There are a myriad of external factors we cannot account for and we wont event bother to address them further (e.g. unemployment, radioactive monster attacks, zombie apocalypse, et al.). Our analysis should be drafted with quantifiable data, presented for review, given the lens of qualitative insights from the business owner and revised accordingly.

The next step will be in taking your universe of customers and in identifying the how each customer falls into each group. At this point we will not identify a value ($$) yet, just understanding the respective populations.

Part 3. – Quantifying GROSS revenues per each product 

amountsPart 3 will be about determining a value for the primary segments (e.g. A, B, C, D and E). This process will usually be very complex – after all, there is usually a reason why nobody has awareness of the LTV value. But, in contrast to the segment definitions – its important that we grab every detail that is DIRECTLY related to that segment. Here we will grab all advertising and subscription revenues.  I will walk through how to determine the values, which can be complicated – especially for Display Advertising. Of note, certain types of business have heavy seasonality (notably programmatic advertising) and businesses often need to forecast for the slumps (in eRPM and traffic) during summer months. We will discuss both an averaged model and a more detailed monthly model.

Part 4. – Gauging duration of lifetime

churnout2Part 4 will be in ascertaining the term (length of lifetime) for each contributing financial detail. For example, We can determine a monthly subscription subscriber stays with the product for 13 months (on mean/average before leaving). For subscription services this modelling can get very complex since there are inevitably active customers we want to incorporate. Its important to note that this model should be computed by month (if possible) so you can accurately watch trends and seasonality. This insight will be invaluable in associating risk sensitivities.

I will walk through lifetime calculation techniques for generalized (anonymous) Web traffic, authenticated web traffic, subscription services and email marketing lists.


Part 5. Arriving at LTV amounts

\alpha=\sum_{i}^{n} \frac{n!}{i!(n-i)!}   Here we will take the previous 4 steps and fill in our VENN Diagram to arrive at the values for each segment as well as a overall number for the combined businesses.

\sum_{i}^{\alpha} LTVi\frac{\mu i}{\mu \alpha}
Arriving at the average, total Lifetime value

Part 6. Risk.. Reviewing the historical trends and associating events

Its important to view the future through the lens of the past for we are often doomed to repeat ourselves – foolishly expecting different outcomes. A good business owner listens to their customers and can quickly tell you what positively and negatively impacted performance during stand-out events. And usually not just “we had a sale” – but “we had a sale on x product, which was very popular and we marketed it <here>” or the opposite – “Our CEO made a very polarizing and public remark which offended a lot of our customers.” I want to succinctly state here, that emotional behavior on the customer is not always directly corollary to the period of the disturbance. Often, customers will opt to not renew their service the following year. If you didn’t have the business owners input, or better yet – the customers feedback, you may incorrectly attribute the churn to something at a later period.

Part 7. Creating forecasting trends

Trending out is useful for budgeting, ‘ceteris paribus’ – but as we know, business is always evolving and there is always a competitor or fire drill at the next corner. Here, I will show you how to use Excel to forecast out trends for web traffic. I will also upload a calculator which will help you try to compensate for changes in attribution, device preference and engagement.

I hope you enjoy the work and find it useful. If you have any questions or want to see anything elaborated on, please drop me a line or comment to the post.

 

 

 

 

 

Article by David

Your host and an experienced analyst dedicated to evangelizing the quantifiable, data-driven measurement of business with an emphasis on the customer relationship.

Leave a Reply

Your email address will not be published. Required fields are marked *