If you are running a droplet with multiple cores 103% may not reflect a lot. If you see occasional spikes but your normal use remains low this may not present a problem. In general scaling to a larger size is usually prompted by a need for additional memory. Running free -m will show your used/available memory and swap. When you max out your CPU use, things may slow down and if these spikes are causing slower responses it may be time to consider an upgrade or optimization of your services configuration. When you run out of memory on the other hand you could see processes killed or services stop so it’s generally a larger (or at least more immediate) concern.

You can use the CPU/Memory upgrade option to compare performance. If you do not resize your disk as well, you can revert back to the smaller size if needed. If you upgrade the disk as well it cannot be reversed due to the risk of data loss with the shrinking of the filesystem.