It’s easy to get lost down the rabbit hole of metrics for your business. When it comes to getting the most out of your website performance, only certain metrics are what you can consider key performance indicators. The trick is figuring out which ones.
What are key performance indicators?
A key performance indicator (KPI) is a quantifiable activity used to measure how a key aspect of your business is operating or how much volume it’s receiving. For websites, this can include “sales volume,” “number of visits,” “average cart value,” and a variety of other metrics.
KPIs are metrics, but not all metrics are KPIs; the terms aren’t mutually exclusive. Just because some information is measurable, doesn’t mean that those measures are informative. It’s important to prioritize the most valuable KPIs.
Macro conversions vs. micro conversions
Your metrics are either leading to “macro conversions” or “micro conversions.” Macro conversions are the primary goal of your website, to convert user traffic into revenue.
On the other side of the coin, micro conversions are actions a user completes that are either on the path to revenue-generating macro conversions or not directly related to revenue-generation at all. Micro conversions are process milestones or secondary actions
This can be like number of checkout visits, pageviews per user, cart adds, etc. But micro conversions are not KPIs – learn more on why you should not focus on micro conversions.
7 Ways that Predictive Analytics Is Transforming Ecommerce
Predictive analytics help you understand what your customers are going to buy before they do.
Which metrics are not KPIs?
On average, the metrics that can take a backseat are pageviews per user, click-through rate, bounce rate, and the average time a user spends on-site. Basically all metrics that are not money. Is it better that someone spends more time per visit on our site, or less? Impossible to say. Is it better if someone spends more money on our site or less? No question.
Other micro conversions include visiting the checkout page, adding a product to the cart, the number of pages visited, and several others. These may indicate interest, but don’t help you persuade the user to click “purchase.”
Should we bother monitoring micro conversions at all?
There is no dollar amount attached to these actions. While these things are great, they don’t lead to getting paid for your services.
That’s not to say you shouldn’t monitor these metrics and micro conversions at all. It’s helpful to follow these metrics because they inform you of which areas of your site or service need attention and lay the groundwork for future conversions.
You should just make your decisions based on the “money metrics”, not micro conversions.
Here are the four KPIs that really matter for website optimization
1) Revenue per visitor
Revenue per visitor (RPV) is simply the average amount of money you made per visitor.
RPV is the god metric because it provides direct insight into which website actions ultimately translate into dollars.
2) Conversion rate
If you can’t measure money (e.g. it’s a lead generation website, nothing is sold), conversion rate is the next best thing.
Unsure what conversion rate is? Head over to lesson #1 in our Beginners Guide to Conversion Optimization.
3) Customer Acquisition Cost
Gaining customers requires a certain financial investment. Your budgetary number for how much you spend to acquire a customer is known as the “customer acquisition cost” (CAC). The initial cost per customer should be less than the profit made once they’re converted (or be lower than LTV if you know it).
The goal here is to perfect your processes so you minimize your expenditures and time, thereby maximizing your profits.
When a model is balanced, it looks like this:
Courtesy of Forentrepreneurs.com
Essentially, the biggest reason for the collapse of many startup companies is because their version of the above graphic is inverse. Their CAC exceeds their LTV, or lifetime value, for their customers.
Once ConstantContact figured out their LTV, they were able to spend much more on acquisition since they know how much they can:
“The average Constant Contact customer stays with the company for 45 months. At their $39/month price point, this means that a customer generates about $1,800 in lifetime revenue, allowing Constant Contact to make money at a CPA (cost-per-acquisition) of $450.”
Be careful giving this too much weight, though: sometimes LTV metrics lie.
4) Customer Lifetime Value
One could make the argument that the most important KPI should be LTV, which is “a prediction of the net profit attributed to the entire future relationship with a customer.” Essentially it’s how much money you make off of a customer over their lifetime as your customer.
Evaluating your most useful customers over their lifespan is essential. You need to establish how much a client is worth over the long haul.
For example, let’s say three customers join your subscription product at the same, paying $50 per month. One stays as a customer for a month, second one for 3 months and the third one for 2 years. While all 3 of them paid $50 when they joined, their LTVs were *very different*: $50, $150 and $1200.
The customers do not provide you equal value, and the quicker you can predict who is a high LTV customer, the better.
However, it can be difficult to measure LTV and only around 40% of companies are able to measure customer lifetime value. The only way is to gather information and assemble it in a predictive model, which features extracted data extrapolated and used to predict customer behavior.
What goes into the LTV formulas?
Calculating LTV depends on your business model. Here a typical formula for calculating LTV in SaaS:
Courtesy of TechCrunch
(MRR is monthly recurring revenue).
A customer making a direct purchase is high value, but these customers also have other behaviors that are valuable to your company as well. (Examples including indirect marketing, such as word of mouth and referrals.) Tracking these traits and behaviors, like site visits, factors into creating the most useful LTV model.
Why are these formulas important?
The main value of these formulas is to give marketers an idea of where to invest their budget and how much to spend to acquire a customer. Having the upper hand on prioritizing and engaging high-spenders with hyper-focus makes marketing much easier.
Additionally, knowing your company’s income over time for a particular customer allows you to decide if you can discount services while still making financial gains, build value upfront, and drive customer traffic to your page.
With the right tactics, you’ll be able to develop a strategy to earn back the cash you spend, and even profit.
How to Model LTV
How can you predict customer’s LTV at the time of acquisition? It’s hard.
The best way to do this is to use a software to automate your data collection, and software services like Custora and RJMetrics are helpful in achieving this. Various Google spreadsheets also exist for doing your own modeling.
Predictive LTV model
So how does one go about building a predictive model? Hire a data scientist! For those of us who don’t have one in-house, here’s how you would go about it.
A good working equation is:
(Average Value of a Sale) X (Number of Repeat Transactions) X (Average Retention Time in Months or Years)
The gym membership
Here’s an example of a gym membership to illustrate this point:
“An easy example would be the lifetime value of a gym member who spends $20 every month for 3 years. The value of that customer would be:
$20 X 12 months X 3 years = $720 in total revenue (or $240 per year)
Now you can see even from this hypothetical example why many gyms offer a free starter membership to help drive traffic. Gym owners know that as long as they spend less than $240 to acquire a new member, the customer will prove profitable in a short amount of time.”
The gym now has developed a budget per customer and can begin the process of increasing their profits. (If you don’t sell anything for money, your main KPI can be a number of leads and conversion rate.)
How does this look in reality?
The folks over at Kissmetrics were modeling the LTV of the average Starbucks customer, using data from 2004-2012. Though the data might be out-of-date in 2017, it certainly serves as a good example for developing a predictive LTV model.
The first step in crafting their perfect LTV model was feature extraction, a process which examines customer data and zeros in on behaviors that lead the user to convert. These should be able to apply in other similar predictive models.
Feature extraction data
In general, three factors went into evaluating an average Starbucks customers’ activity: customer expenditures per visit, number of visits per week, and the average customer value per week. (The average customer value per week was found by multiplying both the expenditures and number of visits per week together.)
Once they had averages of these three datasets containing customer variables, it was time to calculate. Using the average values calculated above, and allowing for certain constants, they were able to create formulas to calculate the overall lifetime values.
Three common formulas were employed: a simple LTV equation, a customer LTV equation, and a traditional LTV equation. The outcomes of all of these equations were averaged together to find the LTV.
Every company’s LTV is going to look a little different. The key is to focus on which customers interact with and purchase from your company over time, and increase their customer satisfaction. Invest in users, rather than expending your resources on “cheap” or “average” customers who ultimately might churn or invest little value.
In general, while micro conversions as metrics can be telling, they aren’t your KPIs. The four key performance indicators that do matter include:
- Revenue Per Visitor
- Conversion Rate
- Customer Acquisition Cost
- Lifetime Value
Knowing how much it costs to acquire customers informs how much money your pour into your merketing budget. Typically, conversion and revenue per visitor rates are relevant because is gives you current, raw data on purchases already made by your customers, whereas predictive modeling for LTV is less accurate and therefore more likely to err.