If you are going to combine net averages from five to three buckets, why not just use three categories in the survey?
Combining the percentages is useful in visually showing at a quick glance the good and the bad. It also makes it easier to show comparisons for your organization versus database norms–you just show the net favorable scores for your company and the net favorable database norm–just two numbers to look at per question. If you instead show five percentages for your company and five norms per question, you’re expecting the reader of the report to be able to look at 10 numbers per question to come to a conclusion instead of 2. That’s not really possible for human beings, especially when you might want to show 7 related questions on a single slide. That would be 70 numbers to find a trend in instead of 14. That’s when it’s useful to show fewer buckets than you asked about.
But there is a benefit to still having the data captured in the five buckets in the first place. In addition to capturing percentages for each response option, I typically ask the data processing firm I work with to calculate mean scores (averages) for each question, where a “strongly agree” might be worth 5 points and “somewhat agree” might be 4 points. You might find that there is no apparent change in the combined percentages for the two “agree” options from one year to the next, and conclude that nothing changed year over year when it really might have shifted to more people strongly agreeing. Let’s say your “net agree” percentage in both years was 70%. But if in the first year you had 10% choosing “strongly agree” and the next year you had 30% choosing “strongly agree,” the mean score will show a strong improvement even though the net percentage remained the same.