How to Forecast User Growth

Issue 21

Good Morning and Happy Teachers' Day 🙌🏻

In the last few posts, we discussed how just looking at MAUs can be detrimental to growth in the long run. Instead, we should decode growth by looking at users' segmentationcohort analysis, and distribution charts. If you haven't read the posts, you can do it later. Building a user growth model isn't reliant on these topics, so you can continue with this post and read them later.

For the new ones here, do check out the other posts that I have written if you haven’t

  1. Referrals: The Holy Grail of Growth

  2. Measuring Product-Market Fit

  3. Decoding Growth - Where is Growth Coming From

  4. How to Engage Users Through Gamification

  5. Understanding Positioning Through a Practical Lens

2,500+ people have subscribed to the growth catalyst newsletter so far. To receive the newsletter weekly in your email, consider subscribing 👇

If you aren’t familiar with the newsletter and online sessions, you can read about it here

Off to the topic,

The Need for a Growth Model

No matter where you work, be it a startup or a big company, people will expect you to set growth targets if you are leading a product or business.

We often take an audacious growth target that looks good to others. A hypothetical sample growth target sheet looks like this.

The company wants to achieve 1.8 million DAU by Mar '21. The argument presented in favor of the target is that "It's just 10% MoM growth, so we should be able to do it. Sounds right? After all, what are we even doing if not achieving this target."

Mar '21 arrives, and we are sitting at 1 or 1.2 million DAU, missing the target by a significant margin. And what do we do then? Take another audacious target. Albert Einstein said,

"The definition of insanity is doing the same thing over and over again, but expecting different results."

What growth teams do while setting top-down goals based on what looks good is nothing short of insanity. So, where are we going wrong?

We are focussing on output metric (DAU) and not inputs. The output or DAU is the final result. What you can influence is inputs, and then only the output will.

Now, if you are reading this, some of you might be feeling, "oh, we do it in the right way." Kudos if you are doing it the right way :)

Majority of teams that I have interacted with, unfortunately, do it the wrong way. Some of you may be feeling, "It's a top-down target. What can we do about it?" You can build a growth model to forecast DAU and MAU. Maybe the management won't listen to you in the next goal-setting exercise. But the one after that, they will.

Let's move on to see how we build this growth model. Where do we start? We start by understanding the inputs of the model.

Inputs and Output of a Model

A model requires inputs to provide an output. We already have the output — expected DAU.

There are two inputs we require to forecast DAU growth.

  1. Daily new users

  2. Day-N retention

Let's be clear on these terms. Daily new users is straight-forward.

Day-N retention measures how many of your users come back to your app on the Nth day. Day 0 is the day user first came to your product. Let's take an example to illustrate Day-N retention.

Suppose 100 users came on 1st September '20 for the first time. Of these 100 users, 50 users came back on 2nd September. The D1 retention here is (50/100) ~ 50%. You can do the same maths for other days. Here is a sample sheet to illustrate this for two different dates. About 100 users came for the first time on 1st September. The other columns represent how many of them came back on respective dates. Similarly, 80 users used the product for the first time on 2nd September. The other columns represent how many of them came back on individual dates.

Let's simplify this table in terms of D0, D1, D2…D7.

Now that we put the table in this form, it becomes easy to calculate Day-N retention.

Once we have understood Day-N retention, we can move forward to see how we can forecast the DAU.

Forecasting DAU

Suppose you launched the app on 1st August 2020, and it has been a week. You have one week's data that looks like this.

As we have 7 days of data, we can get D1 to D6 retention. To do that, we first transform the table into the form below and evaluate average new users and average retention for Day-N.

From the table above, we have the input metrics.

  1. Average daily new users ~ 80

  2. D1 ~ 42%, D2 ~ 38%, D3 ~34% and so on

As we have data for the last 7 days, we can predict the growth for the next 7 days because we don't know how retention will behave in the long run.

Assuming you don't plan to do any additional marketing exercise, the new users will stay the same for the next 7 days. We already know the retention behavior. Feeding it into the model (spreadsheet with all the calculations here)

As you can see, the model employs this formula to calculate DAUs.

There is one caveat in this model. What about users who came in the period "1st-7th August"? For those users, we need D7-D13 retention data to calculate DAU. As in this hypothetical scenario, we launched the app 7 days ago; there is no way to no D7-D13.

What we can do is make some approximations. You can check how retention for apps with similar retention (D1 ~42% and D6 ~25%) behaves in D7-D13. Usually, the dip in retention in the first week is higher than the second week. Let's assume the D13 retention would be around 18%.

Adding both cohorts' DAU contribution, the DAU for this date range will look like this.

So by 14th August, the app will have a DAU of 359. If we want to increase further (say get to 500 DAU), what should we do? One or more of two things,

  1. Increase daily news users

  2. Increase retention

If we don't have a way to increase any of the two, it won't be possible to reach 500.


There might be some questions around this. I am answering a few on the top of my mind. Please feel free to comment/DM me for any questions you may have.

Q: How can this growth model be applied to get MAU projections?

A: For MAU, the average N-month retention and average monthly new users will be the input metrics.

Q: What do I do if I am just about to launch my product?

A: Focus on acquiring the first 1000 users. It would be best if you have new users, not a growth model :)

Q: What do I do if I have an old product (>1 year) with a DAU of, say 1 million?

A: Good news! You can get D1-D365 retention and create a fantastic growth model

Q: How do we set growth targets now?

A: We can set targets on input metrics (retention and acquisition), and basis that an output metric like revenue or DAU

Q: Can it be applied to the Revenue Growth Model?

A: Yes, we can replace NU by New Revenue and User Retention by Revenue Retention and viola, we have a revenue growth model 😊

Q: What's next week?

A: Next week, we will cover "How to create Growth Loops in your product" The best products have strong growth loops: virtuous cycles that accelerate your growth

See you next week.