Data collection & storage

Can you collect contextual data?

Yes. We automatically collect detailed first party contextual information without a need for manual labelling, though our system supports manual label override for specific circumstances. This includes categorisation into interest segments, detection of brands and demographic modelling and revenue attribution. Each of these are then processesed with a confidence or intent score based on the history of the profiles' interactions. For instance, many of our demographic signals (including gender, age and income) are first party and as such we can give a demographic score to 100% of users.

Can you customize what contextual data is collected?

Yes. We provide both automatic and manual classifiction of users to provide both reach and accuracy using our NLP/contextual modelling processes. Our ability to target audiences built on the classifications provided by the contextual data allow for both simple and complex targeting of users. This can be further expanded by utilizing our custom event system to provide unique data insights that can be used independently or combined within the existing contextual targeting system.

Do you use edge computing? Or cloud computing? Or something else?

Yes. We use edge and cloud computing as well as dedicated physical servers as required to maintain real-time processing capabilities and user identity stability even in the event of a browser based change in policy in local storage throttling. We feel strongly that running Javascript on the user's browser as the only and main source of operations creates significant exposure to policy changes and slow-downs in performance. As such, we store and process user data at the edge (browser level), and then utilize cloud and physical storage infrastructure to combine and process user data for action in both client side and server side contexts.

How does your technology allow us to future proof our data from future regulations on top of going cookieless?

Carbon does not rely solely on edge based storage and processing, but instead distributes storage and processing capabilities across edge, cloud and physical resources to ensure performance as well as stability even in the event of browser based changes in policy. We've also built in privacy compliance at the core with a tag and processes that are validated by CMPs to ensure we abide by GDPR, CCPA, and other global regulations.

Can you collect behavioral / engagement data?

Yes. We model your first party data into behavioral and engagement attributes that are considered part of user intent. We go a step further by scoring the relationship of the user to intent and brand based on variables such as recency, frequency and contextual characteristics such as purchase opportunity. We then layer in demographic and revenue data to better understand the audience and value of that audience based on real-time buy-side bid data.

Can you collect on all platforms, devices, and environments?

We are not aware of any current browser/platform limitations (including AMP). We have solutions for normal sites that can use Javascript, platforms that dont allow Javascript and only pixels, and Google's AMP product. From time to time new limitations will come up and we work to address these for our clients. Areas where we are aware there are limitations include: in-app and Facebook Instant Articles inventory.

Can you collect declared/zero party data (e.g. forms, surveys etc)?

Yes. This can be done through custom events to enable targeting and reporting based on this data. Other use cases can be discussed with the support team. e.g. improving the quality of first party demographics using declared data.

Can you customize what type of data is collected?

Yes. We can manually override classifications and also fire custom events and pixels to define audience data. We also work with clients to define and map inventory and classification structures as part of the deployment process, and will continuously adjust brand and content categories as new classifications present themselves. Likewise, our customer success team will work with clients to help define and implement custom audiences as required. Note that our system uses AI to create automatic audience profiles based on reach and value characteristics that are available at any time for activation.

How long does it take to process a user’s data?

Carbon splits processing into two targeting mechanics:

• Fast processing is done in seconds (typically less than 1,000 miliseconds) based on in-session user activity;
• In-depth modeling is accomplished by layering brand detection and demographic data which requires hours of user activity to create a meaningful profile.

As such, creating real-time audiences needs to be balanced with audience quality and performance considerations. Our account management team can help determine the best balance of speed to accuracy for each unique publisher.

Can Carbon increase my first party data collection online and offline?

Carbon’s automated 1st party data categorisation engine with custom integration opportunities, 1st party cookie implementation and key-value pair activation delivers high quality first party data collection and activation.

Carbon’s automated categorisation engine captures high quality signals from every page seen and can be enhanced to capture additional page meta-data, engagement events such as scroll depth and page dwell, exit events such as ad click-outs as well as publisher-specific signals and custom integration with available offline and CRM data sources.

As part of its Revenue Management Platform, Carbon also collects all pre-bid activity in real-time providing first party revenue performance data at impression, pageview, eCPM and total revenue level across all URLs, authors, content categories, UTMs, Ad units, referrers, bidders and any other meta-data available.

Carbon can be deployed as both a 1st party and 3rd party cookie solution and also utilises local storage and browser cache accompanied with key-value pair audience targeting in the ad server to increase reach and targetability of Safari and similarly restrictive environments.

How can we use Carbon to collect more 1st party data?

Carbon provides a number of tools & services to collect, improve and extend 1p data including:

• 1st party cookie activation to extend the core audience reach
• Data activation and source is always declared for full transparency
• Apply higher categorisation quality tools to increase the granularity of the signals captured
• Dedicated data science lead as part of Carbon's client success team
• Utilise revenue analytics to capture the value of site visitors in real time. Audience value is a unique 1p data signal that we can deliver as a platform that does more than a stand alone DMP
• Custom tracking. Most events such as ‘logged in’, ‘ad clickers’, ‘forum contribution’ can be passed into Carbon as 1p data
• Pixel insertion for email, advertisers or sponsors.

How does Carbon's first party data categorization process work?

Carbon automates the process of finding, mapping and optimizing page categories and data signals so publishers can easily identify and package audiences for the marketplace with efficiency. It’s important to state that we deal with contextual and semantic understanding as part of our IP. Carbon doesn’t need to work with third party contextual partners, or ask users to manually categorise pages whilst they are building rules. This is how we surface brand data for example.

Carbon’s categorisation engine captures as much information as possible from the URL, page header and page content in order to place an event within Carbon's taxonomy of over 5,000 intent scored segments. The stream of raw text data is first cleaned to remove structural elements, leaving a collection of words which are then scored based on their importance across multiple pages using TF-IDF and natural language techniques. These scored keywords are then transformed into numerical vectors using K Nearest Neighbour and a supervised classification method with word similarity calculated from the distance metric between two such vectors in order to place a keyword within the Carbon taxonomy.

NB: Keywords are retained within Carbon and can be utilised for audience targeting in additional to segment/taxonomy based targeting if required.