This year’s Song Of The Year Grammy went to this haunting ballad, so I was immediately primed to look out for anything that sets it apart from the crowd in that respect. From a harmonic perspective, the highlight for me is the way the repeating D-C_#m-F_#m progression of the verse and prechorus sections seems for a moment as if it’ll continue when the chorus starts on another D major chord, but that the expected downward root progression then instead moves upwards in the second bar to an E major chord instead. That on its own would already have given the chorus a harmonic uplift, but then we also get a C_# major chord instead of the verse/prechorus C_# minor, not only moving to a more optimistic major mode, but also providing the extra tonal momentum of a traditional V-i perfect cadence into the following F_#m tonic chord.

There’s a lot to admire in the melody writing too, particularly the efficiency with which the writers eke out a small kernel of simple melodic material with just enough variation to keep the repetitions from becoming monotonous. So, for example the basic F_#-E-A-B riff in the first bar of verse one (“things fall apart”) repeats in bar two (“and time breaks your heart” – and that’ll be 50 pence in the hackneyed rhyme box for “heart”/“apart”…), but then just when it seems like it’s going to repeat in bar three (“I wasn’t there…”), it’s varied with a short A-C_#-G_# extension ("…but I know."). Bars 5-8 then seem as if they might be a simple repetition of bars 1-4 (complete with another 50-pence fine for “girl”/“world”), but at the last minute we get a further variation of the fourth bar’s tail, changing the contour from A-C_#-G_# ("…but I know.") to A-B-C_#-G_#-F_# ("…and you both let go."). And in the second verse, again it starts off pretty much as if it’ll be a pretty uncomplicated melodic repetition of verse one until bar six (“more different than me”) brings an unexpected variation of the basic riff from F_#-E-A-B to F_#-A-B-C_# – and 50-pence rebate for a deftly placed bit of prosody!

But the melodic feature that I most like in this song occurs under the words “time” and “sign” in the two main choruses. In the first chorus, “time” has a lovely three-note melisma which starts with that evocative, yearning D-E_# augmented second, but then the E_# leading-note kind of shies away from proceeding to the tonic F_#, instead deflating back down to C_#. And when “sign” arrives, it feels almost like the melody has lost confidence after the previous phrase’s failure to reach F_#, settling instead for a lower-energy D-E-D-C_# contour instead – a change made all the more apparent on account of the clashing E_# in the backing harmony’s C_# major chord. In the second chorus, on the other hand, “sign” finally does push up from the leading-note to the tonic, reiterating this exhilarating melodic movement several times in a row to triumphantly push us into the textural climax of the song at 2:53-3:21.

There is one head-scratcher for me, though, in terms of the songwriting: the odd, low-energy outro from 3:38 that trails on after the song feels like it’s already finished. This seems so redundant to me musically, that I actually wonder whether it may be nothing to do with the songwriting at all – and that I might actually have spotted the first high-profile example of a loudness-boosting hack I’ve been expecting to encounter for a while now. You see, the EBU R128 measurement algorithm that most of the major streaming platforms now use to power their loudness-normalisation engines effectively averages the loudness of the your song’s entire timeline. In simple terms, what this means is that the more lower-level sections you have, the lower the average detected loudness value, and therefore the less the loudness normalisation routine will turn down the playback volume of your production. And in this specific scenario, the meandering outro of ‘Wildflower’ effectively reduces the song’s measured LUFS loudness value by almost a half a decibel relative to how it would have measured if the audio had ended at 3:38. Half a decibel’s loudness hike may not seem much, but in the land of loudness-hacking it’s actually pretty significant.

And if they can get away with it here, then maybe next time they’ll push it further. Doubling the length of that outro, for instance, would have dropped the measured loudness by another half a decibel. It’s a slippery slope…