Blog /

28 June | 4 min read

In 2020, I became tired of not being able to do a single shopping trip without buying and consuming plastic. So I decided to track our plastic waste for a whole year and see how much plastic we (my partner and I) were actually consuming. Luckily, I found MyPlasticDiary to help me record plastic use everyday. For each item you select a category and then record its weight. The app allows you to track your consumption against a weekly target and also gives scores (a combination of item type and weight). Six months into 2022, I finally managed to analyse my data and write up what I learned.

Recording plastic everyday for an entire year was very time consuming. Even though MyPlasticDiary includes preset categories and gives you approximate weights, I often weighed items on our kitchen scale and put in the exact measure (why wouldn’t you?). Some days I forgot to record items and included them the next day. I also didn’t record while away on holidays. So, gaps in our data didn’t  necessarily mean that we didn’t consume any plastic that day.

Using predefined categories in MyPlasticDiary, I could look at our plastic consumption by types of products. Unsurprisingly, bottles and pots/trays made up the biggest amount in terms of weight, but bags and wrappings accounted for 40% of all items we threw away. And I could even look at subcategories, for example we threw away 251 Tetra paks, 78 yoghurt pots, 60 carrier bags and 39 shampoo bottles.

I didn’t just want to track our plastic consumption but also reduce it. Some things we changed (or at least tried):

• Stop grocery delivery to have more control over our shopping
• Use cotton bags for loose vegetables
• Buy larger package sizes for things like yoghurt, bottled water, rice, pasta
• Use reusable coffee cups again (shops didn’t accept them at the height of the pandemic)

We’re still buying some bottled water but also use a water filter for tap water. We didn’t want to start spending significantly more money, eg. in special shops that don’t use plastic. So in the end our plastic consumption decreased only slightly, even though we were even more aware of our shopping behaviours.

While I tried to record our plastic waste as best as possible, it is still an underestimate of our actual consumption. It is missing things we forgot to track, plastic we consumed when being out and about or plastic we didn’t know was plastic (eg. in clothes). For example, I often forgot to record bin bags themselves (each bin bag weighs around 7g). Our data also excludes items where it was unclear how much plastic they are made of, eg. crisps bags which look like foil but actually contain metallised plastic film. Another thing I didn’t track consistently were menstrual products. But recording parts of it made me finally switch to using a menstrual cup.

It isn’t just hard to consistently track your own plastic consumption, but also to find data to compare yourself against. First of all, there are different ways of looking at it. Most sources either focus on single-use plastic waste while some studies have tried to estimate the total plastic waste per person, including single-use plastic as well as waste from businesses, industry and agriculture. Considering that some of our data is missing, it’s not surprising that our consumption is below both the EU (31kg) and UK (44kg) average for single use plastic waste per person. The latest estimate on total plastic waste claims that the average person in the UK produces 99kg of plastic per year.

Plastic is destroying our nature, wildlife and human health. If we really want to tackle plastic pollution, we need a system change, better data and stricter policies.

I still can’t believe that companies aren’t obligated to publish details about their packaging, which would make it easier to get this kind of data. While you can scan barcodes to get calorie information, the same doesn’t yet exist for plastic because we simply don’t know enough about plastic in packaging (how much, recycled or not, what type, etc.). MyPlasticDiary is currently developing a service to help retailers show information about plastic packaging online.

Consumers alone can’t be responsible for solving the plastic crisis. We need reforms that encourage recycling on a large scale and also acknowledge the role of corporate polluters. More importance should be placed on alternative packaging solutions and ban single use plastic beyond plastic straws. If you want to read more on this topic, check out this article: More recycling won’t solve the plastic pollution – it’s a lie that wasteful consumers cause the problem and that changing our individual habits can fix it.

Most of the analysis is based on my own plastic use data tracked using MyPlasticDiary. The app itself gives you some plastic use stats to see if you are on track against the goals you set. At the moment you can’t export data, but the team behind the app was super helpful in sharing a csv file with me when I got in touch. Data for plastic in period products was sourced from wen and the European Commission.

There was no need for extensive data cleaning and I made only small adjustments. I changed the weight recorded for Tetra Paks to 20% of their full weight, as this article suggests 14% of Tetra Paks are plastic sheets and 6% bioplastic caps. As I only started recording on the 6^th of January 2021, I used data from January 2022 for those first few missing days.

Analysis was done in python as well as most of the visualisations using matplotlib and calplot. All my code can be found on Github.

Blog /

26 July 2021 | 4 min read

As part of my job as a Data Analytics Lead I am helping my company to upskill around all things data. So a few months ago, I started sharing #ThursdayTools with my team each week and I thought these might be useful for others too. So here is a list of open source, non-code toools that I use for data viz, data for good, charity data, maps, infographics and more.

Mapping geographic data often means working with postcodes. Luckily David Kane has built Find That Postcode, a tool that lets you upload a list of postcodes and select what geo data to match, including longitude/latitue, Local Authority, MSOA, and much more. It’s such a timesaver and super easy to use.

Community Lens was developed by the Data Collective – a community for people using data in the social sector. It is another great tool to explore local needs. Upload a list of postcodes, e.g. from beneficiaries, and get an output map with your pins as well as heatmaps based on measures of deprivation you select.

The British Red Cross Covid Vulnerability Index is a fantastic resource to explore local needs. It was created similar to the Index of Multiple Deprivation but includes different dimensions (clinical, health, economic, social, socio-economic). You can download data on Ward, MSOA and Local Authority level or use the interactive map.

If you are interested in socio-economic data on cities in the UK, the Centre for Cities has built an interactive tool that lets you select from a range of indicators (eg business, population, housing, employment) for 63 cities. You can look at a point in time or a time series. Download data or as image.

Parli-N-Grams built by Guiseppe Puntofisso displays the frequency of words in UK parliament debates. You can search for words or phrases in the House of Commons or House of Lords and get a timeseries starting at 1919. Also lets you compare different terms against each other. Super interesting to look at trends over time.

360 Giving is a charity setup to help grant makers in the UK publish their data in a standardised way and make it accessible to everyone. You can use grantnav or 360Insights to explore and download grantmaking data from over 197 funders.

If you are interested in data on charities in the UK, head to Find That Charity built by David Kane. You can search for charities by using their name or charity number, upload a list of charity numbers and match data against it, get links to other official registers, etc.

Another great tool to explore charities in England and Wales is CharityBase created by Dan Kwiatkowski. You can search for charities or directly access the API. Much more user friendly than the Charity Commission register itself.

DataBasic is a suite of easy-to-use web tools that introduce concepts of working with data. The tools let you upload data and explore data in a fun and interesting way. You can use it to count words in a text, find out what’s going on in your csv file, compare differences between two texts or run a network analysis.

Turning your data into a stunning data visualisation doesn’t have to be time consuming. Krisztina Szűcs has developed Plot Parade – an amazing data art project providing a number of layouts ready to use, just pop your numbers in and export as an image.

If you got geographic data to map and need something quick and easy, Dark Horse Analytics have have built Map in Seconds which is literally that. Select a region and copy/paste your data, map done. Export map as pptx or png file. Unfortunately, because it is so easy to use, the number of regions are fairly limited/high level.

Blog /

February 2021 | 4 min read

Well, 2020 was a challenging year, that’s for sure. For many, the pandemic meant struggling with health problems, finances or even loss of loved ones. But there was also (re)discovery. With millions confined to their homes, people had to find different ways to fight ‘lockdown boredom’. Old hobbies were revived and new ones picked up, of which among the top trends has been bread making. In spring 2020, homemade bread started trending on social media as supermarkets recorded yeast and flour shortages, and bread making was even declared a meditative and empowering act.

As a sourdough lover and baker myself, I decided to explore lockdown baking in more detail. I wanted to get an idea of how many more people had been baking with sourdough and whether their engagement had changed during the first lockdown period. So I collected all posts in the Reddit sourdough community for 2020 and found some interesting trends.

In 2020, more than 16,800 Reddit users posted a total of 40,525 submissions on r/Sourdough. Both, the number of users and posts increased during March and peaked in April and May. This trend matches with other countries introducing tighter restrictions during spring to control the spread of Covid-19. Well over 100 countries had imposed either a full or partial lockdown by the end of March.

For accessing and retrieving data from Reddit, I used Praw and Pushshift. The official Pushshift Reddit API documentation and articles by the data science community helped me with this task, in particular Using Pushshift’s API to extract Reddit submissions and How to Scrap Reddit using pushshift.io via Python. These general tips for webscraping were useful to combat issues around server overload.

More documentation and code can be found on my Github.

Blog /

14 December 2020 | 4 min read

I recently left my job at NCVO and my lovely team gave me some great leaving gifts, including a book with 100 maps of London. Feeling inspired, I’ve been playing around with some mapping tools for geospatial data. This blog shows some of my experiments using QGIS, an extensive open source desktop software.

QGIS is free and relatively easy to setup. Once you downloaded the latest version, you are ready to go. The tool allows you to import and work with different kinds of geospatial data. It also comes with Open Street Map – an editable map of the world with underlying geodata for example about landuse or buildings. To access data for a specific area, I installed the QuickOSM plugin and downloaded data for my home town in Germany. I then filtered for different types of land (parks, forest, residential), rivers and buildings. This page lists all the map features that Open Street Map uses to describe and tag geographic data. Once I was happy with my map, I used the layout function to include a title and legend. Below is the final result.

With Open Street Map you can visualise an incredible amount of detail for a specific area. I came across this beautiful visualisation of Kew Gardens in London and thought I might try something similar. For my map I chose the Bergpark Wilhelmshöhe, a landscape park in Kassel, Germany. It is the largest European hillside park, and second largest park on a hill slope in the world. The construction began in 1689 and took about 150 years. Since 2013, it has been a UNESCO World Heritage Site. Although I am fairly happy with my final map, I didn’t even get close to what other GIS visualisation experts are able to produce.

Of course, I wanted to use QGIS to visualise some data. For testing purposes, I used a fairly simple data set to look at population statistics in London over time. I downloaded a shapefile (.shp) for London boroughs here and imported it into the software. Then I got the ONS population estimates and calculated the percentage change for each borough. I imported the data as a csv file and then merged the attributes with the shapefile based on the borough geo code. Finally, I created five categories based on the change values and mapped them using different colours.

The end result reveals some interesting findings. All boroughs have seen a rise in population between 2001 and 2019, except for Kensington and Chelsea where numbers have fallen by 4%. In general, the south west and south east have seen smaller increases than the north and north east.

There is endless data you can visualise on a map. Free GIS data lists a range of free geographic data sources including some country specifics for the UK. For the map below, I used data on broadband connections by Ofcom made available on CDCR. This data is collected on output area (OA) level and you can download the shapefiles for London here. With a bit more time, I could have mapped postcode data against the output areas for a more granular view like in this Financial Times article.

Obviously, there is a lot more you can do with QGIS. You can produce more complex maps, calculate fields, do 3D modelling or automate tasks by using QGIS and python. After a week of playing around with the tool, I think I got a good feel for the software and would definitely consider it for future projects.

If you are interested in creating maps, then I’d recommend taking a look at the #30DayMapChallenge. It is an annual cartographic challenge where participants from all over the world created beautiful maps on the theme of each day. You can find some great examples in the official repository or follow the hashtag on social media. The repository also lists a number of tools and resources to get you started with visualising geospatial data.

Blog /

3 December 2020 | 5 min read

Imagine whispering mummies, laughing shadows and invisible dogs … If you like the idea then the three investigators is for you: An American juvenile detective book series first published as “Alfred Hitchcock and the three investigators”. Three young boys investigating baffling phenomena by application of reason and logic. Created by Robert Arthur Jr in the early 1960s, the books were translated into many languages.

The series became particularly popular in Germany as ‘Die drei ???’ (the three question marks). So popular in fact, that from 1993 the German publisher Kosmos continued the series after the American version stopped. The German detectives even got their own audio drama, video games, graphic novels and theatrical plays.

As a huge fan of the series and data lover, I wanted to explore the audio drama in more detail while learning some new data science skills. Thanks to the fansite Rocky-Beach.com, I was able to collect data on each episode, and this is what I found.

In 1979, the German publisher Europa started a radio drama based on the book series. Since then, they produced at least one episode each year, with two exceptions. When the American titles stopped, the production had to pause for one year. A litigation between the book publisher (Kosmos) and the audio drama publisher (Europa) over naming rights caused the second gap in 2006-07. Today, with over 50 million sold copies, Die drei ??? is the most successful audio drama production worldwide.

Since all three investigators and the narrator are male, it’s not surprising that in most episodes the majority of characters are men. However, even if the main characters are disregarded, women make up about 29% on average of the side characters. There are 11 episodes with no female roles at all, while the episode with the highest proportion of women of all side characters is 75%. The proportion of female characters has varied over time, with no consistent trend line. However, the average went up slightly when the original book series ended in 1993.

The three investigators are all about mystery and strange incidents, which is reflected in their book titles. Paranormal subjects are the most common theme, appearing in 48 titles. This includes words such as ghost, demon, vampire or devil. Haunted places such as castles, villas, mountains or canyons are also very popular. Fewer titles include animals, people or words about mystery and death. With five mentions each, the most common words are Schrecken (horror), Drache (dragon), schwarze (black), Rache (revenge) and Spur (trace/clue).

Most of the data is sourced from the extensive fan website Rocky-Beach.com. It lists all of the episodes with title and descriptions as well as meta data such as author, actors, publication date, duration and more. Using python, I was able to scrape this data from the website and combine all individual tables in one big data set. To determine the gender of actors, I used the Genderize API to predict gender based on their first names. Background information is largely based on information from Wikipedia.

All analysis code and details about the methodology can be found on Github.

Blog /

17 August 2020 | 5 min read

As part of my role at NCVO, I lead on the UK Civil Society Almanac – an annual research publication on the state of the charity sector in the UK. Since its first launch in 1997 as hardcopy only, the Almanac kept evolving and went online in 2012. However, when I started managing the project in 2018, the website and print publication felt outdated and incoherent while the content had almost become a copy and paste exercise. I was tasked with a redesign and this blog summarises what I learned.

In line with greater digitalisation, the Almanac is now mainly published online while the hardcopy has changed from 170 text heavy pages to a 50-page professionally designed summary report. Compared to many other research outputs our starting point wasn’t that bad, but at a closer look some issues emerged.

In 2012, the Almanac website was built by a member of the research team using WordPress. Even though this gave us control over content, it also meant that a lot of the team’s resources were going towards maintaining the website. Plugins like HighCharts and TablePress provided great flexibility for data visualisation but required skills in Json and CSS to create nice looking charts. The print publication, which was produced simultaneously, lacked connection with the digital version in terms of messaging and visual design. The print publication had been outsourced to the same agency for more than 10 years and did not meet data visualisation standards that had advanced in recent years.

It became clear that we had to come up with a more consistent narrative and bridge the gap between the different outputs. The publications should be driven much more by the needs of our audience.

You might think a research report is not a brand. This is probably true for one-off smaller publications, but it is highly effective for a project like the Almanac that consists of multiple outputs and channels. Over the years while the Almanac had changed with the times, it had not adopted a consistent communication strategy. I was able to commission Furthr – a branding and data visualisation agency – to run a three-hour workshop to get clarity on the Almanac’s goals, audience, benefits and tone of voice. With external support we were able to pin down what the main objective was and make hard decisions on our main audience and what they should take away from our research. As an overall goal, the Almanac should work for our core audience of senior stakeholders such as CEOs of charities, in terms of content and the way it is presented. This guided all our later decisions on our outputs and how we communicate.

At the same time, there was an organisation wide project on the way to create a new digital publication platform. This meant we could use the new content management system for our Almanac website and the development was led by the digital team. During this process, we met regularly to review the progress and ensure the tool would meet our needs.

For more specific research content, such as tables and charts, we had to find a suitable data visualisation tool that could be integrated with the website. So I signed up to various data visualisation libraries and compared their offers. We decided on DataWrapper as it’s easy to use, has a clean look and a lot of free features. Although it isn’t as flexible as HighCharts, it reduced the amount of time and skills needed to produce professional looking charts that were on brand. For the print edition, I found a few design agencies focusing on data visualisations and put together a design brief based on our goals. We then decided to go with Data Design Studios as they had a lot of existing data skills, including in-house researchers, and their proposal linked well to our brief.

While working on the technical development, we also reviewed our published content to make it more centred around what our users wanted to know. As a first exercise we created a Knowledge Board based on previous user research. Fortunately, we had a range of user data to draw from and drive our discussions, including stakeholder interviews, surveys with charities and Google Analytics data. This drove our decisions around the structure of the report, the topics we would cover and the prominence of each. Although, a lot was limited by our methodology, we also added more additional data sources to meet our users’ interests. We also added a wider range of content, including definitions we’re using and downloadable data tables.

While focusing on user needs, we also wanted to improve our content based on evidence and good practice in communicating research. Previously, most of the heading and chart titles within the Almanac were purely statistical. But research has shown that good headings and titles are more memorable when they are descriptive. The ONS digital team has done a great job in sharing knowledge of how and why they are changing the way they communicate statistical information. To help users understand the content more quickly, we started to focus on the key messages. We also prioritised them in terms of how important and useful they are for our main target audience – decision makers in charities – to reduce overall noise. We did this across all our channels (website, print and social media).

Our plans to create a more coherent content strategy, to change print designers and develop a new website in one year were very ambitious, but we did it. I was extremely pleased with the overall outcome and we received a lot of positive feedback. Since then, we have continued to improve things, including making use of Google Analytics to inform our content and homepage layout as well as adding spotlight sections on interesting aspects based on stakeholder input.