How to build a conversion funnel

What’s in the funnel?

The first place to start in putting together a conversion funnel is to define success. All businesses boil down to a single transaction — the first time a customer pays. Most of the time this is the bottom of the funnel.

The chosen success measure at the bottom of the funnel should be based on successful business outcomes. If your business needs a $100 basket checkout to have the requisite gross margin of 20% then having ‘completed checkout’ at the bottom of the funnel isn’t success for your business and the funnel isn’t helpful.

There are exceptions such as SaaS products which are typically dependent on recurring monthly payments. If the typical indicator of being a recurring monthly customer is paying in month 3, then that’s what needs to be at the bottom of the funnel. Paying in month 1 and 2 are steps in the funnel.

The step at the top should be the first time that a customer comes into contact with the product in a measurable way. Steps in the middle are everything the user needs to do to go from there to the bottom.

Each step should be necessary

The steps need to be necessary to achieve the goal at the bottom — treat them like stage gates that the user needs to go through to reach the end. Frequently, people include steps that have a correlation with the final goal, but are not necessary.

Adding items to a wishlist before adding it to the cart may have a higher correlation with successful checkout than a customer who doesn’t. But it’s a feature that increases conversion, it’s not necessary to get a customer to successful checkout.

Including unnecessary steps makes it difficult to analyse a conversion funnel.

When calculating customers who added an item to their cart, do you only count the customers who added an item to their wish list? The funnel will not count customers who added an item to their cart without adding an item to their wish list, potentially excluding a significant proportion of successful conversions.

To take it further, what if next month less users added an item to their wish list, but more added an item to their cart? It’s important to distinguish between what a customer needs to do and how a product is optimised to drive a user to do it. Conversion funnels should only document the former.

Necessary .. but also valuable

On the flip side, the steps also need to be valuable. If a customer needs to fill in 3 different address fields for a purchasing checkout but the conversion rate from one step to the next is 99% then it’s not important enough to include in your conversion funnel. Doing so results in loads of data points, and superfluous information detracts from valuable information.

Did you do it or not?

Going to the next step in a conversion funnel should be a yes/no answer. Conversion funnels get muddled when you try and quantify the degree of success.

It’s very tempting to think that having a group of users with 3 items in a cart is much better than 1 and this should be reflected in the conversion funnel. But that detracts from the core purpose of the funnel: To measure the effectiveness of your product at getting a customer to checkout.

Customer outcomes, not events

The lazy approach is to aggregate ‘events’ such as traffic (page landings), adding items to a cart and successful checkout. An aggregate count can look like:

The goal of a conversion funnel is to assess the effectiveness of your product at achieving the customer outcome at the bottom. That means knowing the number of unique users who performed a step in your funnel — not having data skewed through single users repeating actions.

A customer-centric conversion funnel looks something like:

The conversion funnel by aggregate count looks healthy but masks the effect of power users. 90% of the items added to the cart were checked out, but only 25% of customers successfully checked out, indicating that the checkout process is actually performing poorly.

Percentage of the previous step when graphing over time

The most common mistake is to measure each step as a percentage of the top of the funnel.

Conversion funnel — As a % of traffic

This approach means that conversion rates are affected by the steps before them. Step 3 will decrease if Step 2 decreases. It makes it clear how many customers reach that step of the funnel, but at the cost of something more valuable — how well that step performed.

The next most common is counting the ‘steps’ over time like below.

Aggregate event counts

The big downside is that when your business grows quickly, it makes it hard to see how well each step is performing over time relative to each other. I prefer to look at this graph in conjunction with the graph below, where you measure the conversion from one step to another.

Conversion funnel — Stepped over time

It’s worth repeating — the most valuable information in a conversion funnel is how each of the steps are performing. The 2nd graph has a heavy bias towards magnitude rather than performance. Looking at the 2nd graph we can see that there was a significant increase in inbound traffic (likely due to advertising in a new place), but we don’t know how whether that traffic was full of relevant customers likely to purchase or not.

The 3rd graph tells us that:

• Users were less likely to view items — it’s expected that advertisements will not be perfectly targeted to relevant customers.
• But those that did view items, were more likely to add items to their cart and just as likely to successfully check out.

Looking at both graphs in conjunction gives a thorough understanding of the change in customer traffic.

Don’t forget

• A conversion funnel is there to measure how effective you product is at getting a customer from A to B. Measure performance per user and ignore the relative size of different customers.
• Don’t try and measure other things at the same time.
• One funnel is not going to tell you everything you need to know about a business’ performance. But one funnel should be a sufficient ‘thermometer’ check. Use secondary graphs to develop a more granular understanding: Like breaking the funnel down by device to catch bugs, or mini-funnels to see if small parts such as filling out forms are performing badly.