Three different questions
Integrated loudness describes the full selected programme after frequency weighting and gating. Momentary loudness uses a 400 ms window, while short-term loudness uses a three-second window. Those shorter windows reveal where the programme becomes louder or quieter; they are not interchangeable with the programme-wide value.
Peak level answers a different question. A sample meter sees the values stored in the file. A true-peak estimator interpolates between those values because reconstruction can overshoot them. The workbench reports both, which helps explain why a sample that never reaches full scale can still need more headroom.
Read the channel layout first
The same signal present in both channels contributes more total energy than one mono channel. This tool measures mono as mono and stereo as stereo. It does not duplicate a mono source to make a reading appear closer to a target. Surround and LFE channels require explicit weighting and layout interpretation, so they are outside this release.
WAV and FLAC are decoded at their original sample rate. A lossy format converted elsewhere may already contain codec changes, and normalization cannot undo those changes. Keep a lossless source where possible and compare the original sample rate, duration and channel count in the report.
Use a delivery target deliberately
A LUFS preset is not a universal quality score. Speech, music and broadcast programmes have different aesthetic and delivery needs. The default is a starting value that you can change; it is not a promise that every streaming service uses that setting or accepts the same peak ceiling.
The exported report records the actual measurements rather than a pass badge for an unknown platform. Verify the receiving specification, check a representative excerpt, and listen at a comfortable level. Never increase playback volume just because a test signal sounds quiet: some signals contain substantial high-frequency energy or isolated peaks.