DEADHAPPY live chat icon

Work experience at DeadHappy

During my time as a work experience student at DeadHappy, I worked in social media learning about and participating in the promotion of the business on a variety of platforms.

I began the first day with an observation of the Wednesday Wake which showed me some of the practices used in the company and set me up for the introduction to DeadHappy which was next. We were given a tour of the office then sat down to go through the core values and foundation of what the company is about. This first day I had less to do than the other days as it was more about getting used to the environment and learning about the message and aims but I enjoyed this as I was eased into the work and could take my time with the tasks set. I began work with Kate in the afternoon, looking at the deathwishes to see if I could find some to post to the socials. Using a personalised company platform, I was able to search through all of the deathwishes for keywords and choose some ready for day 2 in addition to filming and planning some tiktoks.

On day 2, I started off by editing and posting the tiktok made on day one  then carried on with filming the second tiktok that I had planned. This took up a fair amount of time as I had to sort out all the different locations for filming but I had guidance from Kate so there were very few issues. After this, I met up with Nik from the tech team to go through the security on the computer provided and learn how to download new apps that I may need for certain tasks specific to social media work. I was then left to independently work on a blog with a topic of my choosing – new ways to scatter ashes. Even after I had finished at Deadhappy for the day, I was eager to keep working on the blog and editing the tiktok when I got home.

My last day was just as informative and enjoyable as the others, ending my experience at DeadHappy on a high note. I made use of Zoom more frequently as very few people were in the office which, even though I was nervous about, went very well. I talked to Eddie from the growth team about the brand and how it is different from business. This was interesting and even though we ran out of time to look at partnerships that the company could make, I was glad to be able to learn about the thought process behind the scenes. I posted the tiktok that I had edited at home and then talked to Kate about using Canva and Photoshop to create posts for Facebook and Instagram. This was definitely one of the more challenging things that I completed over the three days as I had some issues using the software but after some help from Kate, I managed to successfully complete two posts (the first of which was immediately uploaded). Lastly, I finished my blog on scattering ashes after a round of edits and completed my time at DeadHappy.

Working at DeadHappy was a wonderful experience that I would definitely recommend for other students. I loved every second and I think that even though it is different to most workplaces, the progressive nature and unrestrictive values lend themselves to the fun yet productive atmosphere. I felt that my views aligned with those of the company and this allowed me to cooperate easily with those I worked with and feel like one of the team so that I could really see what it is like to be in a professional workplace. Thanks again so much to the DeadHappy team for letting me work there and making my experience so amazing.

Work experience at DeadHappy

On my first day at Deadhappy, I was immediately welcomed into the workplace by so many nice people and felt right at home. I observed the Wednesday Wake, got warmly introduced to the company and spent the first day learning about Deadhappy itself and the Deadtechy guys I would be working with. I was never ignored or left to figure things out myself, everyone was happy to assist me and give me a quiet, calm place to work and chat. Learning about Deadhappy’s values, website and workers was a great experience, and I let out a huge sigh of relief when learning about the lack of strict rules! As someone who has anxiety, I was very nervous that I wouldn’t fit in, or do something wrong that didn’t fit with the culture there, but Deadhappy made it clear that it was nothing to worry about. I’m bad at holding conversations and keeping eye contact, but I think the kindness shown by everyone there made me much more confident and comfortable.

On my second day, I met with Bal and I got to observe meetings he attended as well as work on the website itself. We went through the architecture of Deadhappy and I got to meet the people he was working on his current mission with. He was patient and took the time to explain everything to me, as I didn’t understand much at all. Before, I was nervous that I wouldn’t be able to complete the tasks he gave me because I don’t learn Javascript at school, but Bal gave me resources to teach myself the basics and taught me a lot himself. Working with him was a pleasure, and I feel like we spoke as equals and not like a teacher to a student. It was nice to be treated the same as everyone else and gave me a healthy experience in the workplace. We worked on a small addition to the website’s account-creating system, but it was important and helped me learn how to do some front and back-line coding. 

Even though my third day was my last, it was just as good as all the others (maybe even the best). I got to join in on the annual Tech Showcase and display the work me and Bal had completed the day before to a large group of people and gained their support and feedback. We then spent time working to get our mini project pushed out. This was interesting viewing how things were put onto the actual website, and got me involved. I spent the rest of the day meeting up with tech members from Deadhappy and learning about their career journeys while asking questions and receiving feedback and tips. I got plenty of free time to spend working on personal projects (like this blog) and was never rushed, stressed or anxious. Working at Deadhappy was fun and challenging, as well as uber chill. 

Overall, I don’t think it could have been any better. I’m very thankful for this opportunity and was never disappointed. I learned key communication, coding, and teamwork skills, and got to have a great experience in a lively work environment. I can’t put into words how good the culture is at Deadhappy, everyone gets along very well and any problems no matter what they are are solved quickly and efficiently. I definitely recommend Deadhappy to anyone interested, no matter how much or how little skills you have in the field of work you want to go into.

Maybe, you are also in a start-up building out your own stack from scratch. I hope you’ll read what we’re doing here at DeadHappy and get some ideas or encouragement.  Maybe when I look back next year I too will be encouraged. Maybe you’d like to join me at DeadHappy? Read this and you’ll see there is so much value you could add. Maybe you are a salesperson or recruiter who will skim read this and come up with a relevant pitch …. Who knows? But I live in hope. 

The Data Community

The data stack at DeadHappy is built upon a community of data people. At the core of the data community are two ‘data dudes’ Kanmi and myself (‘Data Dude’ was not my preferred job title, but it’s stuck).  Our roles are to do anything that needs doing related to data.  Before there were Data Dudes (and for some time after) a good amount of the data stuff at DeadHappy was originally produced by our C-suit. Nick Meredith pretends to be a CTO but he’s really just our reporting analyst and Dominic styles himself as chief operations officer for investors but really he’s our Data Analyst. 

The Data dudes are almost completely remote. This is partly due to covid regulations and partly by choice (we’re both pretty introverted). However after weeks and months of zoom calls I decided to finally determine if Kanmi had legs… this is photographic evidence that he does.

Samantha Grant – Product manager

John Welbourn-Smith – DevSecOps Engineer

Nick Meredith – CTO

Dominic Holton – Data analyst

The problem is that these c-suite data geniuses primarily work in google sheets and geckoboards augmented by Dominic’s ‘innovative sql’ (Kanmi’s term).  Our first priority as a data dudes has been to move away from these sheets towards DeadHappy’s data platform, which I refer to as the Crypt.  Along the way we get support from our Dev/SecOps guy Big Bad John Welbourne-Smith.  

Our data community also includes other DeadHappy people who use our data in their decisions such as our social media sorceress Kate and our product manager (herder of cats) Sam. We call in Big Bad John for help with the DevOps dark arts. 

I think our community works well. One issue however, is that while Kanmi and I are nominally analytics engineers we’ve both got a bit of a bias towards engineering. That’s primarily my fault. We have been perhaps too focused at productionizing our data pipeline, at the expense of really spending the time analysing the end result. I’d hoped that the rest of the folks DeadHappy were ready to consume and learn from data if only we could get it to them more reliably. Realistically, DeadHappy needs more data leadership from us to harvest the insights it can from what we’ve produced so far rather than further developments to the Crypt. Kanmi and I are developing our skills in machine learning and statistics but I think that the next hires in marketing and product missions need to be analysts or data scientists. That said, we’ve laid a fairly solid foundation and, of course, there’s going to be a lot more we want to do with the crypt. 

The Crypt

The Crypt is a tool to answer questions for people and DeadHappy missions. Missions are high level streams of activity such as growth or creating death products to die for. Missions provide vision for lower level streams of activity called ‘challenges.’ The Crypt will be finished when…

  1. It is a part of the fabric of all mission-level decision making. 
  2. It runs by itself, it is secure but available, its documented and it lets us know and recovers when something goes wrong
  3. It automatically proves that it complies with our data policy
  4. DeadHappy people and partners have the right knowledge, reports, tools and support to answer their own questions

When I joined DeadHappy I made the decision to use Google Cloud Platform and especially BigQuery as the core tool for storing and processing data. The DeadHappy tech platform is built on AWS, so for us this meant going multi-cloud but I think it was the right decision.  If you want more info on why we chose GCP over AWS reach out to me you can reach out to me. I’m writing about it further, and I’m happy to give my thoughts.

But the upshot of this decision is that BigQuery is the core of our Crypt.  To the ‘left’ of the Crypt we have data pipelines and ELT and to the ‘right’ of Crypt we have dashboards, spreadsheets, other BigQuery projects and hopefully in the future machine learning pipelines and API’s. 

A High Level view of our Data Warehouse 

Data Ingestion: Overview and ELT

All data lands in a staging project, I’ve called the Sincophagus {read: data-sink / sarcophagus}. We have three basic kinds of data load. 

  1. Snowplow events stream in from user browser sessions. 
  2. Stitch loads data from our various applications including DeadHappy’s own app as well as stripe and typeform. 
  3. Sheets in Google Drive.

If you are wondering why Sheets are a part of our ETL, you know those weird business rules that for some reason are only in someone’s head? I’ve made that people put the rules into a Google Sheet which is then read by the Sincophagus. This is a key move because if someone changes their minds and the rules change, our code doesn’t have to. As much as possible business logic should be parameters read from sheets editable by the business.  

The DeadHappy application is hosted in AWS and uses an Aurora MySQL database. We scan tables for updates and load to BigQuery using Stitch every 30 minutes.  I’d use an ELT tool for ELT every time. Creating manual ETL or ELT processes is time consuming and there isn’t likely to be anything that performing this task in-house adds to your business. Stitch has a good number of free rows a month, so it’s a good choice for a startup like ours. 

The real challenge comes from the database itself. For Stitch to be aware of changes, deletes must be ‘soft’ deletes at least for the period of time it takes for Stitch to next scan the table. By a ‘soft’ delete, I mean that it must be marked for deletion and the updated_at field must be updated. Stitch then ‘sees’ that the row is marked for deletion before it actually disappears.  We’ve had issues with people manually making deletes in the database without that crucial ‘deleted_at’ field and ‘updated_at’ field.  It can be easy for rows to simply vanish or change without Stich passing this new information up to BigQuery. 

Event Tracking: Snowplow vs Google Analytics

Going forward however the database will not be the most important source of information. We’d like to move to an events driven architecture. Our starting point on that journey is web events. Initially we used Google Analytics to monitor behaviour on the website. However, these considerations led us to choose Snowplow as our event tracker. 


  1. The Google Analytics data that we were getting extracted into BigQuery is pre-aggregated. This can make it hard to work with because you cannot join to the data at an atomic level, you must first get everything to the level of aggregation they provide before working with the data. 
  2. We wanted more autonomy about how we would tag up events and what information would be included in which events. Google has some provision for this, however…
  3. Neither Google analytics nor Google tag manager (the container that the universal GA tag sits in) are really version controlled. They have ‘versions’ but this wasn’t working for us.  Everything we do is in Git and for engineers such as ourselves having the tags in Git is more important than a tool that abstracts away from the code base.  
  4. Finally we’d like to move to a system where front and back end events are all tracked using a similar tool. GA only tracks web-events.

We currently have Snowplow tracker sitting in Google Tag Manager along with the Google Analytics tag and quite a few other tags used to track ad performance. This is not an ideal state for a few reasons.  We’ve noticed that quite a few events are simply missed. I think this is due to race conditions on certain events between the Snowplow tracker and other trackers on the website. Basically, we’ve struggled managing Tag Manager. One solution would be to fix up our Tag Manager implementation, however for some of the same reasons we’ve opted to move away from Google Analytics we’ve opted for an in-house solution based on Snowplow and AWS which I’m excited to share this year (once we’ve actually got it working). The basic premise of it is that we will have one DeadHappy tracker embedded in our application and this will inform all the other services of what they need to know. 

The Snowplow Pipeline

The Snowplow tracker sends event data to a ‘collector’ endpoint sitting in Google Cloud Platform. The collector service sits in Cloud Run and passes valid events to a PubSub Topic.  We have Snowplow’s FS2 Enricher subscribing to that topic. The FS2 Enricher is Snowplow’s tool for obfuscating PII and adding geo data based on IP’s etc.  A further Snowplow service ‘the Loader Mutator Repeater’ also sits in a VM. This loads data into our Sincophagus and changes the schema in BigQuery as necessary reloading data that failed due to schema change.  With the painful exception of the two VM’s in this diagram all of the infrastructure is terraform code. I’d like to move to 100% infrastructure as code this year. 


Again, any questions. Just ask. I’m happy to share.

The Warehouse

DBT is a powerful tool for compiling data models as select statements. We use it to transform our raw data into reports and models used for data analysis.  Basically our warehouse is a set of tables and views defined by select statements in dbt.  Roughly speaking data goes through a staging, filtering/cleaning, enhancing, aggregating, joining phases.  DBT handles each of these phases as separate select statements.  DBT uses the command line to interact with BigQuery to compile merge or create table statements as necessary to fill-out the data warehouse. 

The philosophy driving our data modelling is about producing highly redundant highly reusable ‘wide tables’. The Wide tables are something like russian dolls in the sense that they are each based on an object at a different level of composition.

For instance, DeadHappy customers can apply for and buy insurance products: accordingly we have a wide table at an application-product level, a policy_product level, a policy level and a customer level. For web events, we work at the atomic level, at a pageview level, a websession level, a marketing session level, and a customer pathway level. This means that theoretically any analysis can be done at any level without the need to perform aggregations or joins. 

Deploying DBT in Production on GCP

We host dbt as a service in Cloud Run.  Our service has a few features that I think are worth bragging about. The service is revised each time we push to our repo. All the output from DBT are published to our logs and DBT’s errors are reported to the Error Reporting Service which in turn sends me a text message and slacks the data community. The Service reads the headers of whatever https request is sent to it to determine which dbt commands to run. Accordingly I can get it running on a schedule with cloud scheduler and have different models running at different frequencies or even with different  variables.


 Our set-up is fairly similar in effect to this one but we use terraform to create the infrastructure from code instead of using the console. I highly recommend doing things this way because you may find (as we did) that the first region you chose in gcp doesn’t support all the features you’d like.  Porting the whole service to another region or to multiple regions can be a pain to do in the console or the command line because there are so many moving parts. 

I’d like to go into detail about our set-up so I won’t say much on it here. I’m excited by the progress we made in 2021.  2022 is going to be a great year as DeadHappy grows enormously and we adapt and scale our data processes in tandem.

Introducing startup businesses! 

I had the opportunity to have my placement year with DeadHappy, a startup. I was doing part-time work for around Cyber Security accreditations, and here are my thoughts about it.


The fast paced chaotic environment

I started my internship as a Junior Developer looking to do some coding on the platforms but quickly moved on to being a DevOps Engineer and then a Data Engineer (even though I had no clue about data, and honestly still don’t). Hopping around led me to believe that in a startup, the job description doesn’t really matter and I probably wasn’t going to be doing only one thing for the rest of my placement. This was nothing like my life at university where I had weeks upon weeks to complete a project. Releases were going out more often than I was having breakfast, and the small tech team was working like a well-oiled machine, showing the true definition of teamwork, which I couldn’t really see in my academic life, even in group assignments.

How does my business work

Startups are great for another thing – business context. I wasn’t just assigned a task to do, I knew why I was doing it and thought more about the end result and how it solves the problem presented. The metrics, which led to the decisions I was part of, were there for everyone to see and discuss. While it might not be for everyone, I found it insightful to know about the origin of our product, the live data we receive and the future business partnerships. Being involved (even only as a listener) in the non-tech side of things is something I recommend other students do as well while on placement. 



This is probably one of my favorite parts about working in a startup – I can really feel the culture of the business and it’s not something that people just talked about during my onboarding and then never mentioned again. My colleagues are also my friends – yes, all of them – no one is just PersonFromMarketing#1 that I just see every now and then but never talk to. Everyone is a joy to work with because of one simple reason: they give a sh!t, which leads me to the final thing to consider when debating whether to apply for a placement in a startup or not. 

These kinds of businesses are really passionate about their idea, the problem they are trying to solve is the reason they exist. If you don’t care about what they’re trying to do and just want a 9-17 internship, it might not be your place. 

If you take the red pill and choose to see how deep the rabbit hole goes, you might be in for a backstage view of one hell of a show!

I hope this blog has unveiled a new perspective regarding your future placement. Good luck and remember: “A good engineer is always learning!”

Put the lime in the coconut…

The ‘Tech Team’ at DeadHappy consists mainly of men, so maybe it’s my feminist side kicking in, wanting to represent alongside a couple of other women we have on the team. Or perhaps it’s because I raise so many customer bugs from working as a Honcho and feel that it would be cool to solve these myself. Or maybe it’s just my FOMO and that I just want to sit in on their morning stand ups or post more relevant things in the tech channel – not just a picture of me drinking cocktails on my day off (yes I thought that would be a good way to wheedle my way into the team… I was on a bottomless brunch at the time and the strawberry daiquiri told me it would be a great idea…)


Anyhoo, I’ve said women in STEM so many times that some of them decided to help me achieve this – luckily none of them have asked me if I know what it stands for yet *sets reminder to google it later*. 


Dream about (bed)bugs tonight

So this week I’m starting a course that has been recommended by one of the team and because we’re both in the office on Mondays we’re going to complete it together. Another person has taken my incomprehensible notes and ideas about what I want to try and do and is putting them together as a coherent development plan that even I can understand. I’m also going to be helping with debugging and testing with an upcoming feature/release (at least I think that’s what they call it – I’m yet to get my jargon down).


We are the weirdos, mister

Basically working at DeadHappy is super cool, I’ve put myself out there, with a subject I have no idea about and I have been taken seriously – well as seriously as someone can be taken who dresses as a teenager from the 90s. I’m in an environment where growth in new areas is welcomed, where we can interact with each other because we’re not confined to departments and just have bloody lovely people around us who will, at the drop of a hat, tell you that John is talking about Firewalls and not a Cardi-B song…