How to Calculate The Word of Mouth Coefficient in Amplitude

The Word of Mouth Coefficient has become a popular way to measure the effectiveness of word of mouth as an acquisition channel for a product.

Invented by Yousuf Bhaijee, and popularised by his Reforge article, the “WOM Coefficient” is a statistically proven way to understand the extent to which your active users are talking about your product and therefore driving more new users.

Let’s dive in and see how we can set this up in Amplitude!

The definition of Word of Mouth Coefficient

First of all let’s review the definition of WOM Coefficient:

source: Reforge blog

One reason I love WOM compared to its virality metric cousin “K-factor” is that it’s really simple! There are 3 components:

• New Organic Users: users first visiting your product, and not coming from a paid campaign
• Returning Users: all users who are visiting your product that are not new
• Non-Organic New Users: users first visiting your product coming from a paid campaign

Therefore, when we make the calculation, we just need to bucket users in 2 ways:

1. Did they come from a paid campaign or not?
2. Are they New or Returning?

Let’s see how we do that!

How to calculate Word of Mouth Coefficient in Amplitude

Firstly, let’s briefly discuss 3 methods of “attribution” that enable you to identify whether users came from a paid campaign:

1. UTM: the typical approach for web products, paid campaigns are set up with UTM links, and you can identify these users in Amplitude using the UTM Source user property.
2. MMP: a “Mobile Measurement Platform” is the equivalent of UTM for mobile apps, which you can implement with a tool like Branch or AppsFlyer. These can be tricky to set up and are vulnerable to becoming a black box due to ever increasing user privacy on iOS.
3. Attribution Survey Question: I prefer this approach to MMP for mobile app attribution. Simply ask your users in your onboarding flow how they heard about your product, and then save this answer as a user property in Amplitude.

An example attribution question taken from the Agapé app onboarding

Whichever method you use, the Amplitude chart for WOM Coefficient is pretty much the same. Here it is for UTM:

This chart is really easy to reproduce:

1. Add the event New User, and set UTM Source = (none)
2. Also add the event Any Active Event.
3. Finally select the Formula tab and add the formula. Here is the formula so you can copy-paste it.

UNIQUES(A)/(UNIQUES(B)-UNIQUES(A))

So why does this work? We are using the Amplitude formula tab to make a calculation. The formula is composed as follows:

Numerator

• UNIQUES(A)
• … is New User with UTM Source = (none)
• This is exactly “New Organic Users”, which is what we wanted!

Denominator

• (UNIQUES(B)-UNIQUES(A))
• UNIQUES(B) is Any Active Event, which is exactly all New Users and all Returning Users for the time interval
• UNIQUES(A) is New Organic Users
• … so we have New Users + Returning Users - New Organic Users
• Reordering that we have Returning Users + (New Users-New Organic Users)
• And (New Users-New Organic Users) is exactly Non-Organic New Users.
• So we have Returning Users + Non-Organic New Users as we wanted!

And that’s it! We have the formula from the definition of WOM above! You can now see your product’s Word of Mouth Coefficient and how it changes over time.

Feel free to change the time interval to weeks or months if that better suits your product, the same formula still works.

Adapting the chart for MMP or Attribution Question

If you’re using an MMP, you just substitute UTM Source for the equivalent for your MMP and filter for the Organic campaign links (i.e. not paid).

If you’re using an Attribution survey, you substitute UTM Source for the user property where you set the survey answer, and filter for the answers that you consider organic e.g. “Friends” “Family” “Word of Mouth” etc.

Analysing your WOM coefficient

Now you can measure your WOM coefficient, how can you interpret the data? Here’s a few suggestions for things to dive into or look out for.

New Release Impact

Let’s imagine you release a new feature which is controversial and worth talking about. You will likely see a spike in the WOM Coefficient chart.

New Feature Release on Jul 31st creates an uptick in Word of Mouth Coefficient

Seasonality Impact

Here’s an example from an e-learning product for a qualification which has 4 exams a year. You can see the WOM Coefficient mirrors this and also spikes 4 times a year in the month before the exam, when students start to talk to their classmates about studying tools.

Breakdown by key dimensions

As usual, breaking down a metric by its key dimensions — in this case country — is interesting. Here we see the WOM coefficient is highest for India and lowest for South Africa.

Need Some Help?

Through my tiny Amplitude product analytics agency Permutable, I help startups and scaleups to set up and analyse their users using many techniques like WOM coefficient. As well as being an expert in Amplitude, I’ve seen a lot of cases across different industries, so I can help you use your data to identify the problems and opportunities for your product. Get in touch!