Monthly - data

• New lesson editor
• LessonUp
• ISK

Monthly meetup

About data within LU

Peter

Data team

• Daan - CEO

• Peter - Head of Finance

• Ahmed - Business analyst

• Fedor - Data consultant

• Tom - Data engineer

Intro team: Tom

🇫🇷🇳🇱

Tom

Intro team: Fedor

Fedor

When you hear ‘data at LessonUp’, what pops into your mind?

Peter

Do you think we are working data driven at LessonUp?

A

Yes

B

No

Fedor

- modern data landscape

- lessons deduplication

- AI workshops

- Teach activity Index

The Value of Data

Fedor

What is the value of data

What can data bring you

Data literacy → some theory + quiz question?

Insights / KPIs / AB testing / 

Who is using Metabase on a weekly basis?

Data team aims to help you with this

"Without data, you're just another person with an opinion"

- William E. Deming (1900 - 1993) -

Fedor

- We want to make proven decisions that drive value for our users and ourselves.

- Foster a culture of learning, where it is also good to learn what does not work.

- Therefore we need to use our data as a strategic asset

- With AI this is even more important

Data driven organizations are more succesful than their peers

Confidence

People within data driven organizations make decisions with more confidence.

Personalized

Leveraging data can create customized experiences for every individual.

Efficiency

Data helps to identify inefficiencies and optimize processes.

Competitive Advantage

Understanding customers, being able to respond to trends quickly

Fedor

Data Driven Organizations are more succesful than their peers (McKinsey, EY, accenture)

Confidence, more fun, more results. Better discussions.

Personalized: I get lesson A, you get B. I get more images when I ask Maia and you get more text.

Efficiency: Analyze an onboarding flow for example, or internal processes. Make your work easier!

Lead generation

Competitive advantage: Knowing what your customers love

https://newsroom.accenture.com/news/2024/new-accenture-research-finds-that-companies-with-ai-led-processes-outperform-peers

forbes:

https://www.forbes.com/councils/forbesfinancecouncil/2024/09/03/unlocking-success-the-power-of-data-driven-decision-making/

tdwi:

https://tdwi.org/articles/2023/09/20/biz-all-beyond-spreadsheets-data-driven-organizations-outperform.aspx

mcKinsey:

https://www.forbes.com/sites/britneyporter/2023/09/14/how-leveraging-data-can-enhance-your-business-success/

vox media: https://www.forbes.com/sites/britneyporter/2023/09/14/how-leveraging-data-can-enhance-your-business-success/

Data delivers value by

Fedor

This framework works several ways:

- experimentation, look back, hypothesis, predict, validate, decide. Closely together with Product

- Or feature wise:

- learn what lessons are popular

- predict popular lessons

- pro-actively recommend them

- Only works if high quality

- single version of the truth

Value

- Data should create (business) value

Responsible

- Data is a core asset

- Users trust us with their data

- Governance, GDPR, Security

- Ensure Quality

Tools

- A good and reliable toolset to extract, load, transform, analyze and utilize data.

Skills

- Read data

- Write data

- Speak data

- Training

Becoming more data driven requires these pillars to be developed

Fedor

- together, more fun.

- not only data team, but also tech and business teams.

- data team is an enabler and we commit to taking these pillars a level further

Data @LessonUp

Peter

The data landscape at LessonUp

We have multiple data sources which we want to bring from A to B to be able to analyze it. We can distinguish four stages for data tooling:

1. Source data: Raw untouched data from various sources like Google ads / Facebook / Stripe / LessonUp platform etc.

2. Extract data: To pull data from a source using an ELT (Extract, Load, and Transform) tool

3. Store and transform data: Once data is extracted, it’s stored (loaded) into a database or data warehouse. We transform data for easy access and processing.

4. Reporting data: Transformed data is used to create insights

Peter

MongoDB -> DB for devs / coding

Snowplow ->tool where we can track events with

The current data landscape at LessonUp

Peter

Firebase -> tool where all live events are processed and stored (Firestore)

Avo -> event data quality tool to define and implement events+parameters

MongoDB -> DB for devs / coding

Snowplow ->tool where we can track events with

A concrete example: from ads manager to metabase!

Ad is made in facebook and utm code is set

User clicks on facebook ad. On LessonUp site Snowplow registers the event and utm code

The event is stored on BigQuery

dbt add's the correct marketing channel based on utm, user_id, etc...

Create dashboard to see which channels convert best

Fedor

Fedor is working together with thijs and courtney

A concrete example: from lead to MRR/seat calculation!

Signup on platform

Snowplow registers the events.

MongoDB saves user data.

The event and user data is stored on BigQuery

Create lead overview

Leads pushed to Pipedrive via Make, ready for interview process!

Deal won -> deal data pushed to Exact Online via cloud function

Via cloud function to BigQuery and Metabase again in dashboard!

Fedor

Fedor is working together with thijs and courtney

What do you know about our data landscape?

Peter

How many tables does our BigQuery instance have?

A

143

B

1.229

C

12.734

D

162.578

Peter

How many events are generated daily?

A

1.500

B

15.000

C

150.000

D

1.500.000

Peter

55 events per user on average

How often is the data in Metabase refreshed?

A

Metabase is refreshed once a day

B

Metabase is realtime

C

Metabase is refreshed every monday

D

Metabase is refreshed daily at 08:00 and 13:00

Peter

How many metabase users do we have on a working day?

A

13

B

21

C

29

D

37

Peter

On which dashboard can you see how we are performing on our most important KPI's?

A

KPI Dashboard 2025

B

Current MRR Status

C

LessonUp Metrics 2025

D

Performance Dashboard

Peter

What are we working towards to?

A new data platform was implemented by Tasman analytics. Best-of-Breed tooling is used with as much open source as posisble.

2022

We are leveraging our data but struggle with complex models and unclear definitions. We are doing good, but still more than enough room for improvement.

2025

We are working with validated models (lessons, users, events, etc..) that we understand. We use these models instead of raw tables. We are confident in using our data throughout LessonUp.

2026

Fedor

In the upcoming period we make our strategy tangible by defining next steps for each of our pillars

Time

Value

Responsible

Tools

Skills

Marketing insights

User correlation

GDPR

DBT testing

Evaluate AVO

Metabase workshop

...

...

Review datastack

Excel workshop

Examples

Use of AI tooling

Availability resources

Fedor

Data projects

Automation Brevo

✅ Project Summary

We automated the daily synchronization of user data from BigQuery to Brevo, significantly reducing manual work for our CS team.

🔄 How it Works

Source: BigQuery table (Automation User List)

Destination: Brevo (via native API integration)

Sync Frequency: Daily

📈 Impact

⏱️ Saves hours for CS team

🎯 Enables targeted, timely campaigns

🔄 Ensures data is always up to date in Brevo

🗂️ Data Synced (Sample Attributes)

27 user-level and organization-level attributes are synced to power personalized campaigns in Brevo.

ahmed

Refactoring Channel Logic

• Our channel partners have insights in how their lessons are used in the channels dashboards

• The logic for determining what lesson was a black-box and relied on the actions importer.

• We made new logic that...

  • We understand and can explain clearly.

  • Enabled us to turn-off the actions importer, saving 300Eur Monthly.

AI Lesson Generator

• Goal is to generate a whole lesson with AI that is for 80% usable by a teacher and generates SEO value.

• Using the LessonUp library, various python libraries and OpenAI we have made a first version, that is able to:

  • Read learning goals.

  • Generate a lesson including images, video and interactive elements.

• We have made a promising start, next steps: validate with users, investigate integration

Marketing Insights

• Together with Marketing we are working on improving our marketing actvitities

• We checked all our mappings and user_id logic

• Working on setting our UTM-codes correctly

• Goal is to optimize our marketing spend

  • This project gets us in control of our event data and funnel which has lots of additional benefits (e.g. leads, usage, etc..)

• Challenge is to be in control of all this data in order to draw correct conclusions

Cost optimization

Data processing could be quite expensive. For example: costs for running a query on a large table in BigQuery without a correct filter was USD 6 per run (!) Therefore we keep an eye on optimizations where possible.

Some actions we take to mitigate costs:

• Killing old functions running, like the actions importer which is used in the past for event tracking

• Optimization of tables by using partitioning fields --> by using a specific filter it will query more efficiently

• Cleanup/deletion of old not used tables/dashboards → ongoing process

• Looking alternative tooling (for example to replace AVO with Snowplow for event tracking plans, savings of USD 300 pm)

• ....

We are working on automating the "quote-to-cash" process to reduce the workload of the sales team and CS so they can focus more on sales related activities.

• Per 1-1-2025 we started using Exact Online for invoicing our B2B customers. We had a transition period of 3 months and stopped creating invoices via the admin per 1-4-2025

• Next to the manual invoice creation we also created a process for the automated upsells via the school portal

Next steps:

• Merge the old logic (via Admin) and new logic (via Exact) to show seat mutations and MRR calculations in Metabase

• Focus on finetuning the process the make it more error-proof

• Implement cross checks on invoices and seats to make sure errors or missing data is signalled early

Invoicing project

invoice automation process flow

Other ongoing projects

Data security

We need to protect personal data in all of its forms. In case of a data breach with personal data this would harm our company in multiple ways. A lot has been done already by people in the team, but we are not there yet. We still need to make improvements on different levels, like:

• AdminUp / BigQuery / Metabase / Limit personal data on / Old tables in MongoDB / Limitations on access etc.

Data testing

Further improvement on data quality by implement testing so in case data is inconsistent or missing we will get notified.

• # active subscriptions in Stripe vs # Pro subscriptions in database

• # lessons in lessons table vs # lessons shown in Library

.....

Data request / data questions?

Way of working

• Slack your data related questions to Fedor or Peter and we will make sure to create a ticket on our "Data" board in Jira.

• Please note that data related questions could be complex and might require some time to implement.

Exercise

• What data is missing in your daily routine and will help you the most?

  • To generate leads/sales?

  • To analyze results?

  • To work more efficient?

  • To ....?

Submit your data request by the end of next week and we will follow-up with you to see how we can help!

Metabase workshop this afternoon!

Please feel free to join :)

END