People ask me about developmental norms, because we know that the Sanders 1972 norms are unrealistic. Here is a chart of the Grunwell 1987 norms. As you can see, kids are saying nearly all speech sounds by age 4. We don't care about the two "th" sounds, because they typically do not impact intelligibility. The black lines represent 75% of kids; the pink lines represent 90% of kids. Grunwell has a pretty reasonable expectation for kids, because it focuses on omissions and substitutions. It does not focus on distortions, which is great, because we do not focus on distortions either.

There is much disagreement on developmental norms. Every study is different, which is because they collect data in different ways. Here is a link to a variety of "norms." You will see that the studies generally agree that stops, nasals, and /w/ are earlier developing, and "th" is later developing, but the ages differ, and development of other consonants sometimes greatly differs.

I find the Prather, Hedrick, and Kern norms intriguing. They are two-position norms, measuring accuracy of sounds in initial and final position. I think that tells you a lot. Medial position is often measured with sounds in blends, such as the stimulus word "window" on the GFTA-2. If you are a cluster reducer, you will likely say winnow or widow, and does that really tell us anything about the /n/ and /d/ sounds? Initial and final singleton consonants are much more useful developmental information. Note that phoneme-focused norms tell us nothing about pattern errors. A child who deletes final consonants will be recorded as unable to say /p, t, k/ etc., when the phonemes are really not the problem. That makes the norms report later development for sounds, when the sound may truly be produced by the child in other contexts, such as initial position. Process or pattern errors are much more useful to us when we consider eligibility, because they relate to intelligibility.