I have a straightforward goal: Create tracking setups where you never have to ask again if this drop is because of a tracking error or real business impact.
We love event data. We love to track products, subscriptions, purchases, content, processes, flows, and ads. But not simply users - come on, we can do better.
We love correct data, always present in the right and defined shapes.
We love to create tracking plans, help with the implementation, and make your data visible.
We love to be tool agnostic: Adobe Analytics, Amplitude, Google Analytics, Google Tag Manager, Heap, Jitsu, Mixpanel, Rudderstack, Segment, Snowplow Analytics - to name a few of them. And we know them not just by name but from every single doc page.
That is a good question. There is the Amplitude Demo Account: analytics.amplitude.com/demo That's all I know. But maybe it would be a good idea if I could create some demo setups. I have an event data simulator, so I can build any dataset.
@@timodechau thank you for your response. I'm a Data Analyst eager to transition into a Product Analytics and Customer Journey focused Analytics Engineer role. I've noticed your extensive use of dbt in your video, which is impressive. I've already started taking courses on it. Could you please share your recommendations for my next steps? Any advice would be greatly appreciated
It's not that easy to answer. Tooling and data model/prep are two parts of the job. The more essential part is understanding how to help a product team. This includes more knowledge about how product and product teams work. Maybe it is a good idea for a post that I should write.
@@timodechau Really appreciate that you replied also can't wait for that ♥️, BTW I built a simulator to practice the product analytics it simulate a streaming service ( Netflix) with All it's compents ut generate users Data, videos Wich is a part of a series, also i made marketing campaigns budget and social Media interaction to practice CLTV and CAC and of course users traffic Data on the platform to monitor users behavioral patterns I don't know why RU-vid doesn't allow me to post the GitHub repository link here Is there any way to send the link for you!?
Excellent video! if you are able to train the product team to use these tools properly and on their own that is amazing. usually they want the data team to build them the whole thing from A-Z.
And this never can scale. And truth be told, they are the experts in their domain. They have so much background knowledge and will unlock much better insights.
Just want to say thank you ❤ BTW you deliver a high value content, So don't get bother by The number of subscribers or watch count, wish All the Best ❤❤
Unfortunately, not really. This is a good post: substack.com/home/post/p-143579364 and this one too: www.leverslabs.com/article/introducing-metric-trees But I am planning to run a weekly live format where I create metric trees - so keep your eyes out for it. Potentially I will start in August
wow!! you just made a great job thatnk you for your amazing efforts i was trying to find any information about how to make comprehensive dashboard by using mixpanel reflecting the ad spend and you really made easy to understand how to do it so i tried to implement it but found new issue which is the UTM source gives me not set on mixpanel any trying to see solution for this issue thank you again for your efforts and truly appreciate it
I basically did everything as a first project. This was helpful to also get an idea of how these different projects work, which ones are more risky, and which I like to do. I will keep this experimentation until today.
@@timodechau I already do some consulting, but I have trouble finding the types of clients that will generate enough value through analytics for them to pay what I'd like. Currently I offer building a simple but complete data infrastructure ELT with data pipelines, a data warehouse, dbt and any BI tool. What would you recommend me to reach more companies that could benefit from that? I'm in Mexico by the way, but I could sell remotely in the US or Europe if need be.
Hey Timo great video! Had a doubt or wanted a confirmation actually for User Properties Let's say I don't connect to DW Question : For a product, let's say I want to get amount spend in SF in last 1 year by users. There is a User123 who spent some time in SF (6 months) and in California in the next 6 months. Now, the location variable will be updated. Ideally if I take 1 year time duration, this will lead to all amount being allocated to California. Or if i calculated number of user X city, then this User123 should be counted in both but as per Mixpanel's implementation this would be counted in California right? (Want to confirm on this understanding :) ) I know I can pass location as event property with amount as a solution. The reason I said no DW. Cause this requires good data at the source or at my end and we have identified some issues with it which will take some sweet time to resolve : (. Also good edits, zooming etc on a couple of elements + great explanation with the diagrams!
When I got your scenario right, I would solve it with events only, like a purchase event with the amount and the state providing the properties. This should get you the right calculation. It could also been done with user properties, but it would be not really efficient to do so.
Thank you very much for your explanation of the difference between the star schema and the event schema. As a BI practitioner, I use the star schema almost all the time, and I wasn't so sure about the second one. Is there a resource you can recommend to learn more about the Event Schema and Event analytics in general?
Good video! Could you please complete the video though? I am able to initiate all the required instances but not able to move forward from there sadly.
I published longer versions here: www.deepskydata.com/courses/snowplow-on-google-cloud-iglu-server www.deepskydata.com/courses/snowplow-on-google-cloud-2-the-pipeline
Concerning tools - it always depends on the project. Usually from a long list of data tools. First clients - Network. I told friends about my new services and asked them to look out for potential clients. By that, I got my first two projects.
@@timodechau Your welcome! I am starting to use Mixpanel for a new project and your videos helped me a lot. I come from the BI world, mostly using Power BI. How would you face a project like the one you just showed using a BI tool instead of a tool like Mixpanel or Amplitude? Does it make sense to use another tool for reporting and visualization? It is not that I don't like Mixpanel, but it lacks in the board side.
Any BI tool would do if you only want to report on the metrics and how they change over time. But if you want to understand the correlation between specific user behavior and subscription revenue you need a tool like Mixpanel.
@@timodechau I am starting to get that feeling! Have you ever experimented with tree-based machine learning models to predict metrics, like MRR, only to find that the most significant connections are in the lower branches of the metric tree? For example, by using Shap values and models like Random Forest or XGBoost to uncover deeper insights.
Great overview. I think a common confusion is how is this different than just using Google Analytics. Why would a team want to add this to their toolset.
@@timodechau I have millions of questions, but in part because I am new to Mixpanel and event-based analysis. Is there a resource you would recommend to learn more about all of this? Thanks for your amazing content Timo. I am already reading "Guide to product Analytics" from the Mixpanel team as well.
Amazing video Timo. Very insightful! I can't wait to see the next one. Whenever you are in Mixpanel, are you doing the subscription analytics and the product analytics on the same platform (MixPanel)? Seeing the SQL code for dbt opened my mind. This is the kind of learning resource analysts should be seeking for, instead of more SQL tutorials. Thanks for your content!
Thank you for sharing, Timo. This is so crisp, clear, and relatable - for anyone who is venturing into the world of data consulting. The explore and refine process is the one which I feel would help one to specialize and creates his own niche over time.
This is extremely helpful, Timo. I'm really looking forward to the next videos in the series because I'm going through a very similar process in my current company. We just implemented Mixpanel for the first time there and working on getting the subscription data in sync with product events. Thanks for sharing this knowledge.
what do you think about using a dashboard tool like Plotly and Dash or Poer BI? and replacing the use of mixpanel by doing all the back and front by myself?
I would not recommend it. When you use something like Mixpanel, you want analyse sequences. Like customer journeys. This is not really possible with these tools. You can create a funnel viz, but you need to slice and dice it to find interesting segments and that is too much work in these tools.
@@timodechau this totally makes sense. What about when you have product data from one platform (such a mixpanel to name an example) and marketing data from another platform (such as HubSpot)… should I make a unified report then or how can I join the data from different areas like in a metric tree? Sorry for all the questions!
I can't wait for the next video in this series. Thanks for sharing your wisdom with us Timo! There is a huge gap between knowledge that helps you take action and what is taught in online courses and I firmly believe your channel helps us (especially me) close that gap.
@@timodechau I am loving the idea of metric trees and all the related topics like this one. How do you make a data product out of these? I come from BI, and I can't imagine how a tree and it's relationships should be presented besides a static report.
Very insightful, thanks abhi. Especially these 4 ways metrics can drop (components, influences, seasonality and dimensional segments) was a new framing for me. Would be interesting to learn how the latter two could be included in the framework. Would this be part of building and testing hypotheses for influences?
Thanks for your comment Niels. You can check out the board that Abhi was using here: miro.com/app/board/uXjVNq48sQI=/ There are some references for dimensions. But I am also thinking about the same thing. On a metric level, you can always define the dimensions, but sometimes you need them potentially across a set of metrics, which is more complicated.
🎯 Key Takeaways for quick navigation: 00:00 📊 *Ownership en verantwoordelijkheid voor Mixpanel* - Het is essentieel om minstens één persoon verantwoordelijk te maken voor het beheer van Mixpanel. - Deze persoon is verantwoordelijk voor het nemen van beslissingen over gegevensinzameling, toegangsbeheer en configuratie van het account. - De Lexicon, ofwel gegevensbeheer, is een cruciale tool voor het beheren van gebeurtenissen, eigenschappen en formules in Mixpanel. 02:59 🔄 *Aanpassing van standaardrapporten in Mixpanel* - Mixpanel biedt geen kant-en-klare standaardrapporten zoals bij Google Analytics. - Het definiëren van rapporten begint altijd met de specifieke bedrijfsfuncties die worden ondersteund en de gewenste bedrijfsresultaten. - Het is belangrijk om te begrijpen welke gebeurtenissen relevant zijn voor het bereiken van de bedrijfsdoelstellingen voordat analyses worden uitgevoerd. 05:16 🚀 *Begrip van gebeurtenisinzameling en tracking in Mixpanel* - Het gebruik van SDK's (Software Development Kits) is essentieel voor het verzenden van gebeurtenissen naar Mixpanel vanuit verschillende bronnen. - De "track" en "identify" functies binnen de SDK's worden gebruikt voor het traceren van gebeurtenissen en het identificeren van gebruikers. - Het begrijpen van gebeurtenisinzameling en tracking is cruciaal voor het effectief gebruik van Mixpanel voor analyse. 07:07 📈 *Gebruik van geconstrueerde gebeurtenissen en cohorten in Mixpanel* - Geconstrueerde gebeurtenissen zijn gebaseerd op bestaande gebeurtenissen en kunnen worden gebruikt om specifieke gebruikersgedragingen te identificeren. - Cohorten zijn groepen gebruikers gebaseerd op gedefinieerde criteria, waardoor diepgaande analyses mogelijk zijn. - Het begrip en gebruik van geconstrueerde gebeurtenissen en cohorten voegen een krachtige dimensie toe aan de analysecapaciteiten van Mixpanel. 08:42 🛑 *Beheer van gegevensgebruikerservaring in Mixpanel* - Het beperken van het aantal gebeurtenissen tot maximaal 50 om de gebruikerservaring van gegevens te behouden. - Een overvloed aan gebeurtenissen kan leiden tot moeilijkheden bij het vinden van relevante gebeurtenissen voor analyse en het handhaven van de gegevenskwaliteit. - Monitoring en onderhoud van gebeurtenissen zijn essentieel om ervoor te zorgen dat de verzamelde gegevens nauwkeurig en consistent blijven. 10:08 🎓 *Training en promotie van Mixpanel binnen het bedrijf* - Het trainen van medewerkers en het promoten van het gebruik van Mixpanel binnen het bedrijf zijn cruciaal voor een succesvolle implementatie. - Het begrip van de behoeften en uitdagingen van verschillende teams helpt bij het creëren van relevante en effectieve gebruiksscenario's voor Mixpanel. - Een cultuur van training en acceptatie bevordert het gebruik van Mixpanel en maximaliseert de waarde voor het bedrijf. 11:18 🔍 *Gebruik van de gebruikerstab in Mixpanel* - De gebruikerstab biedt gedetailleerde inzichten in individuele gebruikersgedragingen en kan worden gebruikt voor foutopsporing en analyse. - Door individuele gebruikersprofielen te bekijken, kunnen micro-patronen worden geïdentificeerd en technische problemen worden opgelost. - Regelmatige controle van de gebruikerstab is essentieel voor het waarborgen van de nauwkeurigheid en consistentie van de gegevensverzameling in Mixpanel. Made with HARPA AI
This is part of the process. I did so as well in the past. But in Mixpanel, you have good ways to manage it - merge or rename events, or just hide them.
A blueprint is a tricky business since a tracking plan always depends on a business. But I did this extensive free course here: www.deepskydata.com/courses/create-a-tracking-plan-for-airbnb
Thank you for sharing your process. I've recently finished a masters in big data management and analytics, with ~10 years of professional experience in project management and emergency response before hand. My struggle is feeling like I don't REALLY have skills that someone would want to pay for as a consultant. I also think I'm lacking the network as I haven't worked professionally in data science before trying to be a consultant... I like the idea of knowing my secret weapon. It aligns well with everything I know about personal branding. Just the "how" of finding out what the secret weapon is is fantastic!
It is always a good option, to freelance for a consultancy first or work full time for 1-2 years. It definitely helps to learn the typical consultancy things (acquiring clients, contracts,..)
Really interesting stuff. Looks like you do pretty much everything data wise for companies? Defining metrics, setting up tracking, bringing everything into a datawarehouse, visualizations etc. So you're really handling everything- data analytics, data engineering, business intelligence etc. Really cool. If you wrote a book or made a course outlining your processes from start to finish I would buy in a heartbeat.
Thanks! I handle most aspects of a data setup. But definitely have some focus topics like event data. I have a book about event data: timodechau.com/book. I am also working on a new course: Event analytics for Growth. And you can find other free courses I created on deepskydata.com.