Learn how to do it yourself: An insider’s look at the femicide database project

Femicide is a global issue that many countries grapple with and Kenya is not an exception. Recognising that the data about local femicide cases was not readily available, Odipo Dev alongside the Africa Data Hub went on a quest to uncover the details behind these cases in their project “Silencing Women”. With the success of the project, we wanted to get under the hood of what it takes to pull off a project like this and provide journalists with a behind-the-scenes look so they can replicate this in their own newsrooms. Sarah Findlay, the Project Lead of the Africa Data Hub, interviewed Felix Kiprono, the Lead of Odipo Dev’s Media Unit and lead of the femicide project for his experiences. 

‍

This interview has been edited and condensed for clarity. 

Sarah: How did this project come about? Where did the initial idea come from? 

Felix: We kept seeing a lot of femicide cases getting reported in the news, every single day. We sat down and thought, “What do we do about this? How big of a problem is this?” When we went to look at the data, we couldn’t find anything available. Everything was outdated or was not comprehensive. Being a data journalism outfit [Odipo Dev], it was natural for us to decide to get the data ourselves. Because we figured that once we lay our hands on the data, we could tell the story about these women better. We could help unravel the patterns and try to quantify the problem beyond the daily news reporting across different media. So this project was borne out of that need to quantify the problem, to shed light on this issue and really understand the crisis in Kenya in a much more informed manner.  

‍

S: Your approach was to extract data from news stories. Why did you choose this specific approach? 

F: We explored other avenues but it was difficult to find the data we were looking for. In Kenya, our government agencies are not as good at sharing data. For example, our police do not release detailed data on crimes. At the end of each year, they post summaries in their annual report. So we will know that there were ± 4000 homicides in a particular year but there is no breakdown or detail about those cases. Getting the kind of data we were looking for through the authorities was difficult so the natural place to look is in news reports. We decided to use that as a starting point. Now that we’ve tested this approach, we’re looking into other options such as exploring public court records, and collaborating with organisations actively working to fight violence against women and girls and are collecting critical data.

‍

S: How did you choose what data to collect from each story? 

F: So we were interested in: (1) the details of the murder such as location and date of the crime, (2) details about the victim and the perpetrator including their demographics and (3) details about the relationship between the perpetrator and the victim. We also tried to find as much data about the justice side of things as possible such as arrests, court dates, charges, verdicts and sentences. How do these cases move through the system? What are the hiccups? How long do these cases take to be resolved? Most news stories focus on the crime itself, but we wanted to go behind the scenes and see what happens in the corridors of justice. Our whole approach was to gather as much data as possible so that we could identify and reveal valuable patterns and insights to understand this particular crisis and ultimately help develop better interventions and conversations. 

‍

S: Your process included data collection, cleaning and verification. Firstly, how much time did this whole process take? And secondly, how did you go about checking that everything was accurate?

F: We started this process back in July 2023 and we completed collecting and cleaning the data in December 2023 so it took us roughly six months. It took time to understand the syntax to use in Advanced Google Search to get the optimal number of cases, hone in on search terms, filter the stories and clean the data. Our first search returned over 18 000 news articles that we had to filter. By reading each story, we excluded those that weren’t relevant (e.g. differentiating between murder and natural death) and managed to reduce the sample to 1 000 articles. From there, we used the UNODC’s 8-point criteria to identify cases of femicide and then extracted key details from each article into a spreadsheet. This is the dataset we used (you can see more detail on our methodology here). 

Our final step was going through the dataset again and using different pairs of eyes to see that the data made sense and that the details captured from each news story were accurate. For each story and victim, we tried to find as much information as possible to not rely only on that one news source. In some cases, searching for a victim’s name brought up other articles or details, such as court reports, that were used to enrich the data. In many cases, though, there wasn’t additional information available. 

S: You published a database with the details of individual cases as well as a data story that explained the data in summary form. Why did you choose to have both these elements in the project and not just one or the other? 

F: The answer is two fold. Firstly, the reason we went with developing a database with the names and faces of victims is that we wanted to humanise them. It hits home differently when you see someone’s face and realise that they were killed because of this terrible issue. It comes with a different feeling unlike just seeing hard statistics. At the same time, and secondly, we wanted to share the insights widely. Not everyone can analyse data or understand the patterns. We did quite a bit of analysis and some of the insights surprised even the team. For example, the fact that 75% of the victims were killed by someone they intimately knew is a horrifying revelation. This is an insight that a reader might not have gotten if not for the data story. We wanted to break it down and make it as easy as possible for people to understand. And from what we’ve seen, people are using the insights when they are debating and talking about this, not the database itself. 

‍

S: What were some of the biggest unforeseen obstacles that you encountered? 

F: One of our biggest challenges was, despite a robust data cleaning and verification process, some duplicates of cases managed to creep into the dataset. We were so concerned with making sure the details of each individual case was correct that we sometimes missed that the same case was being reported on from different angles and by different news outlets. For example, one case may be covered by the Star in the context of the courts while the same case was reported by the Nation with the basic facts of the crime. It’s an excellent reminder to have multiple processes in place to ensure that these kinds of issues don’t creep into your process. 

‍

S: What is the impact that this project has had thus far? Why do you think it gained so much attention? 

F: It’s been huge. First of all, in and of itself, the database has really created a good understanding of the state of femicide in Kenya. We didn’t have this data before and now that we have it, it has really illuminated the landscape. For us to now begin to understand who is killing who, the scale of the problem as well as where these murders are happening, among other things, means that we, as the public, can start to have better conversations and debates. And we’re seeing those shifts already in online conversations. We’ve seen examples of people online using our data to debunk untruths about the victims of femicide. For example, in Kenya we have a lot of victim-blaming with statements like, “she was looking for money” and in essence implying that she deserved what she got. However, our data shows that most of these cases were at the hands of husbands and boyfriends and this new information immediately shifts the sentiment and narrative about why women are being killed. It means that we can have better conversations about this issue. 

We’ve also been amazed at the uptake of the data with international audiences and global newsrooms. We’ve been cited across the world and this has had a huge impact in bringing the problem of femicide, which is a global problem, to the global stage. 

Our data was also used to inform protests that were being organised about the issue of femicide especially in Nairobi. We provided the data to the protest organisers (we even had tshirts made with samples of the data) and it helped to solidify why these protests needed to happen and why they were important. 

We’ve also seen some policy shifts and government programmes implemented around femicide including the police creating a hotline for gender-based violence cases as well as the judiciary establishing special courts to deal with this. We know that direct impact on this is difficult to fully quantify, we know that our work played a role in spotlighting femicide and we can therefore take some credit for these kinds of changes. 

To your second question about the media attention, the timing of our project launch couldn’t have been better. There were several murders in January 2024 that had made national headlines and there were already heated debates in-person and online in Kenya. We managed to publish our findings and share this new information while everyone had questions about the ongoing madness.

‍

S: What are the next steps / future of the project? 

F: We have a few things in the pipeline. Firstly, we want to get more data out of court records. This will be a big undertaking, but it remains a major gap in our dataset and we think it is absolutely necessary to pull back the curtain on what’s happening in the corridors of justice. Secondly, we want to better understand what interventions currently exist and what does the journey for a survivor of GBV look like? We want to partner with agencies and organisations working in this space and see what data could be used to enrich our current dataset using that additional information. Thirdly, we want to spend some more time trying to get more photographs of the faces of the women victims to add to our database. There are currently only a handful available on the site, but we want to put more effort into humanising as many of those victims as possible. Finally, we would like to expand this beyond Kenya. We haven’t seen this in as many other countries. Now that we have a clean methodology and process pipeline as well as seeing the impact this kind of data can have, we would love to apply this elsewhere. 

‍

S: Final question. What advice would you give to someone looking to carry out a project similar to this? 

F: I have so many answers to this: 

• Most importantly, just do it! It can be intimidating to start such an intensive project like this, but remember that you’ve got this! 
• Be sensitive and offer support to your team as they need: Reading through these kinds of gruesome headlines and articles every single day, month after month, and then picking out the gory details such as women being stabbed and hacked to death can take its toll. So you need to know your limits as well as the limits of your team and seek help if you need it. 
• Be diligent: Even though we had a stringent data cleaning process, we still missed out on duplicates. Have a checklist available about the things your team needs to look through when cleaning the data. It helps to make sure that everyone is on the same page. ‍
• Pivot as you need to: It can be a dirty messy process and often it’s not a clear straight path. You have to keep trying the next thing and the next thing, and you’ll get something from there. If something isn’t working, try another process. Be open to finding solutions as you go and try different things until it works out.