Frequently Asked Questions
Q: I am unable to add new monitors, why is that?
A: If the New Monitor button is greyed out or you don’t have the option to add a specific type of monitor in the setup screen, please reach out to Brian Martin as you have hit your limit for monitors.
Q: I have many inactive monitors, but the system won’t let me create any new monitors.
A: Inactive monitors still count towards your monitor total. You’ll need to delete monitors in order to create new monitors. If you are unable to delete monitors, please contact Brian Martin ([email protected]) and Lizzy Haines ([email protected]).
Q: Why can’t I access HelioSight?
A: Not all users have access to HelioSight. There is a limited number of seats available, if you need access please reach out to Lizzy Haines, your Customer Success Manager ([email protected]).
Q: What’s the difference between a Social Account Monitor, Buzz Monitor and Opinion Monitor?
A: Here is a quick overview of each type of monitor in your account: ForSight Crash Course
Q: Other than social media websites, is there a specific number of websites that Crimson Hexagon crawls currently?
A: Currently, there is no consistent system or pattern used across markets regarding how many sites are crawled/ pinged. The sites actively being pulled into our database are 100% based on what past/ existing customers have requested.
In total, we’re pulling 1,100,000 blog posts per day, 11,000,000 forum posts per day, and 250,000 news posts per day.
Q: If we have to add new websites to be tracked – say 100 URLs of news websites – can we place the request at one go?
A: To request new content, users can go to the “Team Admin” section or simply click here.
Content source additions will take some time to pull back and proper expectations should be set. Also, we are not currently able to gather historical data for new content sources, and will only collect content going forward from your request date.
- Blogs ≈ 1 week
- News ≈ 1-2 weeks
- Forums/Reviews ≈ 4 weeks
Q: How do we determine posts included in the topic wheel?
A: Visualizations under the Explore tab (topic wheel, clusters, etc.) are a sample of a sample. These visualizations are able to analyze up to but no more than 10K posts. This 10K limit remains regardless of how many total cases are included in the monitor (e.g., 30K posts over 3 days, 300K posts over 30 days, etc.). The sampling methodology employed for Explore tab visualizations is essentially the same method used to gather the original sample and should be representative.
For example, if you are looking at data over a 4 day time period with 5,000 posts in each day, the monitor will be analyzing 20,000 posts in total. The 1,000 posts that we see in the topic wheel are a random sample of this 20,000.
Relevant Resource: Sampling Overview
Q: What is the difference between using a Sub Filter and Quick Filter?
A: Sub-Filters may be applied to the total data-set (except for the Compare section).
For example, a monitor may contain 30K posts over 3 days. If a sub-filter is applied to women the total may drop to 15K posts over 3 days, and the whole monitor will be re-run based on this new sample size.
Quick Filters can only be applied to modules in the Explore tab. Regardless of the total posts included in a monitor (30K or 15K), the max number of posts for any Explore tab visualizations is 10K.
For example, if a monitor has 30K posts over 3 days, a word cloud will visualize data from a random sample of this 30K, up to 10,000 posts. Quick filters will only drill down into these 10,000 posts.
Q: How is the geography of a post determined? Are there any limitations?
A: Only 1% of all posts are Geotagged and can be located with precision, and and we use these to train a machine learning algorithm to identify patterns unique to geography.
- The “statistical guesser” draws on data like profile info, time zone, and language to identify attributes unique to a post’s location.
- The attributes IDed are then used to infer the location of posts without a known location. If a strong match is not found a prediction is not made.
Geotagged posts number in the hundreds of millions. However, a specific monitor must pull in a minimum number of geotagged posts to ensure our algorithm has an adequate sample size to make statistically significant estimations.
Relevant Resource: Location Methodology
Q: How are Segments different from Monitors?
A: Segments are analyses that focus on who is saying what, not what is being said. A segment takes a group of people and analyzes what that certain group of people is talking about.
Q. Where do affinities come from?
A: Affinities were created using data purchase from Klout as the foundation. This helped us identify the roughly ~5,000 interests we currently have. Then, we looked at what these users with different interests tweeted about and who they followed and used this data as material for teaching a machine learning algorithm. We then used this to train the algorithm to identify interests for most users on Twitter (those users who could not be assigned an interest(s) with confidence were excluded).
It’s important to note that while we have assigned affinities for other languages, the algorithm was trained using English language conversations. As a result, affinities assigned to Arabic users may have a lower confidence interval.
Also, affinities and segments are actually two different ways of looking at the same data. That is, segments are all users that have a specific interest. Affinities are the likelihood that someone from that segment is engaged in the conversation you’re studying. As such, if you’re ever curious to take a closer look at a given affinity, all you need to do is head over to segments and look it up.
Lastly, currently, affinities are static. They were created a year or so ago and haven’t been updated since. We’re currently working on a new, dynamic version of affinities that will actively change and grow. This version will also have much higher accuracy users tweeting in foreign languages. I was told by our product team that we may expect this update by the end of the year.
Q: What are some best practices for removing spam?
A: Spam is a tricky subject because we may define what constitutes spam very differently depending on the client’s research needs.
Further, keywords used to ID spam may vary depending on category.
If you develop a set of Fidelity-specific best practices that include a set of universal exclusions, you may create a template to easily apply these exclusions in every monitor you create.
Basic techniques to ID appropriate exclusions
First, search for a brand/product on Twitter to ID possible exclusions before creating a monitor.
Second, use the “preview” function when setting up a monitor to further refine search parameters.
While Opinion Monitors may be further refined to identify and exclude irrelevant content via the Brightview algorithm, our best practices emphasize a strong initial design for best results.
Q: When should we consider excluding URLs?
A: AND -(http OR https) will exclude any posts with links from being brought back in your results, including pictures and video.
This may be useful depending on your research question. You may wish to exclude links if:
- You only want to include organic conversation about a topic
- You do not want to include promotions or deals which often include links
- Using this exclusion is also useful if you would like to avoid any news/ photo/ video/ article sharing.
Q. What are some different approaches for brands/ products with commonly used words?
A: Exclusionary Approach:
- An exclusionary approach is useful when a product or brand in your monitor includes a commonly used word when other uses are easy to identify
- The boolean operator: AND –(X) is the primary operator for an exclusionary design.
- When it’s difficult to identify word use because there are too many variations, we may choose to use an inclusionary approach to exert more control over the sample frame.
- The boolean operator AND is the primary operator for inclusionary design (though we will still rely heavily on AND -)
- Workshop: Strategies For Cleaning Up Your Keywords
- Reducing Spam Content in Monitor Results
Q: How can I add new logos for logo detection?
A: Reach out to your Customer Success Manager, Lizzy Haines ([email protected]) to add a new logo.
Q. Why do I need credentials for Instagram (Facebook…)
A: Instagram and Facebook require personal account credentials in order to pull data from their system. It’s meant to serve as a mechanism to prove you are a real person requesting this data, not a bot (similar to captchas). This information is saved in our encrypted database and it is never viewed by anyone. It is used to make generate tokens that in turn allow us to make API requests on your behalf (so that users don’t need to enter their credentials on a daily basis). Instagram and Facebook do not track this information and they receive no metadata from Crimson Hexagon (i.e., keyword searches, company or account info, associated users, etc.).
Q. I made changes to a monitors keywords/data sources, but I don’t see any changes in the data already collected.
A: When you update keywords or add a new data source you may select “Save” or “Save & Reset”
- “Save” implements any changes made from that point forward, no impact on historical data
- “Save & Reset” implements any changes made retroactively and does impact historical data