The Recording Academy had split Grammy.com and RecordingAcademy.com to serve different audiences. In this sponsored Global Career Accelerator project, I examined the supplied website data to decide whether keeping them separate made sense.

The answer was not that every metric improved. People viewed more pages per session and bounce rates fell, but time on Grammy.com dropped. My job was to make a recommendation that accounted for all of that.

Award nights were a different kind of day

I started by plotting daily visitors and comparing ceremony days with regular days using a PivotTable. Ceremony days averaged roughly 1.39 million visitors, compared with about 32,400 on a regular day. The size of that difference made the event calendar hard to ignore.

The assignment used February 1, 2022 as the content-split boundary. I classified the Grammy.com records before and after that date so I could compare the former combined site with the two separate audiences.

Compare behavior, not just traffic

I calculated pages per session from total pageviews divided by total sessions, and bounce rate from bounced sessions divided by all sessions. The former combined site had about 1.86 pages per session. After the split, Grammy.com had 2.25 and RecordingAcademy.com had 2.78.

Bounce rate moved from 41.6% for the combined site to 40.2% for Grammy.com and 33.7% for RecordingAcademy.com. Those measures supported keeping the sites separate: each site appeared to be doing a better job of encouraging interaction.

Time on site complicated that picture. Using the assignment’s average of daily session-duration values, I found roughly 103 seconds for the combined site, 83 seconds for Grammy.com, and 129 seconds for RecordingAcademy.com. I kept that difference in the recommendation instead of describing the split as an improvement on every measure.

The audience also changed the design priorities

I used XLOOKUP to match mobile visitor counts by date. The calculated mobile share was about 73.7%, making mobile experience a substantial part of the audience question.

The competitor comparison added another useful tension. The supplied American Music Awards dashboard showed longer visits and more pages per visit, but also a higher bounce rate. In my written response, I noted that a shorter visit could mean someone found the information they needed quickly. Longer sessions alone were not enough to call one experience better.

Make the recommendation without overstating the test

I recommended that the websites stay separate, based on the pages-per-session and bounce-rate differences while acknowledging the mixed duration results. I also completed the workbook’s extension using supplied control and test values to test whether pages per session differed at a 5% significance threshold.

That exercise was separate from proving that the website split caused every observed change. This was a before-and-after analysis with different audiences and event patterns, not a randomized rollout that I ran for the Recording Academy.

The project taught me to keep the recommendation and the uncomfortable numbers in the same story. A useful analysis should help someone decide what to do while still showing what the metrics do not settle.

  • Excel
  • PivotTables
  • XLOOKUP
  • Web Analytics