Issue 5 - The value of a signal is mostly in the information it carries, not its intensity
We never learned Excel at school
It looks like a quantum physic paradox: at scale, ads work. Marketing is one of the most profitable action company can leverage to grow. It’s scaled business.
But when we ask individuals, nobody admits they have been influenced by marketing campaigns...
Like many, I hate ads. On the web, TV, streets: they are everywhere. Still, I'm fascinated by some brands that release well-crafted commercial movies or signs.
Sometimes the ad itself doesn't convert me to buy the product. It rather inspires me to another search on the web.
I think we need better ads. They exist for centuries, but the digital and capitalist era we live in can bring marketing campaigns to another level.
Targeted ads are good at heart but lack intelligence.
You don't need a TV when you just bought one yesterday...
For an upcoming home move, you're inspiring yourself about the last living room trends. Just looking. You don't need targetted sofa recommendations on each website you will visit then.
You would rather prefer designer mood boards and inspiring things.
Instead of getting ads on web pages, you would prefer to receive an email with last trends personalized according to your interests plus some ideas you didn't think of.
The future of ads is targeted ads. But better.
Ads allowing new discovery, not only focusing on your local optimal interest.
Beyond the product.
Even education ads: nobody taught me to vacuum. Ok, it's easy. And we do like our parents do. Still, I'm sure vacuum ads with optimization lessons - tips and tricks - would definitely be more interesting to see rather than a woman cleaning an already clean house.
I already want to pay Google for their service1. Google Maps is essential, if not vital. Gmail is my digital home. I don't talk about THE search engine...
I would love to give money to an AI agent who will bring me curated content, inspirations, new discovery beyond my usual interests, unknown brands, etc...
Not designed to optimize revenue but personal education, culture, and lifestyle.
Behind the buzzword: feature store
ML Ops is trending isn't?
Beyond these managed services - not mature enough to get a clear opinion - the need for tools to manage and monitor machine learning projects is definitively here.
Runs data pipelines that transform raw data into feature values
Stores and manages the feature data itself
Serves feature data consistently for training and inference purposes
Definitively a good resource to remember the next time you see this buzzy word.
While we can go to restaurants again, frustration comes back too (I’m french, so always grumble): restaurant menus are too often mixed up!
Can't find ingredient details, any photos, daily offer, etc... It's a lot of text and not that clear when you think about it.
With structure and composition details, I think I would be less afraid to order something new. These schemes unconsciously help customers to understand the whole place concept. They make each meal easier to comparate with each other while making each one special.
Data Visualization Designer is an underrated position in data teams. It's often a freelance job, contracted for specific projects.
We focus too much on Data Analysts...
I'm sure there is a place for those guys within data teams. Like graphic designers in marketing teams.
About surgical commit
Git is probably the most underrated tool in engineering schools (even for non-developers) but the most used tool in the end.
Proper git hygiene can be a key to deal with legacy and improve team cooperation. Personally, I used to follow naming rules such as:
git commit -m "[feature/branch_name] some message in present tense"
It's very basic. Too often I’m adding several different files into the same commit. Not surgical enough…
I'm wondering what's your commit strategy?
Should SQL queries use trailing or leading commas?
There is a non-ending debate: SQL formatting.
This blog post goes very deep into the question. I would retain its final argument: "it’s people who do the reading".
ETL in GitHub?
After CI/CD in GitHub Action, why not using it for ETL?
This is the concept behind Flat Data: a project that incorporates three tools to help develop data-driven apps within GitHub.
Deep-dive: "git scraping" ;)
The Blog Post
Working on an upcoming article for a side project, I didn’t plan to write a blog post this month.
Here is my best article according to Medium statistics… Got some awesome feedbacks, hope you’ll like it too :)
Microsoft Excel in the era of big data
Beyond The Bracket
I am not a cinema expert. Favorite movie is probably Interstellar by Christopher Nolan and every year I watch again the whole Fast & Furious saga. Blockbusters.
Anyway, watching the trailer for Nomadland, I felt compelled to go see this movie.
Even if it's not Vin Diesel paced, the rhythm and plays make the moment really pleasant, very exotic. Oddly, you live in the present moment when watching this movie.
I don't think I would have the same opinion if I had seen it on TV, while the big screen and wide sounds help a lot for this kind of movie.
As an amateur film photographer, I wasn't disappointed.
Anyway, the storytelling and shots from Chloé Zhao make the whole experience very thoughtful and peaceful.
Coming out of the movie theater, you feel like you have released a big project. Not that this is the movie of the decade but it really makes me feel that create things to make money is not an end itself.
Don't create things to make money; make money so you can create things. The reward for good work is more work.
That’s all for today!
After a year within a football club in the south of France, I’m moving back to Paris for a completely new adventure.
I’ll probably write about my experience as an engineer in the football industry in the coming months. I think it’s original and a quite opaque scene, so definitely interesting to share. Keep connected if you’re interested.
See you ;)
Even with ads, I would pay for Google services…