/ CX

# Understand and Increase Your Average Order Value (AOV) and Customer Lifetime Value (LTV)

## Introduction

No doubt, there is some confusion around the topics average order value (AOV) and customer lifetime value (LTV/CLTV/CLV). In this post I would like to shed some light on it and clear the mist.

You will also learn how to look up the average order value (AOV) in Google Analytics and why you might want to change the way you look at these numbers to make them become more relevant for your business.

Second, I would like to show how to improve these metrics using the models we discussed before.

Please note: This post requires some basic understanding of Google Analytics. However, I have linked further information so you can easily look up these topics.

Let us get started!

## The Average Order Value (AOV)

No doubt: For most marketers conversion rates are by far the most important KPIs. For ecommerce, average order value (AOV) is critical. Let’s dive into it

### What is the Average Order Value?

Let us look at the underlying formula:

Average Order Value = Revenue / Number of transactions

The AOV measures the average total of all orders over a certain period of time. Like any other metric, the AOV only may become a KPI if you find it is meaningful for your business. Based on this number you may evaluate your marketing (ad) budgets and your business health.

### Common Mistakes When Interpreting the AOV

However, there are some common mistakes when interpreting the AOV. Let us take a closer look...

#### AOV Reflects Sales Per Order, Not Sales Per User

A customer may purchase multiple times at your shop, but for an AOV calculation each order is counted separately.

Let me give you an example. Let us assume we were recording these orders for an online shop over a given period of time:

• User 1: \$10 purchase
• User 2: \$100 purchase
• User 1: \$100 purchase
• User 2: \$10 purchase
• User 3: \$10 purchase

The default AOV in this case would be:

\$230 (sum of all order amounts = revenue) / 5 (number of orders) = \$46.00

The average order value per user however would be:

\$230 (sum of all order amounts = revenue) / 3 (number of users) = \$76.67

\$46.00 vs \$76.67 is quiet a difference. It really depends on what you would want to look at (we will get back to this a later when discussing the lifetime value metric below).

#### AOV Is NOT The Revenue

Don’t take your AOV as the base for your revenue (and gross profit or gross margin) calculation.

Why? Because AOV doesn’t tell you anything about the specifics or nature of your order values (including margins and customer returns).

You may say: “What can I use this metric then for?”. Let us look at another example: An online retailer having

• 3 t-shirts in stock
• priced at \$10, \$20, and \$30
• with an Average Order Value (AOV) of \$12

wants to understand his business. The AOV metric leads to the following conclusions:

• majority of users are likely not purchasing multiple items per order
• low priced items represent the majority of sales

If the low priced t-shirts have the best margin: fine! But what if the 30\$ t-shirt offers the highest margin instead? In this case the retailer might would want to increase the AOV to increase ROI / ROAS.

But it really depends on your current situation. Boosting the AOV doesn’t mean necessarily to increase ROI.

#### Median vs Arithmetic Mean For AOV Calculation

Please also note the average order value (AOV) does not use the median but the arithmetic mean as a calculation method.

Hence, if your order totals vary a lot due to a broad product range and diverse customer segments, you might want to take this into account to not be mislead when calculating your ad spends and marketing efforts.

Let me give you an example. Let us say we see a row 5 orders over a given period of time:

• Order 1 Value: \$1000
• Order 2 Value: \$10
• Order 3 Value: \$10
• Order 4 Value: \$100
• Order 5 Value: \$5
• Order 6 Value: \$10
• Order 7 Value: \$5

The AOV (using the arithmetic mean) in this case would be:

\$1140 / 7 = \$162.86

whereas the AOV using the median would be:

{5, 5, 10, 10, 10, 100, 1000} = \$10.00,

taking the middle (=median) value from a sorted row of all order values. As you can see, a single high value order can easily distort the AOV result when using the arithmetic mean.

The median value is harder to calculate but may give you better and more reliable insights and understandings in some cases.

For operational calculations (e.g. Google Analytics), usually the arithmetic mean is used, but you may easily calculate the median AOV using Excel or any other calculation tool.

### Average Order Value (AOV) in Google Analytics

To see your average order value (AOV) in Google Analytics, you first need to enable Enhanced Ecommerce Data.

Afterwards you may retrieve your website’s AOV by navigating to

1. Conversions
2. Ecommerce
3. Overview

You may easily create a segment in Google Analytics to recalculate the AOV outlined above based on the sessions segment you would like to look at (I am using US users only in the screenshot above).

You may also check out the average order value (AOV) sorted by traffic source (such as organic, paid or social traffic). Simply navigate to:

1. Acquisition
2. All Traffic
3. Source/Medium

and select “Ecommerce” at the top menu:

It depends.

Let us combine users and orders from the above example. Over a given period of time we were recording these orders for an online shop:

• User 1, first order: \$10 purchase
• User 2, first order: \$100 purchase
• User 1, second order: \$100 purchase
• User 2, second order: \$10 purchase
• User 3, first order: \$10 purchase

Let us assume the shop sells two products:

• Product A costs \$10
• Product B costs \$100

The AOV in this case is

\$230 (sum of all order amounts, revenue) / 5 (number of orders) = \$46.00

while the median AOV is

{10, 10, 10, 100, 100} = \$10.00

Knowing the pricing structure of the shop, we can already conclude by those KPI’s that

• more people take low value orders than high value orders (hence the median AOV of \$10.00)
• but still high vs low value orders are somewhat balanced (hence the default AOV of \$46.00)

#### Calculating Ad Budget Via Average Order Value

Let us further assume the shop owner would like to spend 10% of the revenue for advertising:

• a purchase is worth \$46 on average
• \$46 / 10 = \$4.60 ad budget per purchase
• (\$46.00 * 5) / 10 = \$23.00 total ad budget for time period given

#### Calculating Ad Budget Via Customer Lifecycle

While this is fine, we also have made another observation: This particular shop has returning customers (user 1 & 2).

Assuming that user 3 soon also may become a returning customer spending additional \$100, shouldn’t the shop owner take this into account?

What if the shop owner calculates the ad spend as follows:

• a user buys for \$110 on average
• \$110 / 10 = \$11.00 ad budget per user
• \$11 * 3 = \$33.00 total ad budget for time period given

?

In this case the shop owner spends more money on ads while expecting to generate more revenue due to returning customers purchasing more than one time.

The calculation above takes the whole user (or customer) lifecycle into account (a row of transactional and other events generated by a user over time) while an AOV centered approach focuses on the purchase event itself.

### What is the Lifetime Value?

The investor David Skok thinks of the customer lifetime value (LTV) as something that goes beyond the pure cost of acquiring a customer.

Skok believes that over-investing in the initial transaction momentum is a common mistake made by many online businesses leading to a potential failure because the long lasting relationship between a company and a customer is overseen.

Skok highlights that if the costs of acquiring a customer exceed the LTV, than this business model is likely to fail:

Source

...whereas the opposite proportion signifies a healthy business setup:

Source

### How to Calculate the LTV?

Calculating the LTV can be tricky but helps you to leverage your business in a sustainable way. If you look up a formula for LTV, you may stumble upon such impressive creatures:

Source

However, you may not need such a complex formula to get to some results. Let us first calculate the LTV based on the setup we discussed before:

• User 1, first order: \$10 purchase
• User 2, first order: \$100 purchase
• User 1, second order: \$100 purchase
• User 2, second order: \$10 purchase
• User 3, first order: \$10 purchase

#### Calculating the AOV

The first thing you need to calculate is our good ol’ AOV. In this case we already did this before:

\$230 (sum of all order amounts, revenue) / 5 (number of orders) = \$46.00

#### Calculating the Purchase Frequency (PF)

Next, we need to determine the purchase frequency (PF): How many times on average does a customer buy over a given time period?

The formula is simple:

PF = Number of orders / Number of users

...in our case:

5 / 3 = 1.67

#### Calculating the Average Customer Value (ACV)

Then we need to find out the average customer value:

AOV * PF = ACV

...in our case:

\$46.00 * 1.67 = \$76.67

Does this number looks familiar? Right so! We discussed it already right at the beginning of this blog post when highlighting that AOV reflects sales per order, not sales per user.

Simply double check this value by multiplying the ACV with the amount of users to see if matches the overall revenue: \$76.67 * 3 = \$230. Our calculation is correct!

#### Calculating the LTV

As a last step, you would need to take the user lifetime span and multiply it with the ACV.

“(...) average lifetime of a customer is how many years (weeks, or months) your average customer will stay and purchase with you before going dormant and stop. The one true and effective way to understand and see this is by looking at your historical data. To do this, you can view the average time between customer purchases.

Once this time period has been established, and then a customer goes more than two standard deviations past that time period, it can then be safe to assume they are no longer a customer. Therefore, the average time a customer goes before reaching that point, is your store’s average lifespan (t).“

Let us say, the user lifetime span in our case is 3 (weeks). The LTV for this user then would be:

\$76.67 * 3 = \$230.00

...matching exactly our overall revenue for the given time period (1 week in this case).

## Which Way To Go: AOV or LTV?

AOV alone is more often used for one-time purchases (products have to be profitable on the first sale). Your marketing spends simply need to pay off right from the start, as there are no future transactions to be expected. In other words, AOV and LTV are the same value.

An LTV always comes into play if users or customers are likely to create more than one transaction on a website.

Let us look at some examples:

AOV LTV
A shop sells vaults and users will likely create a one-time purchase and never return (they might refer other customers though) A shop sells lipsticks and users will continue making purchases (as long as they like the product and the shop)

Source

Depending on your business, you might prefer one method over the other when calculating your marketing spends and overall performance.

## How to Improve Your AOV or LTV?

After we discussed the AOV and LTV KPIs in detail, it is time to look at marketing measures and technologies that help you to improve this metric and leverage your business.

Whether you are new to AOV / LTV calculation or a pro already, I bet you (at least sometimes) struggle with finding the right tools to do the job...this isn’t 2000 where only a few and very basic technologies were available to drive conversions.

In 2018 it is complicated:

Source

There are simply too many tools on the market to gain an overview and marketers can get very busy staying tuned.

### Marketing Strategies To Increase AOV and LTV

Many marketers get overwhelmed when asked to pick the right tools for their business. So let us step back for a second and look at common strategies used for increasing AOV and LTV:

• Cross-selling (bundle related products)
• Up-selling (market higher-end products)
• Implement loyalty programs

A great way of increasing AOV and LTV via cross- and up-selling is to use recommendation engines that recommend suitable products to new and existing customers tempting them to add more products to their carts.

### Recommendation Engines: Your Best Friend When Boosting AOV & LTV?

Recommendation engines (also known as recommender systems) are on the market for quite some time. Originating from the retail industry delivering personalized product recommendations, the techniques and methods behind this powerful marketing tool soon became utilized in many other environments.

Amazon’s “You might also like (...)” or “Other customers bought as well (...)” suggestions are based on recommendation algorithms acting as a salesmen who automatically identifies the customer taste and then drives conversions via tailored product recommendations.

Recommendation engines showed up in ecommerce first, but nowadays you may also find them in other sectors such as media (YouTube or Amazon Prime, “Recommended Videos”), social networks (Facebook, “People you might know”) or even search engines (Google, “Visually similar images”).

Over time, recommendation systems and engines became more sophisticated. Today systems are available for almost any working environment:

Source

#### Benefits of Recommendation Systems

There are some strong advantages recommendation engines have over other marketing strategies and their related tools:

• Simplicity: Just install the engine, the rest is done almost automatically (data collection and recommendations).
• Reduced Workload: In most cases, installation via a plugin or a script is equally easy to perform.
• Lift: You may increase AOV and LTV with ease, depending on your visitor base and your business.
• Versatility: Many engines and suites allow you to choose the right algorithm for your website.
• Customer Compliance: If not displayed intrusively, product recommendations may increase user compliance due to new products they would have not discovered otherwise.
• Tailored User Base: If your user base doesn’t change much, the algorithm becomes more and more accurate over time being specifically connected to your business

#### Drawbacks of Recommendation Systems

While I generally recommend to use recommendation engines [sic] for any kind of business, there are some drawbacks:

• Sample Size Noise: Adding more items to your catalogue may increase the noise and reduce recommendation quality, especially if items are rarely picked.
• Missing Trust: Users may not trust recommendations that are created automatically.
• Cold Start Problem: You need enough observations (users pick items) to display valuable results. Recommendation quality improves over time.
• Narrow View: The database used by recommendation engines often is isolated (e.g. restricted to the website) not taking user actions into account from other sources.
• Lacking Persistence: In some situations, it may be better to re-show recommendations or let users re-rate items than showing new items.
• Lack of Control: Lack of control because recommendation systems usually run automatically on your website.
• Missing Insights: You may not get enough valuable insights in how your visitor base behaves.
• Missing Segmentation: Rolling out recommendations for your whole audience regardless of its user segments leads to broad results.
• Slow Adaptation: If your user base changes or new traffic sources open up, the system needs time to reiterate and test.

### Ground Control To Major Tom, Or: Going Beyond Recommendations?

Source

Recommendation systems help you to get a broad direction on your market. Look at it as your first step into machine learning, data driven marketing. However, you are still targeting your whole website audience instead of individual user flows.

No doubt: Increasing AOV and LTV can become a tough challenge. Many marketers therefore focus on the whole traffic and simple user actions (such as browsing a product) and tend to forget about the big picture.

### Optimize For Individual Users, Not For an Audience

What matters most is giving the right response to a specific user at the right time.

### Segment Your Users Via a Personalization System

The central idea of a personalization system such as Intempt is to

• Segment your identified users based on tracked website behaviour and third party data
• Change your website individually in real time
• Grow by increasing KPIs that matter to you (e.g. AOV & LTV)

This system is completely customizable, so you may recreate simple behavioral campaigns including recommendations or abandoned cart notifications, but you may also go far beyond by personalizing your user experience across channels using all user data available.

### Cover Any User’s Lifecycle

With a user personalization engine such as Intempt you may cover all phases of individual user lifecycles:

#### Pre-Purchase - Attract (Day 0)

• Which user segment should I target with personalized offers so they make their first purchase?
• How can I re-merchandize products based on browse behavior to make the shopping experience relevant on the first visit (use case below)?

#### In the Conversion Path - Enage (Day 1 - n)

• Where do people get stuck on my website so I can move them along?

#### Retention - Grow (Day n+1)

• Which of my prospects are on the fence? What can I do with those?
• Does my website automatically personalize to users who I sent emails to and have CRM data on (use case below)?

### Advantages of a Personalization System Over a Recommendation System

• Small Segment Size: Because user personalization takes all user data into account (including email or CRM), the segment size used in a campaign can be small
• User Compliance: You might or might not run personalization campaigns that are visible to your users.
• Broad View: Data from all angles is used to create rich user profiles.
• Full Control: You may adapt you campaign events, segments and setups at any time.
• Missing Insights: You may get valuable insights in how your visitor base behaves even without running a campaign. All collected data remains persistent.
• Rich Segmentation: Precisely targeting user segments leads to better results with precise impacts.
• Fast Adaptation: User personalization takes all user actions into account, not just browsing specific product pages. Because the database is richer, the adaption is fast.

### Pre-Purchase Personalization Use Case: The Drop

On Shopify, users get to see category pages displaying products in a default sort order (based on the Shopify settings), e.g. ‘Best selling’ or ‘Newest’.

But visitors are unique: Some (even subconsciously) pick rather expensive products, others browse only cheap items. Then there are users who browse both kinds of products, so no behaviour pattern is visible from a price point of view.

What if you could automatically alter the sort order based on the user’s previous browsing behaviour so it is more likely users convert?

You may say: ‘Users can change the sort order at any time’.

While this is true, wouldn’t it be better to already deliver the right sort order to the right user to increase the chance of boosting the AOV of a first-time user?

Especially when considering the fact that many users don’t use the setting options a shop system offers to them?

#### What is The Drop?

To dive into on-site behavioral merchandising, let us take a look at this Shopify shop from one of our clients:

It is a typical apparel shop with a large catalog containing hundreds of products. In other words: An ideal candidate for our campaign.

#### Change The Default Category Page Sort Order

As you might know, Shopify allows to display collection and category pages with a preset of different sort orders:

• Manually
• Best selling
• Highest price
• Lowest price
• (...)

Let us define the user’s behavior that matches our campaign segment first:

If

• a first-time user
• browses three high priced products in a row

I would consider this visitor to be interested in high priced items.

Then

• I would like to display category pages
• with a sort order different from the default sort order: Highest Price on top
• instead of New on top
• To help the user getting oriented

A default Shopify category link looks like this:

my_shop_url.com/collections/my_collection

whereas a ‘treated’ link with a sort order highest to lowest price would look like this:

my_shop_url.com/collections/my_collection?_=pf&sort=price-descending

Returning users and users from other traffic sources such as social, paid or other referral websites I explicitly would want to exclude from this campaign because I would not want to disturb their user lifecycle or run other campaigns on these segments.

#### Create an On-Site Behavioral Merchandising Campaign in Intempt

With Intempt it is easy to deliver personalized category or collection pages based on the actual behaviour of a user. It doesn’t take more than 10 minutes to create such a campaign.

You don’t need to collect a lot of user data before running the campaign nor trust in any machine learning logic. You are still in full control of the strategy while the campaign runs automatically.

To set up such a campaign, you would first need to install the Intempt tracker - a code snippet similar to Google Analytics - collecting each user’s behaviour.

Afterwards you may create an event to identify high priced items. Simply tell Intempt what the price and a product page is by adding some additional code to your website.

Now you can easily create a user segment. The segment in our case could look like this:

• A first time visitor coming from Google browsing
• A t-shirt
• Above \$100
• More than two times
• Within last 30 seconds (recency)

Note that the first user property is fixed. The Intempt tracker stores all information available from the HTTP header, such as the referral website host, in this case Google.

The other four properties are behavioral properties segmenting the user (browsing) behavior.

Information about the product such as the product category or the vendor can be retrieved

• directly via the page URL (if it contains this information) or
• by using a script submitting this data to Intempt

whenever a user browses a category or a product page.

Data from the referer (Google) is combined with data from the actual website and merged into a powerful personalization campaign.

The last step is to create a campaign. The campaign is used to define the entry segment outlined above but also a goal event for evaluating the success. You may also add additional rules when the campaign should trigger.

Let us now get back to your online store.

You are tracking your user behavior. If this behavior matches a certain pattern (segment) the Intempt server is sending back a signal to the script running on your website.

An additional script running on your store (an event listener) now overrides every category link as soon as the signal arrives.

A default visitor would have this experience when browsing the t-shirt section:

...while a first time user coming from Google and browsing at least three t-shirts above a \$100 in a row would see this:

### Retention Personalization Use Case: Farfetch

#### What is Farfetch.com?

Founded in 2008, Farfetch is an online luxury fashion retail platform that sells products from over 700 boutiques and brands from around the world. It already served more than one million customers and is valued one billion USD.

Products sold on Farfetch are often high priced and target fashion enthusiasts and fashionistas worldwide.

#### Increase Farfetch’s Average Customer LTV Via Email Marketing and On-Site Behavioral Messaging

To increase the average customer LTV generated on Farfetch, the platform would first create an email marketing campaign sending out engaging different newsletters to three user segments based on previous purchase behavior and resulting LTV:

• High end customer (spent > 5000 USD)
• Mid range customer (spent > 2000 USD & < 5000 USD)
• Low end customer (spent < 2000 USD)

Each segment will receive its own newsletter template containing suitable information for this group of subscribers.

Whenever a user (who browsed Farfetch via an email campaign link prior) returns to the site, this user would get to see a personalized engaging offer depending on the previous purchase behavior of this visitor.

How do we get there?

##### Email Campaign Setup

The three segments are created by retrieving sales data from the shop system and storing the results in a email list field.

The email templates also include links to Farfetch with an unique ID attached (using a mail merge field). This ID is tracked by Intempt’s platform whenever a reader clicks on the link attached in the email, allowing a precise segmentation, attribution and data synchronization across platforms.

##### Create Behavioural Segments

To display personalized offers, Farfetch would create three segments in Intempt (mirroring the email campaign):

1. An identified user (tracked via the ID attached to the email link)
2. Who is either labelled (via a Mailchimp list or CRM file import) as a high end customer, mid range customer or low end customer
3. Visited Farfetch more than one time (is a returning user)

In order to identify users properly on the website that have been approached via email before, the email campaign list needs to be uploaded to Intempt using a CSV file.

A high end customer segment in Intempt’s segment editor:

##### Campaign time

Farfetch already added a grey bar to their website displaying offers and announcements (located at the top below the main menu).

Any user who does not fully meet the campaign requirements:

• Is part of a segment (high end, mid range, low end)
• Has clicked on a link encapsulated in the newsletter and browsed Farfetch
• Has later browsed Farfetch again

would get to see default product pages:

Instead, a high end user who clicked on an attached email link once and returned to Farfetch later would get to see this offer:

...whereas a mid range user doing the same steps would get to see this:

...and finally a low end user meeting all campaign conditions would see this:

Please note Intempt can dynamically override any existing page content. Except from the tracker installation no further IT involvement is required, which is why Farfetch would use the existing bar.

It is possible to adapt the segments within Intempt at any time because MailChimp and Intempt share the same database and use user specific IDs.

As an example, you may in- or decrease the spend threshold of the three segments (“high end”, “mid range” and “low end”) simply by changing the segment criterias directly in Intempt.

You may also dynamically override the imported database based on each user’s action. If a user classified as “high end” does a “low end” purchase (e.g. < 2000 USD), then this user could be labelled as “low end” automatically by tracking the cart value at the checkout (using Intempt’s trackcharge feature).

If Farfetch’s email marketer then re-uploads this list from Intempt to MailChimp, the database will be up-to-date mapping the actual and individual visitor behaviour when creating the next email campaign.

## User Personalization Helps You To Increase AOV or LTV

You may increase first-time AOVs by delivering personalized websites helping new users to navigate and browse to the right products they are likely to buy.

Stitching behavioral strategies together by combining channels such as email and on-site helps you to boost your average customer LTVs. Target users based on their behavior across channels while you still retain full control over your campaign.

Just be personal, in a smart way.