#RIUC13

For those of you who weren’t able to attend the 2013 Rapid Insight User Conference, we set a new record for most attendees and largest number of customer presentations. With two full days of dual track programming, the presenters covered a lot of ground. While we wait for some of the video recordings of customer presentations to be formatted, I thought it would be good to do a quick recap here. 

Mike Laracy, Data Geek (at right)

The conference opened with a keynote from our Founder and CEO, Mike Laracy, who talked a bit about the future of predictive analytics. With a mass public education on the value of analytics (from people like Nate Silver and Billy Bean, with a little help from Brad Pitt), as well as significant advances in data storage and processing power, a stronger need for predictive analytics is emerging. The market is shifting towards the view that more data access is better than restricted access, and that given the right tools along with access, smart people – data scientists – can turn raw data into actionable information. Given these changes, the data scientist – that’s you – will be in increasingly higher demand over the next decade and beyond, as will predictive analytics. 

The user presentations covered lots of different topics, and we’ve made all of their slide decks available here; I’d highly recommend checking them out. In addition to what’s there, I’d also recommend checking out some of the interviews we’ve done with customers on building campaign pyramids and using predictive modeling to drive fundraising efforts. The RI staff team also gave a few presentations,  including topics like Tips and Tricks in Veera, Techniques for Improving Your Predictive Models, and An Introduction to Reporting and Dashboarding with Veera.

Another thing worth mentioning is that we announced our partnership with Tableau to provide a complete solution for both predictive modeling and visualization. Now users can use Veera to clean up their data, Analytics to build their predictive models, and Tableau’s visualizations to turbocharge their presentations. For more information, check out our partner page.

I am now a scientist according to @RapidInsightInc. I feel super smart! #RIUC13
— Dustin Mayfield (@dlm0078) June 27, 2013

My favorite part of the User Conference has always been talking to customers about the cool data projects that they’ve been tackling, and this year was no different. Kudos to our users for being so creative and smart with the ways they use our software. We also owe a big thanks to the folks at Yale for hosting us, and to all who were able to attend. Here’s to the best User Conference so far and to making next year’s even better!

Customer Tips: Roundup Edition

Here are a few extra tips for a roundup edition of our Customer Tips series to get your new year off to a good start!

Sometimes it's helpful in an analysis to recode variables, like recoding a binary variable for retention to a binary variable for attrition. 

- Jean Constable, Director of Institutional Research at Texas Lutheran University

Problem: a new input file every two weeks needs to be processed by the same job time after time.

Solution: Use a generic file name for the file in the input node. Copy the weekly input file to the generic file name and run the job. This works as long as the input files have an identical format. 

- William Anderson, CIO at Saint Michael's College

If you throw a filter between your input and the merge even if not populated when you change the input file, you do not lose the merge connections or fields!
-Loralyn Taylor, Registrar Director of Institutional Research at Paul Smith's College

Customer Tips From... Jeff Fleischer (Rapid Insight Inc.)

Okay, okay... Those of you who have worked with Jeff know that he isn't really a customer. But, as the Director of Client Operations here at Rapid Insight and an analyst at heart, he is a wealth of information, so I've decided to share some of his tips. They are:

1. The Format Column node was retired a few releases back in favor of the Convert node. Using Convert, change the data type you wish to format to "text", then select the desired style from the node's Format column. PS: The Format Column node is still accessible - try going through the menu Node -> Add Report -> Format Data. 

2. Merge nodes will forget their setup if you disconnect them from their inputs. Placing a Cache node just before the Merge will keep it from forgetting its configuration if you need to relocate or copy it. 

3. Merge nodes also act as a rename. Just edit the text in the Output Column and use the black arrows on the menu bar to rearrange the column order. Bonus tip: select multiple columns before using the black arrows to re-position several at a time. 

4. Use the menu option Edit -> Convert All Columns to Text to do just that. 

5. Dropdown box controls often respond to single letter entries, avoiding actually having to pick from the dropdown list.

6. Select multiple fields using Ctrl-LMouse to Cleanse multiple fields by setting up a single rule/operation. Note that the fields all have to be of the same type (text, integer, date, etc.) for this to work. 

7. Use the new "File Created Column" option in a Combine Inputs node to identify (with a Filter or Dedup) the most recent records coming from a location. 

Subroutines (Customer Post by Tony Parandi)

Today's blog entry comes from Tony Parandi, Assistant Director of Institutional Research at Indiana Wesleyan University: 

One feature of Veera that I’ve found very helpful is the Subroutine node. This node allows you to feed the output of another job directly into the job you’re currently working on. In essence, it allows you to put a job within a job. This is especially helpful if you have a certain data stream that you commonly use, and do not want to rebuild it each time you need it.

An example that I commonly use the Subroutine node for is the recoding of student ethnicity. In 2010, when the Department of Education mandated the new ethnicity categories, we added two additional ethnicity/race fields to our data system (Datatel). Thus, students could have a wide myriad of ethnic category combinations, which means we have to reform these combinations to match the IPEDS definitions. In order to accomplish this I bring in the ethnic data from our warehouse, and use a series of transform nodes to convert the three ethnicity fields into one final ethnic category:

Rather than recreate this stream for every job, I have it saved as a separate job, called “Ethnic Conversion”. As you can see in the picture above, the Output Proxy node is necessary when creating a job for Subroutine purposes as it connects the output data to the Subroutine to the other job you’re building.

Now when I create a new job that needs ethnicity reformatting, I simply bring in a Subroutine node and connect to my data via Student ID in the Merge node.

The output gives me a single ethnic category that matches IPEDS for every student, based on data brought in via the Subroutine.

Although this is a small and simple example of a Subroutine job, the node is a powerful way to connect jobs without having to do copy/pasting or rebuilding. I have found the Subroutine node to be a great time saver, and I encourage everyone to use it whenever possible.

PS: Be sure to check out the rest of our customer tips series here!

Customer Tips From... Brian Johnson (DonorBureau)

Today's installment in our customer tips series comes from Brian Johnson, the VP of Product and Operations at DonorBureau. Here are his tips: 

1. You can have multiple select statements in the Query node as only the last statement does not have a temp table statement. This has allowed me to automate any complicated SQL queries I have for reports to run every week. 

2. If you need to create a new version of a scheduled report, use the old version and it will inherit the schedule of the original report in terms of run times. 

3. You can save output to Dropbox or Google drive to distribute data to team members without having to fill up their inbox. 

Customer Tips From... Dr. Nelle Moffett (California State University - Channel Islands)

The next installment in our customer tips series comes from Dr. Nelle Moffett, Director of Institutional Research at Cal State - Channel Islands. Nelle is an avid Veera user and loves building analytic processes. Here are her tips:

• Use a Rename node before an output to select only the fields that you want and re-sort them in the desired sequence.
• Use the Cleanse node liberally before any Transform node to remove or replace missing data. This will eliminate unexpected results when the Transform node encounters missing data. 
• To create your own variable labels, create a look-up table in Excel with the original values and the new value labels. Then merge this file with the original file using the field labeled "original" and keep the new field with the desired value labels. 
• Example of value file:

• To calculate the percent of a certain characteristic in the dataset, first use a Transform node to set a flag for that characteristic where 1= has the characteristic and 0= does not have the characteristic. Then use an Aggregate node and select the mean for a flag.
   
• To update a job to the current form of a dataset, first make the connection to the updated dataset. Then double-click on the data node in the job. At the bottom of the window where it says "connection" click on the name of the data file and select the new version. make sure all of the data fields are checked (that should be) and save the changes. If you give the data node a generic (rather than dated) name, then it will still be appropriate as the data continues to be updated to the current date. 

...Have tips of your own? Email them to caitlin.garrett@rapidinsightinc.com!

Customer Tips From... Dan Wilson (Muskingum University)

Our next set of customer tips comes from Dan Wilson, Registrar at Muskingum University. Dan typically uses Veera for repetitive and/or complex reports, including multi-year enrollment history by date, historical majors and minors (by year and department), and IPEDS reporting. Here are his tips:

1. It is important to remember the merge characteristics (all from a, all from b, all from both, only matching, etc.) so the last thing I do in developing any report is to verify each of these. 

2. While Veera's CrossTab feature is quite useful, I find it easier and more familiar to output my results to a target excel file, and then have another excel spreadsheet with my pivot table that has all of the formatting and other features set up. That way I can update the data file without overwriting my formatted "results" file. The same can be done with separate sheets in a file, but some of my reporting files pull data from different queries and Veera reports. For those reports I can run data from several sources, then open up my main file and hit "refresh". 

3. For those instances when a transform looks like a computer program, I'll break it into smaller bits and spread it out over several nodes. This allows me to test smaller chunks of the function at a time and locate any errant code prior to needing valium. (Editor's note: using the de-bugger in the Transform node can also help to find errors quickly!)

...Have tips of your own? Email them to caitlin.garrett@rapidinsightinc.com!

Customer Tips From... Dr. Loralyn Taylor (Paul Smith's College)

Our new customer tips series will feature tips from customers on using either of our software applications. Each entry will focus on one customer’s ideas to make your lives easier.

We’re kicking things off with Dr. Loralyn Taylor, from Paul Smith’s College. Dr. Taylor is a one-woman IR office and Registrar and is constantly looking for ways to save time when creating reports and executing jobs. Here are her five tips:

1.  Take the time to rename your nodes so that you can easily follow your line of thought as you move through the job.
2. Remember that there are multiple ways of doing things. The shortest is not always the best – to me it is often more important to be able to easily follow my thought on how I am working through the problem than to do it elegantly in the fewest number of nodes.
3. Common problems to check for: data format incompatibility (just use a convert node), and sometimes a null is not actually a null (just because something looks blank doesn’t mean that it is).
4. Remember that creating a job is like solving a puzzle; you have to think about it and play with it.
5. I often have to run jobs many times to get them right. Helpful tip: Set up a test data output that you can move around to different parts of the job to see how your data is coming through at different points when you are troubleshooting. 

...Have tips of your own? Email them to caitlin.garrett@rapidinsightinc.com!