We're here to help you get the most out of the Citibeats platform. Answers to the most commonly asked questions appear below, and our friendly team is ready to respond to any other inquiries.Contact Us
Citibeats is a SaaS company with an ethical and social focus, developing AI solutions to detect changes in society earlier than traditional survey methods or the media. "Changes" can be interpreted as potential risks, but also as opportunities for governments, authorities, and private companies to act for a positive social impact. As we like to say, to shape a better future, you need to start predicting the present.
Citibeats is not a brand watching solution.
Citibeats' algorithm analyzes large amounts of unstructured text data and other types of files with extractable text to identify social trends, opinions, and concerns in real time. This information represents precious actionable insights for governments, multilateral organizations, and private companies to act on to improve citizens' lives.
Through data collection and processing, we can obtain different outputs such as:
- Real-time alerts: Data anomalies, relevant clusters, and on-demand cluster monitoring.
- Leading indicators: Social signals that tell you what will change before you see it in traditional survey methods and the media.
The Citibeats system can be divided into three phases:
Collect people's opinions directly from Citibeats or connect your own data. The idea is to let you bring together many different sources of opinions in one place.
Once you gather data, Citibeats can help make sense of all of the noise and answer key questions:
- What topics are people talking about?
- Which topics are rising?
- What are people asking about?
- What are people complaining about?
- What are the differences in conversations between the female and male segments of the population?
- What are the differences in conversations between people and organizations?
- Where are the outliers in the conversation across geographies?
Of course, you may want to combine these analyses. For example, what are the rising complaints of women in Buenos Aires?
Ultimately, Citibeats outputs need to connect to your way of working. There are several output options:
- Citibeats Dashboard: Explore insights in a ready-made dashboard with a private login.
- Citibeats Public Observatories: Make the aggregated and anonymized outputs publicly available to help explain your initiative, gain visibility, and empower other collaborators to use the data.
- API: Connect the analyzed data directly into your other systems via API.
At the beginning of the project, you are asked to select a text-based data source for data collection. It could be a public data source, like news, blogs, Twitter, or other social networks, and/or private data sources you have permission to analyze, like emails, contact center call transcriptions, or CRM notes.
As guidelines, please consider:
- Unstructured text (e.g., a tweet) instead of structured text (e.g., a purchased item in a receipt).
- Minimum 10,000 docs per dataset/period you want to study for reliable models.
- Short text. A text unit should be a maximum of 1,000 characters (a long paragraph). Longer texts will compute as more than one unit of text.
Lastly, you may need to consider language coverage. Citibeats has been applied in more than 50 languages and is flexible to add more. To check current language coverage or anticipated additions of other languages, please check with your Customer Success contact.
You can connect data to Citibeats using the Input API. If you want to proceed, please contact us for the Input API documentation and your Input API key.
If you want to gather data from public sources that Citibeats already connects to, follow the guided steps in the "Create Dataset" section.
Yes. You can mix several data sources in one project and then use a filter to separate the results by data source.
Although you can mix any sources, we recommend that you use the same types of sources to limit bias and provide better results. For example, Facebook comments and tweets are similar types of data, while news articles and tweets are very different.
Citibeats projects work by combining two main elements: categories and datasets.
Categories are topics that you can use to classify opinions. These could be broad (like "environment") or narrow (like "plastic recycling").
Datasets are the groups of documents that you compare across categories. For example, datasets could be split by:
- Geography: You can use cities, regions, or countries as your groups if you want to understand citizens' concerns, improve the tourism strategy, or analyze social trends.
- Companies: This can include companies from different areas, which will help you optimize sales and customer service, as well as compare the social trends and opinions that affect your offers.
- Population Segments: You may want to compare the needs of different groups or see how opinions differ between various subsets.
In the end, the groups you choose are up to your imagination and what you want to achieve with the project.
In practice, there is no limit. We recommend a minimum of 10,000 documents per project. However, this is not a technical limitation so much as the point at which it becomes useful to use Citibeats technology rather than human analysis.
Private data uploaded by a customer is only visible to that customer. On a technical level, we isolate databases for each customer and protect the access using security standards.
In addition, for demographic profiling features, we only return aggregated and anonymized results at a minimum level of 100 people to be respectful of privacy considerations.
Finally, all the data we use for research and development is destroyed once the R&D phase is complete.
The diverse sources and huge quantity of data create the conditions for delivering high-quality insights that accurately represent reality.
Some platforms, such as Twitter, have a significant proportion of bot-generated conversations. However, this tends to be low in Citibeats projects because the type of content posted by bots is generally unrelated to the topics we analyze and not full enough to be counted as an opinion. Moreover, leading platforms like Twitter are working to implement effective strategies to reduce bot content entering their API for analytics use.
Beyond bots, there are several related issues in online data. For example, influencers and "trolls" also post inauthentic content that we may not want to consider as "people's opinions." Citibeats has additional services and techniques to remove these posters depending on the project.
The Citibeats platform supports more than 100 languages, from the most spoken languages, like English, Spanish, Chinese, and Arabic, to less common languages, like Irish, Mongolian, and Waray-Waray.
By "data," we mean "document," "opinion," "comment."
Our platform allows you to analyze data while minimizing gender bias. Once we have a clear idea of the real demographic composition of the population, it is possible to weigh and properly value all the voices coming from usually biased social media, like Twitter and Facebook.
No. Citibeats is entirely responsible for the settings and the training of the system.
Contrary to brand watching, Citibeats offers a complex analysis and interpretation of data. Instead of purely focusing on single keywords, our research methods are culturally focused and adaptable to monitor complex narratives. By complex narratives, we mean any narrative revolving around a specific topic that involves related subtopics and multiple user opinions.