This is the kind of report that will shape a lot of discussion over the next few weeks about where the subscription app market is heading.

Below is my breakdown with key takeaways and reflections across the full funnel: acquisition, activation, conversion, retention, and winback. Some of the trends were expected, and we all saw them coming. But others genuinely surprised me. And a few things I don’t agree with. Let’s dive into it.

Read more

Subscribe now

Today’s issue is all about ARPU - what it’s good for, where it fails, and how people often misuse it.

ARPU shows up everywhere - in experiments, forecasting, and profitability estimations, and it’s easy to treat it as a simple, universal metric. But when I point out problems with ARPU, it often doesn’t go over well, especially with growth marketing. People trust ARPU more than they should. So I’m putting everything an analyst needs to know about ARPU in one place, so we can link to it instead of arguing about it every time..

If your team uses ARPU to judge experiment impact or estimate profitability, consider sharing this. The goal isn’t to force anyone to change how they work, but it’s to make the tradeoffs clear and keep ARPU from being used as the one answer for every decision.

Why do we need ARPU?

ARPU is a finance metric that gives a high-level view of revenue by combining 3 things into one number: (1) revenue, (2) number of customers, and (3) the average value per customer:

Most tools and providers use the same basic definition:

• Stripe: What is average revenue per user? Why it matters and how to calculate it

• Appsflyer: Average revenue per user (ARPU)

• Adjust: What is average revenue per user (ARPU)?

• Chargebee: What is Average Revenue Per User?

• Amplitude: What Is Average Revenue Per Customer (ARPU)? How To Calculate It

• RevenueCat: Realized LTV per Customer Chart (being different, huh)

• Appnomix: A Guide to ARPU: The App World’s Most Underrated Revenue Metric

All of them say the same thing:

The problem is that in real life, it’s rarely that simple. There are many valid versions of ARPU depending on choices like:

• What is the “time period”? (day, week, month, cohort window, billing cycle?)

• Which dates define revenue? (purchase date, recognition date, refund date, renewal date?)

• Which dates define users? (signup date, trial start, conversion date, renewal date?)

• What counts as “revenue”? (gross vs net, refunds, chargebacks, taxes, discounts, promos, upgrades? See why not all revenue is created equal.)

• Who counts as a “user”? (all users, trials, paid users, refunded, discounted users?)

So the “standard” ARPU formula usually hides a lot of decisions. Picking the right version for your product is hard, and it should be defined with finance, not invented separately by marketing or product.

The irony, though, is that marketing and product teams often rely on ARPU more than finance does. Finance usually has better revenue metrics already. ARPU becomes a shortcut metric for teams that need a single number to track performance.

That’s why, over the last decade, every payment, analytics, and growth tool built “default ARPU” calculations so non-finance teams could measure it without deep modeling.

That part is helpful, but it created a new problem: people started using ARPU for everything - onboarding tests, paywall tests, forecasting, growth projections, retention, prioritization, even LTV, often without understanding what ARPU is actually measuring.

And ARPU is easy to game by accident. Averages can go up when the denominator goes down. If an experiment reduces the number of paying customers but keeps a few high spenders, ARPU can rise even while total revenue falls. That means an “ARPU lift” doesn’t automatically mean the experiment was good.

I saw this firsthand at MAU, mobile app AdTech and MarTech conference in Vegas last year: a well-known growth consultant opened with a claim like “I’ve run 3,670+ monetization experiments, each with 60%+ lift in ARPU”. But if you understand the math, that statement is a big red flag: either many of those tests hurt volume, or the metric was being used in a misleading way. In other words, he was effectively saying that either he ran 3K failed revenue tests, or he has no clue what he is talking about.

To summarize, ARPU is worth tracking, but it’s increasingly overused and overtrusted. It’s not a one-size-fits-all decision metric, and it can point you in the wrong direction if you don’t define it carefully.

The difference between the main types of ARPU

Read more

Subscribe now

If you’re a monetization analyst, a new plan launch isn’t exciting. It means rerunning profitability, rebuilding plan mix, updating forecasts, and breaking historical comparability across your dashboards.

Every time I hear the team created a new product_id or price_id in Stripe, or launched a new plan, I know what it means: over the next few weeks, I’ll basically need to redo everything. Did you know that some companies end up with more than 100 product IDs (active or inactive)? On top of maintaining the taxonomy of product and price IDs, I also need to make sure the plan mapping stays consistent so analytics and reporting remain accurate.

After a few years of watching businesses grow, and watching the number of plans grow along with ARR and the customer base, I developed a hypothesis: the more diverse the plan catalog, the faster the growth.

This [secret] hypothesis is based on 26 apps I currently support, which isn’t statistically meaningful in any way. I also couldn’t find any studies or analyses on this, so I was happy to collaborate with ChartMogul, one of the largest analytics platforms for SaaS and subscriptions, and use their data to either prove or reject the idea. And magic happened! ✨

Below is my analysis of data from 8,000+ companies, including historical growth rates and plan data provided by the ChartMogul team. The dataset is anonymized, and it covers a wide range of businesses, from mature enterprise SaaS to small, fast-growing apps. The goal is to test whether fast-growing companies tend to maintain a more diverse plan catalog. I’ll look at how the number of plans relates to growth, and whether more plans are associated with faster growth, or whether fewer plans correlate with slower growth. This post walks through the data, the methodology, and what we learned. This analysis was also published here - ChartMogul Insights.

Pricing Plans: More Options, More Problems

Before diving into insights, let’s talk about plans.

There are countless pricing plan options today - monthly, annual, quarterly, weekly, lifetime, and more. Some companies offer a single plan, while others maintain 20, 50, or even more. Pricing can be a simple flat fee, or it can scale with seats, active users, transactions, revenue, or other usage-based metrics:

Working with plan data isn’t easy.

Maintaining a consistent plan taxonomy is painful for most SaaS and subscription businesses. Plans live across billing and distribution systems, like Stripe or Adyen, and the Apple App Store or Google Play, and each handles payments, stores subscription metadata, and exports data differently

If you offer multiple payment options (for example, iOS and Android apps), you need to merge very different data sources into one cohesive report. That report has to use consistent definitions for ARR, “active” customers, churn, and more, while still allowing segmentation by plan, billing period, price, and other attributes.

This almost always requires a sophisticated, flexible plan-mapping layer, for example, mapping a product_id from Apple to a price_id from Stripe, and translating price_564738589 or IDS_5668 into readable values like pro_49.99_7d_trial or lite_69.99.

If you’re working with intelligent AI systems, you also need to provide plan context, such as which plans are active in each region, which prices users see, how and when discounts are applied, which funnels each plan belongs to, and more.

All in all, reporting on something as simple and essential as plan mix isn’t easy. It requires infrastructure and ongoing maintenance. As an analyst, I can’t emphasize enough how much easier this becomes with tools designed to handle the complexity of subscription data. As you add more plans, it’s important to keep the plan taxonomy and configuration as clean and consistent as possible to make your analysis and reporting accurate.

Getting started: definitions, data, and methods

Definitions and metrics

• Plan mix: The set of active pricing plans a company offers and how customers are distributed across those plans.

• Plan length: The plan billing period (monthly, annual, 3-month, 6-month, etc.).

• Customers: Active paying customers currently subscribed to a plan.

• Tenure: Company age as observed in ChartMogul, specifically, the number of years ChartMogul has data for that company.

• ARR: Annual run rate as reported by ChartMogul.

• ARR growth per year (derived): ARR growth ÷ tenure.

• Customer growth (derived): Net new customers per month.

• Growth score (derived): A combined ARR and customer growth score per organization, weighted 60/40 toward ARR to avoid cases where customer growth is high while ARR is zero.

Data sample overview

• Total sample: 8,193 organizations.

• Organizations with at least one recurring plan: 6,875.

After normalization and removing organizations with missing or incomplete data, the final sample included 4,279 companies with the following ARR split:

Here is the plan mix across all these companies:

We see a small bump in the 4–6 plan range.

💡Overall, 54% of companies offer 10 or fewer plans, and 46% offer 11 or more plans.

I didn’t directly model plan mix, but it still matters when interpreting growth. The number of active plans usually increases as a company matures, improves customer segmentation, and becomes more advanced in how it monetizes. Teams can also expand monetization through add-ons, one-time transactions, and ads in freemium products, in addition to subscription plans. My analysis is limited to plan mix only, focusing on how many recurring plans are offered, their lengths, and their distribution.

Methodology

There are multiple ways to define growth. I started with ARR growth by calculating ARR growth per year for each company, then clustering companies into 5 segments (using k-means) based on their ARR growth.

Next, I did the same analysis for the number of customers, because in SaaS, revenue growth does not necessarily imply customer growth. Given the breadth of ChartMogul’s dataset, which includes SaaS companies of many types and sizes, we have cases of “rabbits” and “elephants,” and an entire zoo (in the sense of Christoph Janz’s 5 ways to build a $100M business). To account for this, I also created a new customers per month growth metric. As expected, customers’ growth looks different from ARR growth:

💡 In our sample, 47% of companies grow by more than +$270K ARR per year, but only 17% add more than 10 customers per month. This suggests either that the dataset is skewed toward companies serving large customers (“whales”), or that a meaningful share of ARR growth comes from expansion revenue, which often correlates with offering more plans!

To build growth signals that were comparable across companies of very different sizes and ages, I used both ARR growth per year (ARR growth divided by tenure in years) and subscriber growth per month (the net change in paid subscribers divided by the number of observed months in the dataset). I then combined these into a single score, weighting the metric toward ARR (60/40). This gave us a simple “velocity” measure: not simply what happened in one particular month, but the average rate of growth over the observed window.

Next, I bucketed companies into 5 clusters:

1. Early breakout (“On fire!”): explosive growth, with more than +$330K/year in ARR and 6K–7K new customers per month.

2. Fast-growing on track (“Winning the market”): strong growth, with around +$270K/year in ARR and 1K+ new customers per month.

3. Mature but accelerating (“Hold in there” / “Don’t touch anything”): steady compounding, with an average of +$165K/year in ARR growth.

4. Stagnant incumbents (“This doesn’t work”): low momentum, with less than roughly +$1.5K/year in ARR growth and fewer than 10 new customers per month.

5. Declining (“Pivot deck incoming”): companies in contraction, with zero or negative ARR growth.

Caveats: Buckets were assigned using thresholds on a combined growth score (ARR growth per year + subscriber growth per month). There was one override: if both ARR growth and subscriber growth were non-positive, the company was labeled Declining regardless of its combined score. Early Breakout was reserved for companies that were not only high-scoring but also relatively young, since compounding rapidly early in a company’s life is meaningfully different from compounding rapidly after a decade of iteration.

What we learned from analyzing the plan data

1. More plans tend to reflect higher and faster growth.

2. Expanding plan offerings is positively correlated with higher revenue and larger customer growth. Companies that introduce additional plans often create more pricing and packaging paths, which can support both monetization and acquisition.

3. There are exceptions. Early breakout companies often demonstrate the fastest growth regardless of plan strategy.

4. Adding more plans may influence not only the magnitude of growth but also the speed at which companies scale.

Let’s break these down.

1. The larger the ARR is, the more plans the company offers

There is a strong positive correlation between the number of plans offered and company revenue

But remember, correlation isn’t causation. The fact that high-ARR organizations tend to offer more plans doesn’t necessarily mean that adding plans increases revenue. Plan complexity may simply follow growth: as companies scale, they introduce tiers for segmentation, enterprise packaging, regional pricing, and bundles. In other words, plan count can be a result of momentum. To be more confident that plan count is a driver, we’d need to compare growth patterns across companies with different numbers of plans.

2. More plans reflect higher growth across all company stages

Companies have very different growth rates, shaped by many factors, most of which we can’t quantify in this analysis. What we can do is bucket companies into groups based on growth (early breakout, fast-growing, mature, stagnant, and declining) and check whether the relationship between growth rate and plan count remains strong within those groups:

Clearly, both the average and the median number of plans are higher for fast-growing companies, regardless of company size or starting ARR. To be cautious, I report 2 plan-count metrics - one normalized and one non-normalized, because the distribution includes influential outliers. The relationship holds across most groups, with the exception of early breakouts, which is also expected: these are newer breakout companies with less tenure, so obviously, less time to expand their plan catalogs.

3. The more plans companies add, the higher the growth rate

Now, let’s see if adding more plans reflects a better growth rate:

The same pattern holds across most segments. Companies that continue adding more plans, including additional products and/or price points, tend to demonstrate higher growth:

The takeaway is that growth is influenced not only by having a diverse set of plans early on, but also by continuously introducing new products and pricing options over time. So interesting how fast-growing companies tend to add significantly more plans (more than 10) than stagnant or declining companies.

4. More plans impact the speed of growth

It’s also important to analyze growth in the context of tenure and speed.

For example, Company A adds $1M in ARR over 60 months (which is about $16.7K per month). Company B adds $300K over 6 months (or $50K per month). Company B is growing faster, so it should rank higher on growth velocity, even though its total ARR gain is smaller.

That’s why this analysis focuses on growth velocity, not just overall growth. The goal was to understand whether offering more plans helps companies not only reach higher ARR, but also reach it faster.

To do this, I defined 5 personas (segments) based on: (a) ARR growth, (b) tenure, and (c) ARR growth speed:

• arr_growth_total = Max ARR – Min ARR

• tenure_months = (Max Date – Min Date) / 30 days

• arr_growth_per_month = arr_growth_total / tenure_months

The last metric is the primary clustering factor. It reflects how quickly ARR is increasing, while still accounting for the magnitude of growth.

Because the sample is heavily skewed with many outliers, I did not use k-means for growth velocity analysis. I wanted to compare growth speed vs number of plans across equal-sized groups, so I used percentile segmentation to create 5 clusters. I named them using a fast-cars theme (rumor has it the ChartMogul team likes Formula 1):

As you can see, the same pattern holds in this segmentation as well: companies with more plans don’t just grow more, they grow faster.

5. Other interesting

In addition to the plan and growth relationship, I also learned that:

• Growth is highly non-linear. Most companies cluster in slow-to-moderate growth, while breakout growth is rare. “Stagnant” is the most common failure mode.

• The strongest breakouts show both signals. ARR and subscriber growth reinforce each other. Seeing strength in only one is less convincing. The healthiest companies tend to be strong on both dimensions.

• Early breakouts are an exception. In the earliest breakout phase, companies appear largely “immune” to plan count. Plan strategy doesn’t show a clear relationship with either revenue or customer growth, and these orgs behave like outliers.

• High plan counts can also indicate price testing. And while more plans can reflect experimentation and iteration, this analysis is based on active plans only (not historical or retired variants).

Final thoughts

I’ve been looking forward to this analysis for a long time. Thank you to the ChartMogul team for supporting the work and sharing their data. Their Insights page is a goldmine of data-backed SaaS research.

Plan expansion alone doesn’t create growth. But across this dataset, companies with broader plan catalogs tend to show stronger and faster growth. The healthiest businesses add subscribers without lowering revenue per customer. And good plan structure often makes this possible.

Plan mix, pricing structure, subscriber growth, and ARR are tightly linked - companies that grow faster tend to offer more plans that fit different customer types and needs, supporting better segmentation and more effective monetization.

Most subscription platforms misreport churn. Every tool provides a generic churn metric meant to work for SaaS, enterprises, big and small B2B, products with subscriptions, ads, and one-off transactions. And that’s exactly why it breaks.

Here’s the churn report from RevenueCat, for example:

And here’s the real churn rate for the same app:

Same story with Chargebee, ChartMogul, Recurly, and Baremetrics (I’ve had the pleasure of fixing reporting after all of them). It’s not that these tools misread your data or can’t segment it correctly. They’re built to deliver a one-size-fits-all churn number (unless you have an exclusive plan with them, allowing customizations).

You have to make a call whether to trust that plug-and-play metric and bet your growth and budget on it, or build something accurate in-house - something designed for your product, your subscription plans, your grace periods, and your trials.

Below, I’ll recap how to approach churn reporting, how to work with churn vendor data, and how to know your churn data is accurate.

Making sense of vendor data

I want to be clear: I’m not saying you shouldn’t use tools like Chargebee (though this one I personally wouldn’t), Baremetrics, or ChartMogul. Most of them are solid and genuinely helpful. I rely on RevenueCat data every day, and their scheduled exports are often the main source of my reporting. I use ChartMogul in my data audit a lot, and I also like Baremetrics subscription reports.

The issue is that data reporting is not the primary feature of these tools. Their dashboards are add-ons, not analytics platforms. We should treat them that way, instead of opening support tickets every time a number looks off in a freemium UI.

Depending on your product, you may not want to ingest Apple Pay, Google Play, Stripe, and PayPal data yourself or run servers to capture and store every event. That’s expensive to store and maintain. But at the same time, proper analytics requires more than aggregated totals with no user_id - this is exactly what payment stores deliver via API by default.

This is where subscription tools help. They process payment-store data for you, consolidate it, map it to a user, segment it, and send it back in a clean, consumable format. They do the heavy lift, and it’s great that we can take advantage of it.

What you should do is own the metric definitions inside your database, built on top of the raw ingested data from these tools. What you build might not match their dashboards, and that’s fine. But it lets you narrow TTP (Trial-to-Paid) to your actual trial length (which they can’t do), segment subscriptions across all 20 plans you offer (which they do poorly, especially if you use Stripe), and exclude lifetime plans or free offers from ending contracts, so your retention is accurate.

Building churn reporting the right way

The hardest part of building an accurate churn metric is deciding what it’s for. Making it timely and actionable is one thing. Using it for benchmarking is completely different.

You will need to develop 2-3 variations of the same KPI:

1. Churn for the board and executives for benchmarking.

2. Churn for leadership to develop a strategy and run forecasts.

3. Churn for product and marketing teams to act on and measure their initiatives against.

The logic, formulas, and calculations behind these 3 are not the same. Vendor dashboards usually give you something close to #1. That’s what they use for their monthly and quarterly benchmark reports.

But what you actually need is #2 and #3 - and you need them to be as close to each other as possible.

💡I often notice how people share their preference for the “right” KPI calculation and debate which one is more accurate or appropriate - for example, bounded or unbounded retention, excluding new users from DAU or including them, excluding trials from churn rate, or including them. The truth is that we need all of these KPI variations for different use cases.

1. Churn for benchmarking:

Metric: A simple consolidated blended churn, merging all plans and reflecting your overall loss this month, one for revenue and one for customers.

Report frequency: Monthly and quarterly.

This is exactly what subscription tools do.

For example, here is RevenueCat’s churn formula:

([Churned Actives] / [Actives]) × 100 = Churn Rate

The problem is that not all “Actives” are actually “Ending.” RevenueCat notes that Ending customers differ from Beginning customers, but they don’t remove from the Ending group the customers who aren’t up for renewal this month. That means the denominator includes subscriptions that won’t expire and can’t expire, which artificially lowers the churn rate.

That’s one issue. Another is plan migration. If a customer switches plans, many tools count the original plan as “churned,” even though the user is still paying under a different plan. This inflates the churn, and so on.

What you should do instead is narrow “active customers” to only those who are actually up for renewal this month. Exclude new subscribers (they can’t renew) and include only monthly plans plus the small slice of 3-month, 6-month, and annual plans that expire this month. That group becomes your denominator. This is the only way your blended churn rate will be accurate.

2. Churn for strategy and forecasts:

Read more

Instead, the author, who happens to be ChartMogul’s Director of Sales, argued the opposite: don’t include discounts in MRR. It made me raise an eyebrow at first, because one of the first things you learn in analytics is to always include discounts (and refunds) in LTV, in revenue reporting (which, please don’t tell your CFO, is often just SUM(usd_values_received) in your dashboard), basically, in everything. Otherwise, you’ll over-report.

But analytics isn’t finance (though I wish it were). There are 20 ways to define revenue and even more ways to calculate ARR. And the author is right - we shouldn’t include discounts in MRR.

MRR is NOT the total sales collected in a month. That’s where most confusion starts, and one of the reasons semantic layers rarely work for everyone. The revenue metric you use to measure growth isn’t the same as the one finance uses to close the books.

To make things worse, many growth teams are still (badly) stuck measuring marketing and product initiatives against MRR, LTV, or ARPU. Apparently, measuring A/B tests by something simple, like net new paid customers or transactions, isn’t trendy anymore. It has to be a finance-driven, overly complex, cumulative number with at least 4-6 moving parts. Then, you know… it feels serious. 🤦‍♀️🫠

So today, let’s talk about:

• Whether discounts should be included in MRR.

• When they shouldn’t and why.

• How to define revenue for your business.

• Why using consolidated financial metrics (MRR, ARPU, LTV) for growth experiments is misleading.

• And most importantly, how to make revenue metrics actually work for analytics.

Life is easier when you stay away from revenue reporting

If you think about it, analysts and data scientists work with transactions, subscriptions, customers, purchases, and funnel conversions. Very rarely with financial forecasts. Maybe the reason is that they tend to overcomplicate things, running regressions and building models, when all executives really want is a moving average or a historical growth rate. What I learned - executives don’t trust what they can’t replicate in their spreadsheet, especially when it comes to revenue reporting.

So, to begin with, we’re not the owners of revenue reporting. We don’t own billings, MRR, EBITDA, debt-to-sales ratio, or bookings. That’s accounting and finance territory, based on data derived directly from bank statements and payment processors. We don’t report revenue from Google Analytics, and we definitely don’t pull the cash balance from a BigQuery or Snowflake table with 14 upstream models. But we do report everything else, and somehow, we’re expected to make it align perfectly.

Read more

So I’m locking it all down here - grace periods and billing retry timelines, in one place, that I can find fast.

This is for anyone working with subscription data and responsible for growth reporting, retention, churn, renewals, or other subscription metrics.

Grace periods may seem like a small, optional detail. But if they’re ignored or set up incorrectly, your metrics get noisy, new and active subscribers won’t match your starting totals, and your reporting gets messy.

Today, I’ll cover:

• What grace periods are

• Why they exist and why they’re important

• How they impact subscription reporting

• What happens if they’re set too short or too long

• Best practices for choosing the right time window to give customers enough time to come back, while still reporting churn and retention accurately.

What grace periods are and why they matter

A grace period is the extra time a customer gets to fix a failed payment before their subscription is canceled. It’s like a buffer zone. For example, let’s say a monthly subscription ends on March 31. The renewal fails because the customer has insufficient funds. Instead of canceling the subscription immediately, the platform gives the customer X days (the grace period) to update payment info or retry the charge.

Grace periods are important because they directly affect how you count renewals, churn, and re-subscribers:

Subscription start → subscription end → grace period → retry attempts → canceled vs. renewed.

Here’s how RevenueCat explains grace periods:

Grace periods exist because payment failures are often temporary. People update cards, transfer funds, or retry. Without grace, you’d cancel paying customers too quickly.

Grace Periods vs. Billing Retries

A lot of analysts don’t know about the difference between billing retries or confuse them.

• Grace period = how long the subscription remains active after payment fails.

• Billing retries = how many times (and how often) the system tries to charge the user again.

The yellow section in the image above represents the grace period. In most cases, customers can still access all features during this time. The red section represents the billing retry period. Here, the account is on hold - customers can’t use the product, but if payment is recovered, they return to their original subscription. If the billing retry period expires and payment still fails, customers must start a new subscription when they return.

How grace periods affect subscription growth reporting

This is where it gets tricky. Grace periods create those annoying 2–4% discrepancies between Renewals, New subscriptions, and Active subscriber totals.

These gaps happen because not every “expired” subscription should actually be churn. The grace window buys users time to stay:

Read more

Today, I am continuing my deep dive on customer movements - how to measure and report customer upgrades and downgrades.

A few weeks ago, I introduced the concept of re-subscribers in Re-Subscribers: Are Your Customers Coming Back?, where we covered:

• Re-subscribers are customers who previously paid, churned, and later came back to pay again (e.g., they started a new paid subscription).

• They are one of the strongest signals to measure how well your product delivers, whether the pricing is justified, and whether there’s room to increase it.

• Yet, most analytics tools get them wrong: they either misclassify them as new subscriptions, expansion revenue, or worse, grace-period recoveries.

• There are many types of re-subscribers, but only one group truly matters: those who left and returned by choice. This group is the clearest indicator of long-term product value and growth potential.

• I shared my SQL approach to correctly identify and segment true re-subscribers - the kind who churned and came back by choice, not due to billing edge cases or else.

Re-subscribers often get bucketed with upgrades and downgrades - and that’s where most analysts get it wrong. They treat winbacks as either expansion or contraction, when in reality, they’re neither. As a result, 99% of winback reporting out there overestimates how many returning customers the team is actually bringing back.

In your reporting, you need to separate re-subscribers from customers who upgraded or downgraded. Below, I’ll walk you through how to do that - how to make your winback reporting accurate and how to properly segment all types of returners.

A recap: why re-subscribers are important

It's much harder to retain users than it is to convince them to make a purchase.

But bringing users back after they’ve canceled a subscription is even harder.

Once someone has tried your product, there’s not much marketing can do to win them back - if the product didn’t deliver on its promise. Or if it did, but not in the way the user expected. Or maybe it wasn’t worth the price.

So, the % re-subscribers from your total active customers or paid MAU is the clearest indicator of seasonality, value recognition, and product stickiness. They voluntarily paid again [at the same price?], for something they already used.

Read more

Subscribe now

Growth leaders are often on the hunt for that one gold formula or a score to evaluate the success of their business, whether it’s a completely useless (and inaccurate) subscription loop calculator or some magic business equation. They need to know where the product sits, rank the app, and measure growth against competitors.

First off, I don’t believe business or product can be distilled into a single formula or bucket. And even if that were possible for some very simple, single-purpose apps (like calculators, converters, compasses, alarms, etc.), I still don’t know what to do with that “75% - above average” score. Drink champagne?

Instead, I use my own set of signals to assess product maturity and how well it delivers value. One of which is re-subscribers.

We already know it’s much harder to retain users than it is to convince them to make a purchase. But bringing users back after they’ve canceled a subscription is even harder.

Once someone has tried your product, there’s not much marketing can do to win them back - if the product didn’t deliver on its promise. Or if it did, but not in the way the user expected. Or maybe it wasn’t worth the price.

To me, the % of re-subscribers tells a lot: how well your product delivers, whether your pricing is justified, and whether there’s room to increase it, or signs that you’re priced too high.

Today, I am back to subscriptions analytics and reporting, focusing on one of the most underrated and overlooked subscription metrics - re-subscribers. Below, I’ll walk you through how to define them, locate and query, segment, and interpret re-subscriber rates and milestones.

Who are the re-subscribers?

Reactivation, winback, resurrection, restoring purchases - every PM or tool have their own term. But to me, these are all variations of re-subscribers.

Re-subscribers are users who previously had a paid subscription, churned, and later came back to start a new paid subscription:

I don’t consider re-subscribers users who started a trial, never converted, then came back 4 months later with a 50% discount and purchased a direct subscription. I personally mark these as Directs, not Re-subscribers.

However, if you use RevenueCat, these will be attributed to converted trials (backdated to when the trial started). If you use ChartMogul for reporting, they’ll be marked as expansion. If you use Chargebee, these subscribers will get bucketed as either new paid subscriptions or recovered grace-period users.

Read more

Subscribe now

Let’s talk about benchmarks. Why are we so obsessed with them?

I’ve yet to meet a founder, product leader, or marketing exec who isn’t asking me for benchmarks. Everyone wants benchmarks. Some of my clients have even asked to embed market and industry intelligence …

Read more

Subscribe now

To be fair, the Churn Rate and Net New Anything would theoretically be the most difficult KPIs to work with. However, the Trial Success Rate falls into that category as well.

The tricky part is that it seems so simple - you take converted trials and divide them by started trials. That’s it. Yes and no. Analysts often underestimate how much precision is required to define this metric accurately and make it actionable.

In practice, there are different types of trials: one-time eligible trials, infinite freemium, and reverse trials, all of which require customized logic for trial conversion. Also, trial durations can vary - 7 days, 14 days, 30 days, 90 days, or worse - a combination of a few - which makes mapping them into monthly or weekly metrics reporting tricky.

To make matters worse, no existing application provides an accurate Trial Success Rate. I've used Chargebee, Recurly, RevenueCat, ChartMogul, and Baremetrics, and none of them gave me an accurate Trial Success Rate. As Stripe or Apple dashboards. I always had to rebuild it myself.

Below, I’ll walk you through my approach to calculating an accurate Trial-to-Paid Rate and addressing different trial-length conversions in weekly or monthly reports.

Why we can’t have nice things

Or why most SaaS or Subscription apps fail at the TTP Rate

First, some subscription analytics apps do not offer insights into trial usage or trial-to-paid conversion rates because the data they can access comes from the payment stores only after users have verified and billed. As a result, trialers can be “invisible” in some subscriptions or SaaS apps (which was the case for ChartMogul and Chargebee)

This is also a common issue with the payment_success event in Amplitude or Mixpanel when the trial flag is not passed. Even when it is passed, setting up the TTP rate chart via a funnel is often inaccurate.

The biggest and most common challenge with subscription reporting in RevenueCat, Recurly, and similar apps is that they don’t have a concept of a re-subscriber. This means they cannot accurately differentiate between:

1. New subscriptions activated from trials.

2. New subscriptions activated by users who are no longer eligible for a trial.

For example, let’s say a user starts a 7-day trial on April 1st and cancels it on April 4th. The dashboard records a failed conversion on April 4th. Now, imagine the same user returns three months later and subscribes directly to a paid plan on July 1st. Since they’ve already used their trial, they are no longer eligible for another one, so they activate a paid subscription immediately.

Here’s where modern subscription reporting apps go wrong: they incorrectly mark this new subscription as a TTP conversion and tie it to the initial trial start date (April 1st). As a result, on July 1st, the dashboard updates the TTP Rate for April 1st instead of logging a new conversion for July 1st. 🫠

This is a critical issue because the user's return may be the result of a win-back strategy, but there’s no way to measure it.

That’s why most subscription analytics apps inflate TTP rates - they count both successful trial conversions and re-subscribers. This is especially problematic for mature apps, where re-subscribers can account for 20% of new subscriptions, leading to significantly over-reported TTP rates.

This becomes even more complicated when you also start offering different plans or prices. It becomes almost impossible to attribute the correct plan to the right conversion for the appropriate timeline.

How the TTP Rate should be calculated

Read more

This is most relevant for B2B SaaS and B2C Subscription businesses that offer multiple subscription plans:

Preparing data and metrics for B2C subscription reporting is very different from B2B SaaS. Many analysts (and PMs, and CFOs) are unaware of the nuances of subscription reporting and simply borrow methods from SaaS into B2C. 

It sounds simple: if a user cancels Plan A and starts Plan B within x days, define it as an Upgrade (or Downgrade)

Well, not so fast. 

Below, I will share common mistakes, caveats, and nuances of subscription reporting and guide you through different methods for setting up analytics to report upgrade and downgrade subscription KPIs. Hopefully, this will help you avoid my mistakes when setting up such reporting (and then rebuilding it. More than once). Do it right, and do it right early, before too many reports, ML models, or data models are tied to your tables.

Read more

Subscribe now

Today, I want to continue the conversation about analytics for subscription apps and touch base one more time on trials to answer the eternal question: How long should your trial period be - 7 days, 14 days, or 30 days?

There is no clarity or consensus on the optimal trial length among growth experts these days:

• SaaS and B2B researchers suggest no trial or short trial periods, citing the urgency effect and reducing the number of people who free-ride without ever buying.

• Growth experts (particularly Elena Verna or Kyle Poyar), along with the PLG movement, suggest making trials indefinite for everyone with their freemium all the way and reverse trial signature. For some, it seems to work. For others, it doesn’t. Oops.

• Product experts advocate for longer trial periods to give users enough time to explore your features “at their leisure,” become familiar with your app, and finally fall in love with your product.

For me, it comes down to the Trial-to-Paid rate between short and long trials. Let’s dive into what the data actually says.

First of all, we all read and follow the same growth experts, but we often confuse frameworks offered for SaaS with those for classic B2B, B2C transactions, and B2C subscriptions. The SaaS B2B dynamic of freemium doesn’t fit into B2C freemium.

Speaking of SaaS and B2B, I am not a supporter of reverse trials because so much nuance goes into product onboarding, activation, and monetization, and these specifics vary greatly between companies. Some SaaS offers extensive rich freemium, while others put metering or limits on the usage of their features (MTUs, servers, admin users, etc). Every product is unique, and I don’t believe you can borrow monetization frameworks for reverse trials, customer onboarding, or upselling. Additionally, I don’t like how growth experts cherry-pick a particular product (which is quite unique) and use its success story to advocate for the framework, which naturally doesn’t sit well with most other SaaS.

But let’s get back to subscription apps.

What data say

RevenueCat: Longer trials convert better

Read more

Subscribe now

It seems like every app now offers subscriptions. As I continue to receive more questions on reporting subscription metrics, I’ve decided to launch a Subscriptions Series to go over the technicalities and reporting challenges of every subscription lifecycle:

Trial → Subscription → Renewal → Churn → Re-subscription

Today, many subscription plans offer a trial. Thanks to payment providers, setting up plans with different trial durations is easy. Rarely do mobile apps and companies offer plans without some form of trial.

Trials, however, can be tricky when trying to calculate LTV as they may affect the LTV:CAC ratio and result in over-reported subscriptions. Yet, if used wisely, the volume of trials can serve as an early predictor of user churn or as an indicator of revenue shifts during price experiments.

In this publication, I will discuss the different types of trials, cover the top trial metrics, compare the dynamics of freemium vs. reverse trials, share trials and TTP benchmarks, and walk through subscription reporting from common subscription analytics apps, including RevenueCat, Recurly, ChartMogul, etc.

Read more

Subscribe now

Question: I know it's not best practice to measure retention without tying it to a period. But is there a way to measure "overall retention"? In your article on churn, you gave a formula that abstracted away the period, start date, etc. Wondering if there's a similar logic that can be applied to retention measurement for exec-level reporting purposes.

This question, which I received last week, is particularly timely, as (a) this month happens to be a Metrics Month in my newsletter, and (b) it touches on a challenge I have spent a lot of time figuring out: what is the right measurement for churn and retention in a way that not only reflects your overall business growth but also provides accurate and actionable insight for your team. It’s still frustrating to me how often these two pieces are disconnected.

So today, once again, I will share my take on growth metrics, including cohorted vs. non-cohorted retention, the nuances of blended churn, the connection between churn and retention, and effective ways for reporting overall retention.

Below, I will dive into:

1. When and why do you need to report on non-cohorted retention and churn? When cohorts are not sufficient?

2. What retention and churn measurements are venture funds looking for, and how are these KPIs different from the ones you already have in your dashboards?

3. Why we can never ditch massive spreadsheets with business KPIs, and why efforts to automate this often fail.

4. How to report the blended churn rate. Why are many flavors of this KPI misleading?

5. How to report overall non-cohorted retention. What are the different variations of retention KPI?

6. Why I give preference to dynamic and flexible metrics reporting over automation and “locking” metrics in semantic layers.

7. Why subscription growth and revenue reporting are unique and difficult.

This publication is aimed at senior analysts and above with experience in reporting retention and a solid understanding of SaaS and subscription lifecycle concepts, as well as revenue and growth metrics, including the math behind them.

This is also a follow-up to my deep dive into cohorted retention in the : How to measure cohort retention and into churn in the : Everything you need to know about Churn.

If you are involved in growth or subscription reporting, please read these two guides first.

If you are just getting started with cohorted analysis or need a refresher, here is a great introductory guide on what cohorts are, why they matter, and how to use and read them appropriately: The P9 Guide to Cohort Analysis in SaaS, written by one of my favorite bloggers, Christoph Janz. While Christoph’s focus is on SaaS, this guide is also useful for analytics in B2B and B2C.

While cohorts are vital and essential to report, additionally, you will also need to own a non-cohorted view of metrics. So now, let’s talk about use cases and common methods of measuring and reporting non-cohorted and blended retention and churn.

Why and when do you need to report on non-cohorted retention

Read more

Subscribe now

Did you know there are over 1.17 million subscription apps (as of data from September 2023), and their growth is expected to increase by over 13% by 2025?

Subscription apps are currently the fastest-growing business model. Every day, more products and services (and newsletters) offer subscriptions, leading to a surge in demand for monitoring and measuring subscription growth and revenue.

That’s why, over the past few years, we’ve noticed a wave of companies begin to deliberately target the subscription analytics space (e.g., Recurly, Chargebee, RevenueCat, ChartMogul, Qonversion, Paddle, Baremetrics, and many more). All of them bring the same value by offering ​​dashboards and comprehensive subscription reporting into subscription metrics to help businesses optimize plans, pricing, subscriber acquisition, and revenue:

If you’re already taking advantage of such tools, this publication will be less relevant for you. That said, if your company is investing in data-driven decision-making via matured in-house analytics and data infrastructure, this article will help you understand how to work with subscription analytics. I will walk you through ways of properly setting up subscription analytics for reporting and point out common data challenges, as well as ways to solve them.

Read more

During the recession, the first thing you’ll notice is an increase in the churn rate. You may spend many hours cohorting it, segmenting by every attribute, re-calculating, and rematching with every data source you can access. I wish I could save you days of work and assure you that it’s increased for every business, and it’s a known and expected pattern that you’ll have little control over. But it doesn’t mean your marketing initiatives are not also making it even worse.

Since it’s up for everyone, people panic. They recalculate their marketing budgets, blame product owners, and question analysts. I have seen a few publications recently with incorrect or flawed churn definitions that inspired/forced me to write this piece. In this hot mess, let’s align on SaaS basics, on what churn should be for your product or business, how to measure it, and how to report it.

Read more

I received a question from one of my readers about driving and supporting analytics for a company with a freemium model, and the best accompanying KPIs to scale growth and monetization. So today I will talk about the difference between freemium, free trial, and reverse trial, covering common metrics and challenges in analytics to measure and report the impact and growth for each. 

In this publication, I won’t focus on the foundations of go-to-market strategies or break down monetization models (read Kyle Poyar, Elena Verna, and Jason Lemkin to learn more) but rather on the dark side reporting free trials / freemium analytics.

How PLG generation changed analytics

In today’s new PLG era, sales-driven models take a step back, while product freemiums and free trials become the foundation for growth and business strategy. Analytics has had to adapt as well, and as a result -  a new product analytics domain has emerged to serve PLG's mission. It requires a slightly different skill set, mindset, and perspective from analysts compared to “classic” data analytics or business intelligence (which is on a journey into the sunset to retire soon). 

What does this mean for analysts? 

• The speed of insight and report delivery outweighs its precision. More attention is put towards instant data monitoring. You have to learn to be efficient and super fast with SQL or Python to have a map or a distribution ready to show right away.

• Data democratization is here and now at the fingertips of PMs, marketers, and everyone who adds “data-driven” to their CVs. Accept the mercy of product analytics tools like Amplitude, Heap, and Mixpanel and brace for the impact when things go south.

Read more

If you’re not a paid subscriber, here’s what you missed this month:

• Why You Shouldn’t Stop A/B Tests Early - covers the most asked questions on experimentation in analytics: how long should I run an A/B test? Why is it recommended for 2 weeks? How do I stop the test early? Why are slow (phased) rollouts dangerous? And more.

• How To Pick The Right Chart - recommendations and guides on how to appropriately choose the right visualization for your analysis.

• Top Data Trends That Will Shape The Upcoming Years - what are the future data trends across ML and AI, and how do we prepare now to accelerate business growth, data team productivity, and efficiency?   

  Subscribe now

I am in Austin, Texas this week and thought it might be a good time to touch on analytics reporting. Today’s topic is dear to anyone who is on a mission to monitor SaaS metrics and KPIs and report growth to their board, investors, executives, or whoever you happen to owe your soul to. I’ll walk you behind the curtain of reporting to illustrate  how both easy and challenging it can be to create a true story of the business's health and growth. 

Caveat: below I’ll dig down into common SaaS metrics, so you have to be familiar with most SaaS KPIs and their relationships to follow the article. If you are a beginner or getting started with SaaS analytics, please refer first to The SaaS Metrics Cheat Sheet from ChartMogul or this nice roundup of 15 Metrics Every SaaS Company Should Care About from Hubspot.  

As someone heavily involved in SaaS health monitoring and reporting for many years, let me assure you that the Board always asks for these 5 preferred KPIs:

1. ARR/MRR

2. ARPU

3. CAC

4. LTV

5. Churn

Of course, you have to provide data on many additional metrics including Net New Everything, ASP, ACV, expansion revenue, retention, NPS, and much more.

And (obviously) investors expect KPIs in a table view including MoM and YoY stats, and broken down by regions. And, of course, cohorted! Nothing today is reported un-cohorted. We passed that stage back in 2003 when the second movie of Harry Potter was released and search engine marketing (SEM) became a thing. Today it’s expected that retention, LTV, and churn should be cohorted. I believe that less than 20% of companies actually can report cohorts cleanly, but who cares about the details like missing data, wrong data, or low precision and lack of trust? I do. I care. That’s why I created this publication. 

KPIs to be cautious with

All KPIs mentioned above (except ARR and MRR, which are more straightforward) open an ocean of opportunities for analysis to deviate and mislead (and not necessarily on purpose). For example: 

ARPU 

ARPU can be a vanity metric or outright misleading. Contrary to its purpose, it will rise when you are losing customers. It’s just simple math:

Over 6 months, with MRR and the number of customers declining, ARPU shows a 95% increase. Let’s celebrate!

And, like any average, it can betray the true health of the business. If not correctly used, blended ARPU (which is often used when a small segment of your customers accounts for a large part of your revenue) can effortlessly promote you from a fly to a mouse. Bye-bye, dear little fly!

Christoph Janz Five ways to build a $100 million business and Five years later: Five ways to build a $100 million SaaS business

To be fair, following Christoph’s logic, no amount of ARPU calculation error will get you from a rabbit to a deer or a higher bucket on a quest to reach 100M ARR. So his ARPU segmentation makes complete sense. But for the “consumer-focused” business reporting (with ARPU $100 or less), it’s possible to screw up and fool that chart.

Analytics can get messy, but if any KPI becomes green when you lose a customer or revenue, that’s not a good KPI to follow :) At some point I will dedicate a publication to ARPU - why it’s tricky, and in many cases quite frankly useless in monitoring growth. 

CAC and LTV

Why these SaaS KPIs are challenging: 

1. CAC is a dark world, full of mystery.

2. LTV segmentation per marketing channel often isn’t trustworthy. 

3. Gross profit data (promos, discounts, refunds) are not easy to get. 

4. Any LTV has to be maintained and tweaked.  

Here is the thing about CAC: no one knows what their true CAC is today. Given the trend toward enhancing user privacy, it’s very tricky to calculate how many users you acquired via paid vs organic channels (and how they overlap). Tracking multi-touch attribution is getting more challenging, and most customers are taking longer and multiple channels to consider a brand before they buy. You may get a cleanish view into your blended CAC (cost on upsell, expansion, and renewal efforts) and your sales, but estimating precise marketing costs for acquiring new users is very difficult today. 

The main reason investors watch LTV is to track how much you can spend on marketing while still remaining profitable. That's pretty much it - to use as a benchmark to compare and contrast different marketing channels. Thus, if you don't segment it per marketing channel / campaign (and I have never seen it cleanly done, but prove me wrong), your lifetime value for a customer is useless. 

I think I went through at least 5 different LTV calculations and approaches throughout my analytics journey (contractual, not contractual, including free trials, excluding trials, inverting retention, the Patrick’s way, etc). And none, I believe, represented a true user value. Given that CAC is a dark world filled with the unknown, I have seen many companies exclude it from their LTV and report it simply as user revenue multiplied by user lifetime. But that makes it a different metric that does not represent the actual value of a user anymore. I’m not even quite sure if it represents anything. Many SaaS businesses have multiple subscription tiers, and plans, so LTV becomes more of a passive metric that does not lead to any actual insight that you can act on. 

And now imagine that (1) you have a very approximate CAC that can deviate 4x times either way (a true story), and then (2) a very questionable LTV that is (a) not sensitive to upgrades and downgrades, or worse - (b) excluding gross profit via promos, and refunds (and that % can be quite high), and now you have to try to explain to your dear investors why your LTV is not 3x higher than CAC. It becomes dangerous and flawed as executives tend to abuse LTV by using it as a baseline for revenue forecasting and then for planning budget spend based on its forecasted revenue.   

Good luck estimating your Month To Recover CAC with all of that.

Churn 

Why churn is challenging:

1. Churn is not a cancellation.

2. Involuntary churn data often are not easy to get.

3. Blended churn is misleading.

4. New and not up-for-renewal users have to be excluded from churn calculation.

I’ve written a lot about churn already. Monitoring it and cleanly reporting it is crucial for SaaS (unlike B2C where it can be less damaging). Surprisingly, many executives and VCs don’t quite get the concept of churn. They monitor it wrongfully by looking either at blended churn, or pulling new customers into the denominator, or equaling churn with cancellations. Read more - As The World Churns. 

Overlooked KPIs

I should have called this section What the Board Should Be Looking At Instead, but frankly, they look at everything with a heavy focus on financials (and you can’t fool with balance and income statements like you’re able to with growth metrics). And yet some KPIs, especially the ones related to user behavior, are underrated and often overlooked, such as: DAU/MAU ratio, expansion customers, trial conversion rate, Day 7 retention, network score or referral score. To highlight a few:

DAU/MAU Ratio

Just because your customers pay you for the product, it doesn't necessarily mean they actually are using it in a way that you’d expect. 

DAU/MAU Ratio shows what % of monthly customers use your product every single day. 

You can use WAU/MAU or DAU/WAU ratio depending on the expected frequency of usage for your product. If you are not sure how to get that, read - User Engagement and Activity Histogram Analysis. Some SaaS products or services are expected to be used daily. Others are  weekly or even monthly. So make sure you run that analysis first thing to know how to approach your metrics. 

I haven’t met a business yet that would have strong financials with low customer usage. In my experience, frequent and strong engagement precedes and is an early indicator of future MRR growth (via re-subscriptions, resurrections or expansion revenue). A low DAU/MAU ratio is a predictor of Churn. This is the reason why at MyFitnessPal we pay a lot of attention to our product usage: the more often users log foods using our app, the more likely we will retain them and help them reach their goals.      

Another reason why I like DAU/MAU ratio is that you can't fool with it or deviate. It's simple and yet very precise to reflect your actual engagement (unlike MAU which may include visitors, or retention that can be based on the app_open or received_notification events and not real user activity, or churn that might contain users-who-can't-churn in a denominator, making your rate look small or negative). 

Expansion customers

I mean customers, not revenue here, as we usually break MRR on segments, but not customers. In a given month, for example, we segment recurring revenue on reactivation, expansion, new, churn, and contraction:

Revenue breakdown is important, but as you know, customers are not transactions. One customer doesn’t mean one payment. Even though expansion revenue is often  small compared to total MRR and its segments (I think the benchmark is 30%), but % expansion customers can be quite large. One of the reasons for such discrepancy is that often businesses offer a discount to encourage users to upgrade. But the value of an upgraded user is different. 

A few years ago I was in a meeting watching the Board argue about if switching from a monthly to an annual plan with a 50% discount should be considered an Upgrade or Downgrade. From one side, we were getting significantly lower annualized MRR from those users, so it should be regarded as a Downgrade. From another side, getting these users into annual subscriptions increased LTV, and given annual users are less likely to churn than monthly, it should be considered an Upgrade. 

If your Northstar is LTV, you should treat these as expansion (upgraded) customers, even though they won’t be fully reflected in expansion revenue. If your goal is to maximize profit short term, you probably should treat them as contracted users.    

Referrals and referees 

This is relevant for products where users can bring in more (new or potential) customers by sharing, referring, affiliating, or else. I remember how we scaled Change.org from 50M to 250M active users by doing User Segmentation and Power User Analysis and monitoring the influence score (a very simple proxy to measure network effects) that represented the proportion of referred (we called them recruited) users for each Share action event.

Power user’s stats often are invisible for investors and are not requested or reported. In analytics, these “whales” mimic and reflect DAU/MAU rate and speak of your pool of opportunity for revenue expansion.

“Since Power Users are generally a small share of the user base, companies frequently overlook them. Unfortunately, this results in missing high-ROI opportunities to achieve revenue wins or unlock larger product improvements.” - from Reforge The Power User Trap.

Disclosure: after working with too many 360 customer view or data attribution and "campaign lifecycle data" platforms over the years, I gradually lost my trust in marketing growth (no offense to my fellow growth hackers). I have seen companies burning too much too fast to get too little traction, and CAC and LTVs data often didn't reach even a low acceptable bar to trust or act on it. From an analytics perspective, acquiring "external" data (generated or passed by ads platforms, social media, plugins) has always been and will continue to be a challenge which is likely to get worse with time. 

Product analytics, on another hand, mostly utilizes "internal" user activity data that you can access, validate, cross check, and scale. And it's inspiring to watch companies grow by heavily developing their product and business strategy based on user activity and behavior analytics rather than optimizing paid marketing with last-click attribution.

Using the same data, analysts have the power to draw either a compelling story or make it look like a sinking ship. Playing with a user definition, activity, averages, and rates, we can tailor LTV to look higher than CAC, Churn to be negative, or ARPU to be high. Investing in analytics, due diligence, and tooling should secure clean reporting with high precision and trust.   

Thanks for reading, everyone. Until next Wednesday!

I used to do this to make my stakeholders happy. Because of this:

Read more

One of the great things about being a blogger is that you are constantly being invited to fancy parties and fireside chats, all with the opportunity to connect with experts ac…

Read more