Matt Clarke @EcommerceMatt

I'm an Ecommerce and Marketing Director with an MSc in Internet Retailing and I write about machine learning and data science at Practical Data Science. practicaldatascience.co.uk UK Joined July 2008

Tweets

3K
Followers

1K
Following

442
Likes

164

Matt Clarke @EcommerceMatt

4 years ago

Setting the correct dtype on categorical variables in Pandas dataframes can make a huge difference to memory usage. Here's how to do it. #datascience #pandas #python practicaldatascience.co.uk/data-science/h…

0 0 4 336 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

The Pandas pipe() function can be used to create data pipelines and works well with the modern Pandas method chaining technique. Here's how to use it. #datascience #python #pandas practicaldatascience.co.uk/data-science/h…

1 0 4 342 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

I initially wrote off method chaining in Pandas as being harder to read and debug but, in retrospect, I think I was wrong. It's actually a really handy technique when used sensibly. Here's how to do it. #datascience #python #pandas practicaldatascience.co.uk/data-science/h…

0 0 0 202 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@LeeFootSEO I did it like this and then created similar ones to categorise the underlying subcategory levels based on product data.

0 0 1 57 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@LeeFootSEO I've done this too, Lee. I used deepset/sentence_bert from Huggingface. Worked well.

0 0 1 36 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

Learn how to use perform custom named entity recognition in Pandas with Spacy by analysing the skills employers are looking for in job advertisements. practicaldatascience.co.uk/data-science/h… #datascience #python #pandas #nlp #spacy

0 0 3 325 0

View Details

Julia MacDonald @julia_m_mac

4 years ago

At McKinsey, we charged $400k+ per presentation. Here’s the simple 7-step framework we used (steal it for free)🧵:

706 4K 21K 4.1M 24K

View Details

Matt Clarke @EcommerceMatt

4 years ago

Working in ecommerce? Not sure how data science might be able to help you? Here's some inspiration... practicaldatascience.co.uk/data-science/1… #datascience #ecommerce #python #machinelearning

0 0 0 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@NewPrediction Many thanks for the mentions, Brian!

1 0 1 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@NetForce1 @mandrillapp @MailchimpStatus You might be experiencing what I have been seeing over the past week or so. GeneralError code -100?

0 0 0 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@mandrillapp For the past week, I've been getting GeneralError -100 failures, causing lots of remedial work. I raised a ticket but didn't get a reply. Are you aware and working on this, please? Thanks.

0 0 0 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@chris_byrne Thanks, Chris! Appreciated.

0 0 1 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

Recently moved to Klaviyo. Great platform but the sales process failed to tell us about the huge increase in costs for "unsuppressed profiles" required to get good results. It's made deliberately difficult to suppress them too. Are there any better alternatives? #klaviyo

2 0 2 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@AnalyticsNinja I did find one useful workaround. I wrote a Python script to run a webhook for the Klaviyo API to suppress users added to a flow (so no need to upload a CSV, which is v. clunky). However, this should be built-in, not a dev task.

1 0 2 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@AnalyticsNinja Yep, agree. Loads of nice features. During the sales process we weren't told about the hidden costs of unsuppressed profiles so our costs have gone up massively. We need these to get good results from abandonment emails. They also seem to make it deliberately hard to suppress.

1 0 0 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

While it's no match for the mighty Advertools (my preferred web scraping tool), if you want to quickly scrape every page in an XML sitemap to Pandas you can do it in a few lines of code in EcommerceTools, and extract links and stuff. #datascience #pythonseo #python #webscraping

1 1 5 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@GiordMarco96 Yes, have this issue with one of my e-commerce sites. A massive amount of clicks and sales come from one big seller. Hugely lucrative and hugely risky at the same time!

1 0 1 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@GiordMarco96 Cheers, Marco. It doesn't yet, but I've done some stuff with Pandas, MLXtend, and (only today) FPGrowth, so can add this in.

1 0 1 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@GiordMarco96 For me, the 80/20 "rule" does roughly apply. 80% of my clicks come from 18.8% of pages.

1 0 1 0 0

View Details

Matt Clarke @EcommerceMatt

4 years ago

@GiordMarco96 Here's another way to check this. This fetches the data from GSC and calculates the cumulative clicks and the cumulative percentage clicks. The top 10 pages by clicks on my site generate 41% of clicks. The top page generates 8% alone.