The WIN Gender Tracker analyses text only and currently includes four metrics to measure the prominence of women and the inclusion of their voices and opinions.

PROMINENCE OF WOMEN

• 1. % people mentioned who are women
  A simple way to track how prominent women are is to count the sheer number of mentions that they receive in an article. This metric divides the total number of women mentioned by the total number of people mentioned. A ‘mention’ includes any reference to a person (man, woman or unknown). This could be a name, a pronoun (she, he, her, his, him, etc.), a title (Mr, Mrs, Miss, Ms, Sir, Lady, Dame, etc.), or a gendered noun (girlfriend, boyfriend, daughter, son, wife, husband, etc.). The Gender Tracker counts every single mention, not just unique mentions. So if ‘he’ is mentioned 20 times in an article it is counted 20 times.
• 2. % main characters who are women
  You can also look at prominence by looking who the story is about – the main characters or subjects. A story can have more than one main character. This metric uses salience and defines a main character as someone with a salience of more than 0.5. The number of women main characters is divided by the total number of main characters. A main character is only counted once (unlike metric 1).

INCLUSION OF WOMEN’S VOICES AND OPINIONS

• 3. % sources who are women
  Tracking the gender of sources gives a clear idea of the level of inclusion of women’s voices and opinions. A ‘source’ can be someone giving an account of a personal experience or opinion, a witness, a spokesperson or an expert. A ‘source’ includes someone who is quoted either directly or indirectly. This metric counts the number of direct and indirect quotes by women and divides it by the number of direct and indirect quotes by people. This metric counts all sources and not unique sources.
• 4. % authors who are women
  Author (byline) is a simple and powerful metric for voice and opinion. This metric counts the number of women authors and divides it by the total number of authors (where the author is specified). This metric is only included in the media app and not the quick tracker or individual apps.

WIN understands gender is a spectrum, and adopts a gender-inclusive approach in our programmes and activities. While we confirm that gender is non-binary, due to limitations of the technology behind this automated tracking tool, the WIN Gender Tracker looks at the representation of “men” and “women”. We hope to improve the WIN Gender Tracker in the future to be able to capture a wider spectrum of gender identity, especially those related to gender-neutral language.

If you would like to know more about the methodology for each indicator please contact us.

WIN is working on expanding the WIN Gender Tracker to look at the third category of gender balance: Portrayals of women. These metrics will focus on the language used to portray women and will identify language that is stereotyping or overtly sexist.

All of WIN Gender Tracker’s metrics use % scores which represent the % of women for that metric. For example, if the score for “% main characters who are women” is 27%, this means that 27% of the total number of main characters are women. The WIN Gender Tracker gives each article an overall score: Average % women. This overall score is simply the average (mean) of all of the metrics and each metric is given an equal weighting.

Truly gender balanced content means that women and men are represented equally. In numbers, this means 50% women and 50% men. Technically anything under 50% women or over 50% women isn’t perfectly gender balanced. But WIN sees anything over 50% women as a good thing, because for every article that overcompensates on its representation of women there are 3 articles where women are underrepresented. For now, this means that the WIN Gender Tracker’s metrics have a target score of 50+% women.

The WIN Gender Tracker uses a combination of a third party natural language processor (NLP), a number of WIN-developed keyword databases, and a gender-name API. These are used and integrated differently depending on the indicator.

The WIN Gender Tracker uses a natural language processor (NLP) to recognise ‘entities’ or people. These entities are then identified as being a woman, man or unknown using a gender-name API as well as an evolving WIN name database. When a name is gender neutral or the gender is unknown, the Gender Tracker runs an additional step and looks for related pronouns and titles that may indicate the person’s gender. Otherwise, the name is identified as ‘unknown gender’.

While the WIN Gender Tracker currently has between a 90 and a 95% accuracy rate, mistakes do happen when identifying names. The more the Tracker is used, the more it learns. If you spot any errors, please send us an email at website@womeninnews.org.

At the moment the WIN Gender Tracker is available in English and Arabic only.

Tracking your content and understanding how gender balanced it is (or isn’t) is an important first step. Knowing how to improve it is a different story. WIN has developed a practical Gender Balance Guide for the media to help journalists, editors and media organisations alike to make their content more gender balanced.

We want this tool to be as good as it can be. Please help us by telling us about any errors or omissions that you spot, big or small. Please contact us at website@womeninnews.org to be put in touch with one of our Gender Tracking experts.

Yes, WIN has access to the content you input into the tracker and the resulting report. This data remains confidential, in line with WAN-IFRA's privacy policy.

WIN will only use the results of reports anonymised and aggregated with others to produce statistics and trends on gender balance in content. Individual organisation reports, data and personal information will never be used.

Your personal information will not be shared, and will be stored under WAN-IFRA’s privacy policy.

The WIN Gender Tracker has 90 to 95% accuracy compared to manual tracking. The Gender Tracker uses a combination of a third party natural language processor (NLP), a number of WIN-developed keyword databases, and a gender-name API so it is constantly learning and improving.

If you spot any errors, please send us an email at website@womeninnews.org

WIN recommends tracking at least 40 pieces of content per quarter (3 months) for a total of at least 160 articles annually.

For 5 days each quarter, select 8 articles from your website’s homepage. Please make sure the articles selected are from different sections of the website (politics, economics, culture, sports, etc.) to ensure a diverse and representative sample of your organisation’s content.

WIN also recommends spreading the 5 days when doing the content tracking over different weeks each quarter to ensure the diversity of the content analysed.

After each quarter, you can generate an aggregate report from the WIN Gender Tracker to look at the results, assess the strengths and the weaknesses in terms of creating gender balanced content.

Questions to consider when looking at the results:

• Do we need more women authors?
• Are some of the organisation’s sections more gender-balanced than others?
• Do we need to create a more diverse source list?

You can find a user guide here. It will explain to you how to set up an account, track your content, and generate reports.