DEV Community: Mary Jonah

Checkout Flow Optimization Analysis

Mary Jonah — Sun, 09 Aug 2026 11:06:01 +0000

I recently completed 365's Checkout Flow Optimization Analysis with SQL and Tableau Project and decided to share publicly what my analysis is.
Tech Stack: MySQL, Tableau Story.

Client Background
Northstar Metrics
Total checkout attempts/ Monthly checkout success rates
Count Total Carts/ Monthly Abandonment Rate
Error Messages and Devices
Hypothesis for A/B Testing
Recommendations

Client Background

365 DataScience is a Bulgarian based educational platform whose courses are mainly focused on Data Science and AI. The company makes a claim to have more than 3 million students and 132 courses in these 2 areas.
They have 2 pricing plans and a free plan that allows any person who has signed up access to their projects. All activities performed by users on the website are logged and stored in their database for further analysis.

They have been generous enough to share 2 tables: checkout_actions and checkout_carts from their 365_checkout_database. They both capture details such as the user’s id, date of activity, specific activity performed, error message encountered (if any), device the user accessed the website on, and the date they were able to complete purchase of a plan.
This analysis is intended to be shared with the Head of Marketing on checkout details spanning 1st of July 2022 to the 31st of January 2023.

Northstar Metrics

Total carts created
Total checkout attempts
Monthly:
- Checkout success rates
- Cart abandonment rates
Error Messages
- And device types from which the error messages were generated from

Total checkout attempts/ Monthly checkout success rates

Checkout attempt is the final stage, where a user with a selected plan in their cart, initiates the process to pay for the plan.
Total Checkout attempts encompass both successful and unsuccessful payment of a plan after
A successful checkout is one where a user was able to successfully pay for the selected pricing plan.
On a monthly level, 1 out of 3 carts are cleared without any issues
A surprising finding is that November, which had ~5x of January’s total checkout attempts, does not have a corresponding high successful checkout rate.
- From the video, it mentioned there was a campaign (think Black Friday, Pre-Christmas sales, etc.) hence the sudden increase in total checkout attempts can be seen as an outlier and not the norm.

Count Total Carts/ Monthly Abandonment Rate

A cart is abandoned when a user selects their preferred pricing plan but does not get to the stage to provide their payment details but leaves the 365 DataScience website.
The key difference between an abandoned cart and an unsuccessful checkout attempt is that the unsuccessful attempt had a user providing their payment details.
- But due to issues either from their end or 365’s, the payment was not successful.
Back to the analysis, there is an interesting story that has been told of where October 2022 is the month with the highest percentage of abandoned carts.
- Juxtapose this with the data in Checkout Success Attempt Rate.
- In that screen, September 2022 had the lowest success attempt rate.
- Whatever blocker been faced by customers bled into October thereby causing most of them to not finalize their checkout process
- This needs its own independent investigation
The highest bar that corresponds to the total count of carts in November 2022 proves the assertion of a campaign that was held as discussed in the earlier section.

Error Messages and Devices

The error messages are those shown to the user when the checkout process fails.
From this analysis, there are 2 different messages shown based on the device used.
- Desktop: “year field is required”
- Mobile: “number field is required”
80% of the error messages are from users who used Desktop to access the website.
Unfortunately, the data associated with successful attempts does not include device information hence we cannot draw a conclusion about which of the devices we get most purchases from.
Fixing both issues will use the same technique hence it will be cost effective and reasonable to do so at the same time.
- Additional emphasis should be placed on resolving the number field as it directly influences the revenue.
The issues can come from myriads of sources: small input fields on screens, real-time card verification as soon as the user completes filling in the number field in the form.

Hypothesis for A/B Testing

Adding real-time inline validation to the card-number field on mobile /desktop checkout pages will reduce failed attempts caused by missing card details, which in turn will increase the checkout success rates for both devices.

Recommendations

As soon as the user fills any other field outside Card number, and Card number is empty, highlight it.
Make it visible when the user moves to a CVC field and Expiry field is empty.

Does setting a daily reminder actually cause better retention?

Mary Jonah — Wed, 22 Jul 2026 16:52:02 +0000

Building on the post Who are our customers in Mindful Minutes product's journey?, a detail we have on our customers is whether they have set a daily remainder to use the Mindful Minutes application or not.

This is a simple notification with encouraging words on the importance of meditation, a remainder to breathe, and reminding them to not break their streak among other positive words. Below are some examples:

Ever since its introduction, there has not been any analysis to see if the feature has a positive impact on the company's revenue. So, this article will do just that.

Insights

As anyone would have guessed Yes. Within the remainder on group, most of them continued to use the application.

There is a sharp dropoff by the 1st week. A few questions comes into mind.
- Is the feature available for during a 7-day trial period for new users? (As this simulates such a situation)

Who are our customers in Mindful Minutes product's journey?

Mary Jonah — Sat, 27 Jun 2026 20:56:10 +0000

Welcome back to the second post on Amplitude's Mastering Retention Playbook. In this post, I am going to talk about breaking down your users into cohorts and having separate retention processes for them. A lot of times, users are clamped into active and inactive users. But this hurts as users are on a different journey in their active life of a product.

Chapter 3.2 of the Playbook shares 4 different stages anyone interested in understanding who their users are:

New
Current
Resurrected
Dormant

New User:

A new user is one who is using the product for their first time. With such a customer it is important to make their first session very engaging, so they come back as often as possible.

Current User:

A user in this category has been using the product in the previous product usage interval and the current usage interval (Refer to Product Usage Interval for a refresher if this term is new to you.).

In Mindful Minutes:

A current user is one who has completed a meditation session at least once last week, and then another this current week.

Dormant user:

A user here is no longer using the product or service within the span of data been analyzed.

In Mindful Minutes:

Since October 2025 to January 2026, we have more dormant users than any other user based on our records.

Resurrected User:

This is a dormant user who has completed a Critical event recently.

In Mindful Minutes:

After our recent email activation, we got about 10% of dormant users completing a meditation on the application.

Below is an analysis of our users in the various stages:

Another helpful chart to look at is the rate of growth which is referred to as Pulse. It is calculated using: New, Current, and Dormant users. A value greater than 1 means, the product is retaining much more than losing users and vice versa when the value is lower than 1. The week by week calculation is displayed as a line chart below.

.
It is imperative that measures are put in place to halt this loss as there was no week which crossed 0.2% so this is a leaking basket situation with the users.

But was the picture all gloomy from the first week in October 2025 through to the first week in January 2026?

It turns out no, and that there is a glimmer of hope in the more active than dormant users in the final weeks leading to January 2026.

Finding product usage interval and mapping retention type for a meditation app

Mary Jonah — Wed, 17 Jun 2026 13:26:51 +0000

I came across Amplitude's Mastering Retention Playbook and I have been hooked ever since.
More often than not when we want to increase revenue for our digital products, we concentrate on increasing the number of new customers for our products, but the playbook reminds us there are 2 other groups we should include:

Current users becoming power users (so they spend more)
Dormant users (users who are no longer using the product) rather than focus on our current users and those who were once users.

This article starts from basics; it covers details from Chapter 1 up through to Chapter 3.1. I hope this serves as a practical resource for anyone who picks up the material. At the end of the article, my expectations are you would be able to explain:

Calculate the average number of times a typical user uses your application (performs a transaction on an ecommerce store, plays game, completes a meditation session, comments/like/share for a media mobile app or website)?
How to measure retention in multiple ways for your product.

Here I use a fictional mobile application called Mindful Minutes to explain the terms: *Critical Event, Product Usage Interval and talk about the 3 ways to measure retention

Critical Event

Company	Expected interaction with company's product
Youtube	Watch content
Amazon	Buy products
IC Wealth	Buy Ghanaian stocks
Instagram	Post content, like, comment

The interactions above are called critical events. These are the core actions which the various companies expect the bulk of their customers to undertake. Anything outside these is fine but they may not have a direct and positive relationship with revenue for the company.

Critical Event for Mindful Minutes

We want users to complete a meditation session. Hence that is our critical event. The more users complete this, the more money we make.

Product Usage Interval

This relates to how frequent users interact with the application. There are general expectations for apps: daily (social media, weather apps), weekly (e-commerce [could be daily for some users]), monthly (an application for tax or payroll purposes).
But the fact that your product s categorized under social media does not necessarily mean by default your product usage interval is daily. You need to calculate this for yourself.

Product Usage Interval for Mindful Minutes

From our data, most of our users use the app on a *weekly * basis (not excited to hear this but wait this is why we are reading this playbook to move it to daily)

3 ways of measuring retention

It may surprise you to know that there are multiple ways as mentioned in the Playbook.

N-Day (week/month) Retention:
This type calculates the percentage of users who performed the critical event on a specific day. This measure informs you how many of your users returned on say 3 days after performing the critical event.
Unbounded Retention:
The difference between this and the N-Day retention measure is this value informs you of how many users performed the critical event on the specific day and over. So, if you have 30% Unbounded retention on Day 7, then 30% of users meditated on the 8th day and the days after when they completed their first session.

1-Week Retention vs Weekly Unbounded Retention for the Mindful Minutes app

How to read the graph?

N-Week Retention Curve: 8.64% of users came back on exactly 1 week after their first completed meditation session
Unbounded Retention: 31.96% of users came back after their first week and the weeks there after

Bracket Retention: There are some applications that are used infrequently. As an example: a university build a portal where students can view their transcript. There can be varying arguments that it fits the N-day measure but I think it is infrequent. More students will take a look at it during vacation or when school resumes. So why not group these days/weeks/months and then capture your retention within this period. This explains how bracket retention is calculated.

Bracket Retention using 3 week bands for Mindful Minutes

There you have it, my take on Chapter 1 through to Chapter 3.1 of Mastering Retention Playbook by Amplitude

Go through it, share your thoughts, I am excited to read them and let me know where I falter so I can correct.

Resources:
SQL:

Product Usage Interval
3 Retention Types

Should We Even Have a Summer Sale? A Data-Driven Debate 🌞

Mary Jonah — Thu, 11 Jun 2026 07:50:43 +0000

Summer is just around the corner, count down in less than 2 weeks and by now, most companies are already out with good deals on their products. Unfortunately, my fictional company was dealing with other issues, so they are now in action and are hurriedly preparing to slash prices of their products for their customers.

The Marketing Department is excited about this, and the Chief Marketing Officer (CMO) wants the plan completed by end of week. During the first meeting of the team setup to draw the plan, the senior data analyst was asked "Which category of products should we concentrate on based on the sales data your team has collected?".

He answered, "Should we have the summer sale in the first place?". "Because because", muttered the CMO.
"Let me share a simple analysis of revenue grouped by seasons for 2024 and 2023", the analyst said in response to his own question.

"Gross Merchandise Volume is at the highest during the Summer for the past 2 years. Should we not rather boost sales in other seasons?"

"But we will be the only company without a sale, how do you think we will be perceived in the eyes of the public: a company that does only wants money", the CMO interrupted the analyst. "Last quarter financials, showed that a third of our customers purchased more than once within that period."

The story of "liquid gold" from 3 oil fields in Ghana.

Mary Jonah — Thu, 04 Jun 2026 10:02:33 +0000

Still in the 2025 Ghanaian budget, it had a section that highlighted Crude oil production from January up till 25th September 2025. Data from 3 oil fields Jubilee, TEN (Tweneboa Enyenra Ntomme), and SGN (Sankofa-Gye Nyame) captured in Table 8 included:

Daily average production (in barrels)
Number of production days in the month
Quantity for the month (Daily average production x Number of production days)

What I like

Table with all the data needed to come up with visualizations.

What will I change?

Create 2 visualizations:
- Line graphs which show the total quantity (in millions of barrels of crude oil) produced per month from each field.
- Small multiples of line graphs which shows the average production per day for each oilfield.

Before

After
Total output from all 3 oilfields

Average production from each oilfield

Tools: Excel
References:

When things fell into right places

Mary Jonah — Thu, 28 May 2026 06:59:31 +0000

In the 2025 Annual Inflation Report, there was a happy story in 'bars' that needed to be told.

What I like:

Data labels so one can use it in recreating the chart.
Legend was at the top (was eliminated but this is better than most diagrams leaving them in the default position which is beneath the chart)

What will I change?

Axis titles:
- y-axis title: Include "Inflation" title and % to communicate the values are percentages.
- x-axis title: 2025 moved to be an x-axis title as the data spans the same year no need to include it in each month
Highlight the drastic drop in May - June
- Capture that as the story in the title
- Match the background colour of the title with the line between May and June.
- Include the inflation details for these months for both categories
Include the series titles directly on the lines.
Provide details of story as a description beneath the title.
End with the inflation numbers for both series at the end of the year with minimal emphasis on them.

Before:

After:

Tools: Excel
References:
- FIGURE 2.2: GOODS VS SERVICES INFLATION (% YOY), JAN.-DEC. 2025 - Annual Inflation Report (Page 21)
- Exchange Rate Details - Bank of Ghana
- Jan to Dec monthly gold prices - Stat Muse

Give us money and collecting from our citizens, Q1 - Q3 story of financing Ghana's expenditure in 2025

Mary Jonah — Thu, 28 May 2026 06:51:08 +0000

The initial diagram can be found under the section 2025 Q1-Q3 Budget Balances and Financing Operations in the 2025 Budget Statement and Economic Policy.

Here my presentation is about informing the citizens about meeting financing goals from international and domestic sources.

What I like:

The figures in the diagram, there was no table with the figures, so I was able to recreate it using what I see in the diagram.

What will I change?

Both colors will be replaced with Dark Teal, Accent 1 and Orange to eliminate any bias or assumption associated with the existing colors.
The line and figures associated with the totals do not add extra details; they can be added in the text description.
Add block of text to provide more insights on title.
Replace Outturn with Actual, it is too technical for the audience I seek to communicate with.
Use slope graph and its slope to communicate the rate of change.
Replace the default legend with colour-coded text at the top of the graph.
Move the quarter information to the description instead of repeating in x-axis titles.
- Also bolden the text financing rather than repeating it in legend.
Left-align text all text to follow the default way humans read text from top-left to bottom-right.

Before:

After:

Tools: Excel
References:
- Evergreen Data - Make a Slopegraph in Excel
- Figure 6: Summary of Q1-Q3 2025 Financing Operations - 2026 Budget Statement (Page 24)

Introducing "Four-Eyed Insights 🤓"

Mary Jonah — Wed, 27 May 2026 06:42:58 +0000

Hi Everyone,

Welcome to Four-Eyed Insights, a play on the fact that I wear 2 extra 'eyes' in the form of spectacles👓.

This series will be a fun one, here I will be re-doing visuals I find complex so they can be easily understood by most people (my target is not the whole world).

Yes, there will be a before and after section (do not take it personal, everything here is my point of view, and opposing ideas are encouraged so share them in the comments section). I will also share my thought process.

DEV Community: Mary Jonah

Checkout Flow Optimization Analysis

Table of Contents

Client Background

Northstar Metrics

Total checkout attempts/ Monthly checkout success rates

Count Total Carts/ Monthly Abandonment Rate

Error Messages and Devices

Hypothesis for A/B Testing

Recommendations

Does setting a daily reminder actually cause better retention?

Insights

Who are our customers in Mindful Minutes product's journey?

New User:

Current User:

In Mindful Minutes:

Dormant user:

In Mindful Minutes:

Resurrected User:

In Mindful Minutes:

Finding product usage interval and mapping retention type for a meditation app

Critical Event

Critical Event for Mindful Minutes

Product Usage Interval

Product Usage Interval for Mindful Minutes

3 ways of measuring retention

1-Week Retention vs Weekly Unbounded Retention for the Mindful Minutes app

Bracket Retention using 3 week bands for Mindful Minutes

Should We Even Have a Summer Sale? A Data-Driven Debate 🌞

The story of "liquid gold" from 3 oil fields in Ghana.

When things fell into right places

Give us money and collecting from our citizens, Q1 - Q3 story of financing Ghana's expenditure in 2025

Introducing "Four-Eyed Insights 🤓"