Frequently Asked Questions

Questions about Capture Behavioral Engagement

1. How much time is this going to take for our IT department?

Hardly any. We have a very light footprint on your website, using a similar approach as Google Analytics. Adding our tracking code can be accomplished in minutes.

2. How fast is implementation time?

Quick, measured in person-hours, not days or weeks. And if Capture manages your instance of CBE, we do almost all of it for you.

3. Can this technology differentiate between more than one IP address at one home

No. CBE uses IP addresses primarily to identify anonymous visitors at the CITY level of precision. The IP address for the home will generally be the same for all visitors from that home. We WILL be able to identify multiple different KNOWN visitors from the same home (e.g. siblings, parent & child) based on other tracking technology that we use.

4. Does this work across devices?

Yes! A visitor can engage with the school’s website(s) via any device (computer, mobile phone, tablet) and CBE provides a mechanism to stitch together all of the visits into a comprehensive profile of the visits over time.

5. How does this system migrate into my current CRM/ERP?

Yes! A visitor can engage with the school’s website(s) via any device (computer, mobile phone, tablet) and CBE provides a mechanism to stitch together all of the visits into a comprehensive profile of the visits over time.

6. What are the differences between CBE and other marketing automation tools (HubSpot, Marketo, etc.)?

Hubspot is a very successful general commercial marketing automation product, aimed at small to mid-sized businesses. By contrast, CBE is aimed exclusively at Higher Education institutions and focused primarily on enhancing schools’ enrollment recruiting. Some other key differences:

  • Capture offers a ‘managed’ version of CBE, and we can tightly integrate it with Capture’s other recruitment services.
  • Using aggregated behavioral data, our team of data scientists will be building predictive models to help improve the conversion rates and to help our partners get more of the students they want. We expect the value derived from this data to be one of the most compelling benefits of CBE.
7. My school is already using Google Analytics on our website to track visitors. Why would we want to use CBE?

Google Analytics tells you ‘how many’. E.g. how many page views did you have on the Financial Aid page yesterday. How many people visited your site from a particular banner ad you have running. How many people clicked on a particular link.

CBE tells you ‘who’ visited and ‘what’ they did. For example, CBE can tell you that Jessica visited the Financial Aid page, plus it can tell you the exact pages she visited before and after the Financial Aid page. It can also tell the school exactly what times she visited this month, and the exact pages she looked at, right before her status changed.

Soon CBE will use all this data Capture is collecting on student behavior to build ‘predictive models’ to help understand the path students take on their decision journey to help the school influence that journey, increase their conversion rates and attract more of the students they want. Google Analytics doesn’t do any of this ‘who’ stuff.

Further, CBE gives you a variety of tools to recruit students ‘real time’, based on their engagement and their specific interests and behavior. Examples of this are the dynamic website content (e.g. toaster message, image swap), triggered personalized emails & texts, and alerts to the school’s admissions staff when students exhibit certain behavior. Google Analytics doesn’t have any of this. By communicating with students real-time, based on their current interest and behaviors we believe we can further influence the Students’ decision journey, increase conversion rates and help schools attract more of the students that they want.

8. What is the current process for managing multiple users on the same device (twins, siblings, etc). How do we track that data?

Unfortunately there’s no way currently for us to distinguish between multiple students on the same device. Both students’ histories will be merged together and associated with the first of the siblings who clicks on an email to identify herself. Our guess is that it’s not going to happen that often or end up being a significant issue.

9. What technology is used for your ‘toaster’ dynamic website content? Does it work on smart phones too? Can it be blocked by pop-up blockers?

We’re using custom JavaScript to enable this feature. This runs in the visitor’s browser and is compatible with all popular browsers on all major platforms (PC, Mac, mobile devices). It cannot be blocked by pop-up blockers.

Questions about Envision

1. What’s an ensemble model?

An ensemble model blends together multiple predictive algorithms, to maximize the strengths and minimize the weaknesses of each model. In our case, we blend together a mixture of 21st century linear and non-linear models. Our competitors rely on a single 20th century linear model.

2. What’s the difference between a linear and non-linear model?

A linear model gives a consistent weight to each variable. So if a student visits campus, it says that increases their chance of application by X percent. But we know that a campus visit from an in-state student is not the same as a an out-of-state student and neither is the same as for an athlete. A non-linear model captures the interactions between those variables.

3. I’m used to looking at p-values and R-squared values for my models. How do I know if your model is accurate?

P-values and R-squared values are measures of how well a model describes a given set of data. They are not measures of how well a model predicts data it hasn’t seen before. Instead, we use measures of how well a model predicts a holdout set of data.  For an applicant model, we consider an RMSE below .4 and an AUC above .75 to indicate a high-performing model. For an enrollment model, we consider above 90% accuracy on the holdout to indicate a high-performing model.

4. What do your performance statistics mean?

Accuracy is the ratio of predicted group level enrollment predictions to the actual group enrollment total. Root Mean Square Error (RMSE) is an industry standard test statistic for predictive model performance where smaller values indicate better model performance. Area Under the Curve (AUC) is a test statistic that varies from 0.5 to 1, where 0.5 is no better than random guessing and 1 is perfect prediction.

5. Why do you predict from prospects to application, not to enrollment?

From a large pool, only a very small number of prospects will enroll in your school. From a large pool, only a very small number of prospects will enroll in your school. Let’s say .2 percent of all your prospects will enroll in your school. If you predicted a 0 percent likelihood of enrollment for everyone, you’d be 99.8 percent accurate. That prediction isn’t very useful, but it sure looks good. When everyone’s prediction is near zero, it makes it impossible to create meaningful distinctions between those students. In other words, it invalidates the ranking. On the other hand, the number of applicants is a much larger percent of the total prospects, enabling a meaningful ranking of those prospects.

6. What features are in my model?

Capture appends about 10,000 contextual and behavioral data points to your data. Unfortunately, the mixture of those data points is proprietary and protected by our pending patent.

7. We already have more deposits than the total number that Envision is predicting. Does that mean the prediction is wrong?

Keep in mind that Envision enrollment models predict the total number of enrolled (post-melt) students. Just because you’ve made your deposit goal doesn’t necessarily mean you’ll make your enrollment goal. Take action accordingly.

8. My enrollment model is predicting that we’re below target. What should I do?

Admit more students until the model predicts you’re on target.

9. When should I submit data for another iteration?

For an applicant model — after a large influx of names, or before an initiative such as a big mailing where you’ll need a prioritized list. For an enrollment model — after a significant number of new admitted students or when wanting to test a different scenario of hypothetical admits.