Demystifying Ecommerce Lifetime Value: How to calculate and improve your LTV

Lifetime value (LTV, for friends) may seem complicated but doesn’t need to be. 

And, boy, is it a worthwhile metric to track! 

As an ecommerce business, you have two main tasks. 

One is to acquire customers. The other is to get profit (or impact) from those customers. Both tasks are related in more ways than the obvious ones, but we will get to that soon enough.

What makes LTV a tiny bit harder to measure than other KPIs is the fact that you need to define it and adopt a measurement approach that is relevant to the context of your ecommerce business. We’ll also get to that. 

But first, have a look at the contents of this guide and see if you want to jump to any particular section.

What is LTV?

Customer lifetime value is the average profit from customers’ past purchases. This is what is also called a historical LTV. 

However, it can also take into account a prediction of future profit, in which case we’re talking about a predictive LTV.  

Put it simply, LTV tells you what you typically “get out” of each customer, after they become one. 

What does LTV tell us? What’s all the fuzz about? 

Lifetime value is a super-metric. 

Measuring it brings a lot of benefits, and can help you identify worthwhile projects to increase it over time. Here are some of the biggest benefits of measuring and improving your LTV.

• It directly impacts revenue

More lifetime value, more revenue. More revenue, (should be) more profit. By measuring LTV and understanding the factors that make it up, you are in a position to improve it, which will directly improve revenue and profit. 

• It tells you whether you are spending the right amount acquiring customers 

Lifetime value is useful because it tells you how much you’re making per customer, which in turn helps you assess if you’re spending the right amount to acquire each customer in the first place. In other words, if your CAC (cost to acquire a customer) and LTV are in a good balance. 

If you’re making £20 per customer’s lifetime and spending £21 to acquire them, you are acquiring customers at a loss. This means you will run out of money and don’t have a viable business.  Later, we’ll discuss what makes a good LTV and how it relates to CAC. 

• It flags up the most profitable customers 

You can calculate LTV for your whole customer base, but you can also conduct further analysis on the impact of different customer attributes on LTV. 

For example, a sports company might find that people buying water sports products are amongst the most profitable and therefore decide to invest more in product development for this segment. Or a hotel booking company might find out that solo business travellers are the most profitable and therefore develop better product features for them. This can only happen if you can measure LTV by customer and customer segment.

• It sharpens your acquisition efforts. 

Similarly, it can help orient your acquisition efforts and the targeting for your performance media. 

For example, you might find that geographical regions perform differently over time. If you can find a sweet spot where CACs are low and LTVs are high, this could be a fantastic audience segment for further penetration. 

Consider a pet accessories brand. They might find that London, where pet ownership is the lowest in the country, offers the lowest LTV. Based on this insight, they might decide to limit spending in London to double down on regions that offer higher lifetime values. 

LTV in ecommerce  

You will think differently about LTV in an ecommerce business than in a SaaS or other form of subscription-based business. 

Ecommerce is typically non-contractual. Customers come to your site and purchase under no obligation to purchase again. This means that it is hard to predict when each customer will buy next, (if they do at all). This makes predictive LTV hard to calculate. 

Having said that, more ecommerce businesses are now offering subscription services either standalone or as a type of available purchase. This essentially means that there is a form of contract in place. 

And in most cases, these companies will offer both “one-time-purchase” and subscription, which means LTV needs to be calculated separately for each, although it can also be calculated in aggregate for both as well. 

We’ll clarify this a bit more, but before that let’s look at the two main types of LTV calculations we mentioned at the beginning: historical and predictive. 

Types of LTV, by the method of calculation: historical vs predictive 

Put it simply, there are two types of lifetime value calculation.

• Historical: the easy and relatively useful one in most cases. 

This approach is based on past behaviour. It looks at what your customers have spent in the past as the only guide, and without inferences. 

Example: If you’ve made £1,000,000 revenue from 10,000 customers, the lifetime value will be £100, multiplied by the gross margin ¹

The main problem with this approach is that it’s not taking into account the length of time of customers. It simply uses all-time data, which over time gets higher. This introduces errors to the data ²

Therefore, for a historical approach to be useful at all, you need to introduce intervals, this is a period of time that the calculation is based on (from the first purchase to today). When you do it this way, you will get data such as 

• 1-year LTV is £5 (the average customer drives £5 of gross profit in one year)
• 2-year LTV is £7.5 (the average customer drives £7.5 of gross profit in two years)
• 3-year LT is £8.2 (the average customer drives £8.2 of gross profit in two years)

This LTV calculation is a lot more useful because it tells you the length of time it takes for a customer to reach a threshold of spend, and it also shows you the rate at which the spend decreases proportionally. 

Historical calculations with set intervals are useful for established businesses that have been trading for some years, especially if the frequency of purchase is relatively high. 

If you haven’t been calculating LTV at all, this is a great place to start. Here’s the formula:

1-LTV = (Total Revenue Last 12 Months / Number of Customers) * Gross Margin (%)

Let’s walk through the formula. You add all the revenue in the last 12 months for all your customers, then divide it by the number of customers to calculate the average revenue per customer in the last 12 months. Then you multiply this by your average / historical gross margin.

• Predictive: the harder, fancier method

Predictive LTV measurement needs to make use of assumptions and inferences to predict the future profit coming from each customer.

Depending on the amount of historical data you have (in particular the number of transactions you have per customer), you can use a method of calculation or another one. 

The simplest method (and the best one for businesses with little historical data and transactions per year) is a heuristic method that divides the average profit from customers by their churn rate, in a given period.

Predicted LTV (set period) = Average profit from customers (set period) * Churn Rate (%)

Churn rate is theoretically impossible to know in traditional ecommerce, one-time-purchase non-contractual models. Just because Joseph hasn’t bought any shoes in one year it doesn’t mean he will never do it again. But it’s a safe assumption we will have to make. And in many ways, it’s not as important how you define Churn Rate, as it is that you remain consistent about it. 

Let’s think of a fashion business, one with over two years of data. You could take a snapshot of your customer database with purchase data for these last two years and calculate:

• What is the average profit per customer for the last two years. Let’s say this is £20. 
• What is the rate as to which they churn, looking at how many customers didn’t buy anything in the last two years, vs those who did. Let’s say this is 30% 

Predicted lifetime value in 2 years = £20 * 30% = £6

This means you predict to make £6 per (existing) customer over the next two years, and you can add this to your historical data for a slightly conservative calculation of LTV (based on 4 years, two historical and 2 predicted). 

You can make this calculation provided you have access to your customer database and with basic Excel skills. 

Let’s go with the harder method now. 

This method tries to model the churn rate per customer (as opposed to in aggregate or by segment) through the use of a statistical model. 

The thing is, the likelihood of someone churning depends on their uniquely past behaviour. If someone typically buys every month and they haven’t done it in 3 months, they are likely to have churned. But if someone buys every 6 months, we wouldn’t assume they have churned after 3 months. Therefore, it’s more precise to model the churn rate per customer. 

To model churn rate depending on historical data, you need to know three things. 

• Customer tenancy (how long they’ve been a customer of yours) 
• Number of purchases 
• Recency of purchase. 

With these three metrics, the model can answer the following question: 

Is the time since the last purchase (recency) greater than the typical frequency between purchases? (the number of purchases divided by customer tenancy). 

The model could then say that Mary has bought on average every two months but hasn’t in the last 6 (so it assumes churn) whereas Josh only buys every 7 months so the same 6 months of recency do not indicate churn. The model then assumes Josh will continue this frequency and drive LTV, whereas Mary is now “dead” to us. 

This calculation is a lot more precise. 

To use it, you will need to use a Python or R package, such as Lifetimes  

LTV in Subscription vs One-time purchase business models. 

Subscription as a business model is growing dramatically. Companies love it because it commands retention naturally, and customers love it because it’s convenient. 

Besides funerals, there are subscription services for just about everything. 

Subscriptions come in different shapes and sizes but they all have in common that the customer purchases by doing nothing, as opposed to one-time-purchase when the customer will purchase by doing something.

This means that under a subscription model, the customer has to tell the company that they wish to churn. At least from the subscription. At least at this time. 

Therefore, subscription businesses can easily calculate churn and make it a central metric of attention. When do people churn? Why do people churn? What kind of people churn? How does product usage impact churn? What about customer service interaction? 

With one-time purchases, which is the majority of ecommerce purchases, churn has to be inferred (as discussed in the previous section) 

The business model also influences the type of calculation

OK, I think it’s time for a table 

	Historical / Intervals	Predictive
One time purchases	Easy to calculate  Uses metrics you’re already tracking such as:  * Revenue per user * Age of customer * Profit margin	Hardest to calculate  Churn has to be inferred through frequency, recency and length of tenancy analysis. Or calculated in aggregate, which is a bit blunt.  NB – Lack of regularity in purchases (eg novelty purchases) will increase error in the model
Subscriptions	Easy to calculate but doesn’t offer as much value as predictive LTV.	Easy to calculate taking into account average churn rate to infer how long we expect customers to stay.

What is a good LTV? 

Let’s say that you have followed one of the methods above and arrived at £38 as your lifetime value. 

Is that good? Is that bad? 

No metric can be good or bad without a point of comparison. 

You have three possible ones. 

Firstly, comparison to the past. If you’re improving your LTV over time, that’s great! 

At some point, gains in LTV might come harder to come by. This might be a sign that you’ve exhausted this line of enquiry. Here, you might be better off decreasing CACs (if your objective is short-term profit) or investing in brand awareness (if your objective is long-term growth; please note this will also decrease your CACs, but only eventually).

By having interval-based LTV calculations (1-year LTV, 2-year LTV, etc) and tracking those monthly, you will know how these are trending, and therefore the impact of your initiatives. 

Secondly, market comparisons. This can be a bit harder and more expensive. In essence, you need to know what’s the typical lifetime value in your category. To do this, you can do a survey of your target market and include survey questions aimed at understanding typical spending patterns from customers. If you find that an average homeowner spends £200 on gardening products annually, but your average customer spends £20, you have some way to go, and some questions to ask yourself.

Thirdly, you have CAC:LTV ratio, which is the most meaningful way to interpret LTV. 

The cost of acquiring a customer (CAC) needs to be lower than the gross profit you’ll make on that customer (LTV). Not just lower but at least half the size (ie CAC:LTV ratio of 1:2) because besides variable costs such as the cost of the product, shipping and payment gateway, you need to pay for the semi-variable and fixed costs of running the business such as staff, offices and technical infrastructure. A conversation at this point with your mighty finance team could be helpful to set ideal CAC:LTV ratios. You’re probably safer with a lifetime value that is three times your CAC, but it’s worth aligning with everyone about it.

What factors contribute to LTV? 

Let’s break down the factors that make up your LTV in a super practical way. 

These factors are your levers to influence LTV. 

You can pray for a higher LTV and you can wear special socks to conjure a growing LTV, but if you want to take action, you need to go down one level into its individual components.

Let’s consider these three factors: 

• How much do customers spend per transaction, or “average order value”. The more your customers spend per transaction, the higher your LTV (of course, increasing AOV is great in and of itself anyway)

• The frequency with which they purchase. A customer purchasing 5 times a year will most usually have a higher LTV than a customer that only purchases once. 

• How satisfied they are with their experience. A happy customer will be more likely to transact again. To translate this fuzzy idea into an obedient number that you can track, compare and graph, you can implement NPS in your business 

What is great about this way of thinking is that you now have three potential projects or workstreams that you can prioritise and tackle individually, as they will react to different types of implementation. 

Let’s quickly cover this in the next section 

How do I improve my LTV? 

To improve your LTV, you need to increase the amount of revenue per purchase, the frequency rate of purchase, and/or the overall customer satisfaction (as a proxy for retention).

Here are some ideas on how to do this

Improve Average Order Value (AOV)  by providing a richer experience to users 

Improving AOV can be achieved by engaging customers as they search, browse and add items to their baskets. 

It can be useful to think about a traditional retail experience. When customers are in a store, they can easily browse through a lot of the items. Signs and cues help customers navigate the store, and as they engage with a shop assistant, recommendations for alternative or additional products may happen. Because of this rich, interactive experience, customers will commonly leave the shop with more, not fewer items than they initially considered. 

Ecommerce businesses have the opportunity to replicate this experience of discovery, rather than falling back into a vending machine approach. In doing so, they can see their AOV increase, whilst giving users a more memorable experience. 

To do this, you can consider tactics such as product bundling, where items are bundled together for customers’ convenience. 

The best way to bundle a product is by use-case, as in this example from Tommee Tippee where baby fundamentals are grouped in a “Baby Get Ready”. This is perfect for new parents who usually don’t have the time or the knowledge to pick all the products they need. 

Other tactics include cross-selling related products or adding content blocks promoting curated collections on popular touchpoints such as the homepage or category pages. 

Improve frequency of purchase through personalised shopping moments. 

Customers shop with you when they are in the market for your product. There is a trigger moment (in fancy terms “category entry points”) that gets the customers to become active, consider the brands they know, and do some research to narrow down options and finally, purchase.  

But you don’t have to wait for those category entry points. You can make this happen yourself. 

Email, and to a lesser extent social media are great channels for this. This is because your audience is (mostly) existing customers who already have experience of shopping with you (and conveniently they might have accounts, payment details on record or even loyalty points accrued). 

On email, you can make use of segmentation so that you can address customers based on past purchases or audience attributes you collect such as gender. This way, you can personalise these shopping moments for greater relevancy. 

Keep it simple. These customers already know you and like your products, so nice photography and a bit of thoughtful copy might be all you need to persuade this customer to consider a purchase. 

In this example by Rapanui, a fun picture of a dog alongside some entertaining copy is the perfect call to action to a product description page. 

Improve length of tenancy (customer’s age) through NPS analysis

Net Promoter Score is a great metric for ecommerce businesses because it tells you a lot, and it’s terribly simple to implement. 

NPS is a score ranging from -100 to 100. To calculate this score, you first need to get your customers to answer a very simple question. 

(from 1 to 10) Would you recommend [our company] to a friend or colleague? 

Depending on their answers, customers get classified as “promoters” (who scored 10 or 9), “passives” (who scored 7 or 8) and “detractors” (who scores 6 or lower). Once you have the answers, you subtract the percentage of detractors from the percentage of promoters. That’s your NPS score. 

It’s however very important to interpret this score in the context of your category. To do this, you can check Delighted’s NPS benchmarks.  

Analysis of NPS studies shows that this humble score carries a strong correlation with business growth because it achieves an unbiased understanding of customer satisfaction and the likelihood of continued purchases. 

NPS surveys tend to also have a follow-up, open question that typically reads “Could you tell us why did you give that score”.

We’ve found most people that answer these surveys tend to give quite thoughtful answers to this second question (especially detractors!), and it’s here where you can find leverage to improve your customer’s satisfaction, and with it, the expected length of tenancy. 

By reading and classifying answers by topic, you can put together a list of the major pain points of customers, prioritised by the number of mentions. 

As a bonus, you also get to understand why people love your company, by reading the comments of promoters. You can then use this to refine your messaging themes in customer acquisition or brand development. 

Footnotes

1. Gross margins are not always easy to find out, specially at SKU level. Knowing an approximate gross margin based on historical data is a great start. You have the choice of accounting for the cost to serve a customer, return costs, shipping costs, payment gateway fees for a more precise LTV calculation, or you can simply use the product mark-up as your profit margin, which is a lot simpler. If you do the later, you need to allow more padding between your CAC and LTV.
   
2. As time goes by, your lifetime value calculation is based on a longer length of time, which will make the values fluctuate for reasons outside of your control. In some ways it will grow because your customers are spending more over time (they can’t spend less over time), but in other ways it will decrease because you might be bringing new customers that spend less than your previous cohorts of customers. This is very common, because businesses tend to grow disproportionally in light buyers as they gain market share (ie, you already acquired your biggest fans and heavy buyers in the category earlier on so over time you only acquire lighter buyers)