Funnel Analysis

What is Funnel analysis?

In one sentence: A powerful method of analysis which depicts the most pivotal points in the entire life path of a user. Or in a diagram:

funnel analysis

funnel analysis in one picture

How do Funnel metrics work in practice? What are the biggest questions and stumbling blocks? How could we extract „actionable” knowledge from them? How can you create it for yourself? Let’s see! (Note: the below are not strict rules, they serve more as guidance. In special cases you can always be creative. :-))

Why is Funnel analysis good? When should you use it?

This type of measurement is typically useful if you want to check out a linear usage-flow. Basically, it is about going step by step and calculating how many users arrived to a certain point of the given process. Or from a different perspective – how many users churn out to this point? The process itself can be a simpler thing. Like e.g. filling out a registration form…

gmail funnel

Gmail registration + funnel metrics (the picture is just an illustration)

… or it can be a little more complex, like e.g. an Onboarding process (e.g. the picture in the first paragraph, which depicts the steps of a memo creating app.) But you can do a measurement on the complete page/product from the first visit to the point of purchase.

In each case the key word is linearity. So we try to model processes with these funnels where you cannot skip through certain points (those strictly follow one another), and where we can arrive to the destination on a clearly marked path. On a webshop page for example the path can be set up easily:

Step 1: Landing at the page.
Step 2: Browsing through the products.
Step 3: Visit a specific product.
Step 4: Put a specific product in the basket.
Step 5: Filling out the purchase details.
Step 6: Placing the order.
Step 7: „Thank you for your purchase!”

IMPORTANT: This method considers the viewpoint of the user, we measure their development. So even if the user browsed different products, we count it as one. We should always consider our users as those who have „gone a level up”: so if someone placed a product in their basket, tick the 4th step, and the goal will then be to get them to the 5th step. If someone puts 80 products in their basket, they still only get one tick for the 4th step. (Exact numbers are measured on other metrics anyway.)

The pages of SaaS (Software as a Service) and UGC (User Generated Content) are roughly as simple as E-commerce. The situation gets more complex with media pages, which is why we rarely use the funnel analysis with those.

I would recommend the most well-known funnel model. Based on Dave McClure’s AARRR model a user goes through 5 steps when using our online product or webpage:

1. They arrive at the page
2. They start using the product
3. They return to the page
4. They make a purchase
5. They refer the page

Dave McClure AARRR model

Dave McClure AARRR model

As you can see, with Dave McClure linearity is not 100% (for example, someone can recommend you without making a purchase), but observations suggest that in most cases this will still be the realistic order.

How can you create it?

1. Define certain steps! For ease of understanding I would suggest making between 5 to 10 steps. If you have a lot more, you risk getting lost in the data, if you have much less, then you don’t have a Funnel. 🙂

(Additionally: Regardless of this, you can always create funnels at any point in time. So e.g. you have a process overarching a page where one step is registering, then the process of registration can have a separate sub-funnel. This logically builds into the large cone, but on a visualization level it’s worth separating from the large one so you don’t get distracted.)

2. Try to measure them somehow!

The measurement can be:
• Google Analytics event settings
• Mixpanel
• Own data tables (e.g. Tableau or with the support of GoodData)
• With anything that comes easy to you

3. Once you are done with all of this, chose the simplest, most easy to read visualization. So NOT this:

shitty visual funnel

Unfortunately this is a real chart

And not this:

shitty visual funnel

Incorrect funnel analysis visualization

And only if you must, then this:

 

shitty visual funnel

Google Analytics-type Funnel visualization

The simplest, best visualization is the Bar Chart. It’s no accident that I mentioned this in the first paragraph.

funnel analysis

Funnel analysis in one picture

I recommend this in each case!

How can you get many action items from one analysis?

This is a key question. Each measurement is done so we can make a change or improvement!

So how can we get useful information from a Funnel analysis? In 3 ways.

One is the so-called bottle neck examination. So we check where the chart drops the most. If we see that during the registration everyone gives their name, email address and password, but almost everyone disappears at captcha, then we can suspect that there is a problem there (e.g. the captcha is not readable). But of course it’s not always this easy. There are more difficult steps where we don’t expect many users not to churn. E.g. when entering data into the bank details field, there’s always a big drop. So we should not search for the „bottle neck” where most users disappear, but where this is relative – compared to our expectations – where most people drop out. (Our expectations are adjusted to many things – we can use our „common sense”, but using market benchmarks or past data is more advised).

Another favourite metrics of mine is measuring the elapsed time between certain steps. That is, how much time if takes for someone to put a product in the basket, starting from when they begin to check out the product, which tells us a lot about our product and communication as well. The time it takes to purchase a $2000 laptop will never be quick as buying a $2 battery.

The third measurement method is segmentation. If we can find user-segments who are more successful at certain steps than others, then we receive great feedback on who to target in the future, as well as on what brings things to a halt. For example, if we can see that men buy female lingerie without a problem, and a large portion of women eventually don’t buy any, then it may be worthwhile to think about our main message to target men to be „buy your wife some lingerie”.

Summary

1. Funnel analysis is a strong tool with which you can measure the life path and development of your users!
2. Define 5-10 steps and visualize them simply and in an easy to understand way.
3. Analyze your Funnel: Check at which step most people drop out at, and which segment is the strongest! Good Luck!

And if you want to be notified first about new content on data36 blog (like articles, videos, handbooks, etc.), sign up for the Newsletter!

Tomi Mester

AS A SUPPLEMENT:

1. According to Tamás Geiger : In the GA premium we have individual channel functions as well!
2. Supplement to Tamás Lindwurm: “if we want to broaden our funnel then it’s better to begin at the bottom (with the most committed clients) and moving up from there”.

share:Tweet about this on Twitter4Share on LinkedIn5Share on Facebook1

← Previous post

Next post →

2 Comments

  1. Hi,

    Very interesting subject. The topic take my atemption, because I work with bioinformatics and a way of analyzing the results is using a graph in funnel. But, I don’t find any resource about the subject.

    Congrulations

Leave a Reply