As our last post looked at how Facebook topic data allows analysts to go way beyond what social analytics had been able to achieve before, now it’s time to get tactical. We believe that Facebook topic data is genuinely transforming social media analytics, as well as overall marketing and advertising programs, delivering unparalleled insight into audiences and their preferences.
So with that in mind, we want to help everyone that uses PYLON for Facebook Topic Data to get the best possible return. This post will provide practical tips that will allow them to do exactly that.
Creating insight from data
We have written previously about how PYLON has been installed behind Facebook’s firewall. Not only does this comply with DataSift’s privacy-first approach, but it also enables the receipt and processing of the anonymized stream of data generated by users posting and engaging with stories on Facebook.
Marketers can then create insights from the data by following these five steps:
- Create a filter – create an interaction filter to select the required data from the huge volume of real-time activity happening on Facebook using Curated Stream Definition Language (CSDL), DataSift’s own language for processing human data.
- Add custom classification rules – add domain expertise to the data by creating classification rules using VEDO, DataSift’s own classification engine. For example, create classifications for brand names, key product features, or to discover intent.
- Run a recording – when an interaction filter is run, the data that matches the filter is recorded into a private index, which can only be analyzed by the filter’s creator. The recorded data includes any classification.
- Analyze a recording – submit analysis queries to the index and incorporate the aggregated results into an application or analysis. There are a wide range of attributes that can be used for analysis covering demographics, topics, links, sentiment, etc.
- Validate and optimize filters – PYLON gives access to Super Public posts that match an interaction filter. With Super Public posts, filters can be validated to determine if they are working as expected and to remove unwanted noise.
Techniques to go that extra mile
With the basic workflow above in mind, here are some techniques that will help analysts and marketers to get maximum value from Facebook topic data:
Create better filters – data scientists can spend more than half of their time simply cleaning and processing data before exploration. So an effective filter to remove unnecessary noise from the massive volume of Facebook data is essential. To minimize data cleaning, here are some useful tips for creating better filters:
- Filter with precision using topics and keywords – the built-in topics and topic categories in Facebook topic data make it easy to create powerful and accurate filters. Facebook’s natural language processing technology allows analysts to start with a very simple filter, using one or more topics. When the analyst is comfortable analyzing the broad range of collected data, filters can be refined with keyword filtering.
- Expand filters with link titles – Facebook topic data comes with the ability to filter on link titles. If terms are not mentioned in the URL but are mentioned in the page title, it is still possible to capture this data for analysis. Link titles extend the reach of filtering allowing the capture of more relevant shared content.
- Validate and optimize filters – there are a number of techniques available to validate and optimize filters that include leveraging the verbatim text from the Super Public posts, reviewing links and hashtags results, and analyzing the results of custom classifiers.
Add value to the data – as more companies are using Facebook topic data, an analytics product that stands out from the crowd needs to add unique context and value to the data. VEDO allows analysts to:
- Classify topics – apply domain knowledge and tag topics in interactions.
- Identify emotion and intent – tag keywords and phrases that identify emotion and intent.
- Reuse rules across recordings and projects – write classification rules once and use them on multiple projects.
Interactions are recorded to the index with any classification. The classification is then available for analysis. Adding classification allows a wider range analysis options and provides unique insights into the audiences.
Analyze the data – there are many techniques that can be used to extract meaningful insights from the recorded data:
- Multi-dimensional analysis – the multi-dimensional nature of Facebook topic data allows for analysis of otherwise hidden insights. Make analysis queries about a specific parent attribute and break down the results by another attribute, and then another. For example, instead of only generating analysis of positive stories about a brand, breakdown the results by the author type and location attributes.
- Segmentation based on demographics – PYLON allows access to demographic details of users, including age, gender and region. This enables analysis based on self-declared demographics. Analysts can easily segment an audience and associate activities and interests with a certain demographic group.
- Comparison using baselining – a great technique for comparing results against a random data sample for the same audience to see what is normal and what is notable. Uncovering the audience segments that are uncommonly interested in a brand or product can lead to new marketing campaigns, new products, and new business strategies.
Out-of-the-box PYLON features, combined with the structure and richness of Facebook topic data makes a compelling way to extract deeper, richer insights quickly. Our final post in this series will look at the considerations involved when building Facebook topic data products.