At SparkPost we believe that no one single source of data can provide you a comprehensive view on your email performance and that senders are best served by a holistic view supported by diverse sources. This is an approach grounded in data which provides multiple benefits:

  1. It provides multiple views to pinpoint and aid in decision-making for optimizations or to resolve deliverability issues to drive a better return on investment from your emails. 
  2. It helps eliminate outliers that you wouldn’t likely see if you only had one or two sources. 
  3. It allows for the use of sparse but highly accurate sources (like permissioned panel data) to serve as a training and calibration source for AI and ML-based techniques (like Signals Health Score and Intelliseeds) to ensure that they’re as accurate as possible.

Transparency is one of our corporate values that we live and breathe each day. As the world’s largest email sender with the largest analytics footprint (since acquiring eDataSource),we are proud to have the largest data network of any delivery analytics provider, not just in terms of overall data points, but in the depth and diversity of the sources we pull from. The SparkPost data network currently consists of eight sources:

  1. Active Sensor Network: This source is comprised of recycled and pristine domains owned and operated by SparkPost.  This data provides greatest insight around senders – on and off our platform – with very sloppy list acquisition and management practices and helps provide telemetry which we can leverage internally to help customers as well as to detect and remediate bad senders on our platform and (often) avoid them in the first place. 
  2. Passive Sensor Network: This sensor network allows us to monitor typo spam traps, recycled spam traps, and parked domains but do so using passive DNS and signature-based methods which keep user content completely private from us. Passive DNS methods are a well-established technique used for security monitoring across many different areas. This passive sensor network, which provides similar coverage to that provided by the active networks of other providers, serves as an effective way to monitor list hygiene and sending best-practices compliance.
  3. Third-party Pristine Trap and Content Scoring Data: We receive third-party data from industry leaders Abusix and Cloudmark, which provide extremely accurate information on aberrant behavior. Though the number of hits on these sources is low compared to the traditional informational trap networks that other analytics providers deliver, it’s because these are attuned to extremely serious list issues versus simply doing ineffective new user validation, typo collection or poor engagement-based suppression.  
  4. Permissioned Panel Data: Our permissioned panel data is a set of users who have explicitly opted in to sharing the commercial email in their inboxes so that we can understand behaviors around interaction and inboxing for real human users. With representation at 42 different service providers around the globe, the permissioned panel data provides insights into real human interaction with your mail. The breakout by country on our panel aligns very closely to internet penetration by country, giving a balanced view of how consumers respond to your emails globally. It also serves as a training data and calibration set for our other ML-informed functionality like Intelliseeds and our Predictive Health Score, to help those sources provide the most accurate results even for senders whose mail streams may be too small to get meaningful data directly out of the panel.
  5. Traditional Seedlists: This is a basic and table stakes-level source that shows inbox delivery. This is what other delivery analytics providers have as their primary solution. Traditional seedlists are a well-established solution, but lack accuracy as providers have moved to a more engagement-based AI-driven filtering.  As major providers have made the move to engagement-driven models for inbox/tab placement, the accuracy of methods like traditional seeds, where seeds never engage with mails (or always engage with mails) produce sub-par results. 
  6. Intelliseeds: To address the emerging accuracy issues with traditional seeds, we introduced Intelliseeds in 2019, which take traditional seeds and drive their behavior in terms of reading, skipping and deleting mails off a model trained against our permissioned panel data. By engaging with mails in a more realistic fashion, you bypass the fundamental issues with traditional seeds. You can learn more about how Intelliseeds differ from traditional seeds in a future blog post. Seeds and Intelliseeds provide accurate inbox performance signals for senders whose volumes may be too low to be well-represented in our permissioned email panel.
  7. IntelliX Virtual Persona Network: The IntelliX Virtual Persona Network allows us to automatically subscribe to marketers’ email lists and receive their campaign mailings at email accounts we fully control. What makes the Virtual Persona Network unique is that we use a distinct and unique email address for every email list. We never re-use the email addresses with the Virtual Persona Network. This allows for incredibly powerful insights, including: the world’s largest indexed repository of opt-in commercial mail, ability to analyze brand marketing shifts over time, and the ability to detect list compromises and list selling. We are able to provide the raw HTML content of the email message, full images, and because we run all of the Virtual Persona data through our own anti-abuse filters we are able to provide detailed spam scores and insights that are often not available to senders.
  8. Signals Health Score: SparkPost Signals’ Predictive Health Score helps you understand the real-time health of your campaigns through predictive email intelligence. Using a machine learning model trained on email leading indicator signals from deliverability and engagement data from your sending such as: bounces, opens, clicks and spam trap hits, and more. SparkPost makes this possible by comparing your sending signals with historical performance from thousands of other senders so you can optimize your email engagement and avoid reputation issues before they impact your business.   

We’re always looking for new ways to improve and expand the data sources that provide SparkPost’s clients an unmatched view of the email ecosystem and how to expand our massive data network as new sources become available (either home grown or through third-party sources). When a new source is added, we will certainly share it with you. 

~George