HPLC says 98.7% but independent lab shows 1.4% shoulder

A 1.4% shoulder at 11.3 min is hard to hand-wave. Did both labs use the same wavelength? UV 214 vs 220 can change area a lot.

That’s what I thought to.o My last COA gap was 99.1% vs 97.6%; source method didn’t baseline-resolve the impurity.

99.1 vs 97.6 is a big gap. Was the impurity visible in both chromatograms or only one?

If only one chromatogram shows it, raw data matters more than the COA PDF. The 11.3 min shoulder might be integration artifact. Different threshold, 1.4% can appear or vanish.

Retest in two weeks. If the shoulder grows, degradation. If it stays 1.4%, method. Been burned by one sample before.

I've been down this rabbit hole. The number on a COA is only as good as the method behind it. If the source lab uses a short gradient and the independent lab uses a longer one, you can easily see a 1-2% difference on the same sample. What matters is whether the impurity is consistent and identified. Without LC-MS, you're just comparing two sets of peaks. I'd request the full sequence, blank, and duplicate injection from both labs before deciding anything. If their duplicates disagree by more than 0.5%, the method is the problem, not the sample.

An 11.3 min shoulder would send most people down a rabbit hole, but the one 1.9% mystery peak that's ever turned up traced back to a vial that had been living in a car console for 10 days — same lot, and a fresh vial from that same lot came back clean. Weekly waist and weight averages didn't move either way.

The tracking that actually tells people something tends to be Sunday morning waist, sleep hours, and restaurant meals, not COA decimals. The only two weeks where waist ticked up were 4+ takeout meals and under 6 hours of sleep — not the weeks with messier lab traces.

Common surprise here: the boring explanation usually wins. Did the sealed vial from that same lot get run, or just the six-week sample? If it's only the old one, heat and storage is a way more boring explanation than method variance.

Nineteen days of flat scale against the exact same batch number is what finally killed my COA spreadsheet habit — I spent weeks convinced a "bad" batch was sabotaging me when the actual pattern was that my rough weeks lined up with a work deadline and takeout, not with peak area. Over 14 weeks my waist went 36.5 to 33 inches while the scale sat dead still through week 9, then dumped 2.8 lbs in four days once I got back to 7 hours of sleep instead of 5.5 and stopped eating restaurant food at 9pm. So I'd gently push back on the framing: chasing the numbers on the COA gave me nothing actionable, and the boring log of sleep, sodium, and step count gave me everything. Genuinely curious what your six weeks of tracking shows — does the shoulder sample week actually look different from the other five in how you slept or ate, or is the weird week just the one where life got messy?

My cheap fabric tape has been a better bullshit detector than any spreadsheet I’ve kept—waist at the navel every Sunday before coffee, and over seven weeks it went 38.5 to 36.9 inches while my weight only moved 4.2 lb. The surprise was a 5-hour-sleep week that added 0.6 inch even with an identical food log, which taught me to check context before freaking out over one odd reading. If I were you, I’d freeze the test-result spiral for a second and log sleep, sodium, and weekend eating for 14 days—my bet is you’ll see more signal there than in a single weird number. Did anything change in your routine during those six weeks, or was it same food/sleep/stress the whole time?

Honestly, cOAs are just a lab's oinion until someone else reruns it. Method changes, shoulders appear.

One thing I would ask the independent lab: did they run a blank and a forced-degradation control alongside your sample? A 1.4% shoulder at 11.3 minutes is small, and without a blank it is hard to know whether it is sample-related or just buffer and column noise. Also, was your six-week aliquot stored the same way throughout? I track weekly averages, and I only trust

For what it's worth, did the independent lab do any identity work on that 11.3 min shoulder, or is it UV area only? A 1.4% shoulder with a different mobile phase can also be a method artifact

Since you’ve been tracking this sample for six weeks—do you have an earlier chromatogram from the same batch or vial? If that 11.3 min shoulder stays ~1.4% across time points, it’s probably a process impurity, not degradation. If it grows, that’s a different story. Also, did the lab run a duplicate injection? A shoulder that shifts with injection volume or concentration can be a method artifact. I’d ask for the overlay before reading too much into one run.

The 1.4% shoulder at 11.3 min—did they run a load-dependent injection? If it collapses at 5 µL but grows at 20 µL, suspect solvent mismatch or column overload, not an impurity. Ask for PDA spectra and RRF: area% assumes identical extinction coefficients, and unknowns rarely cooperate. I’ve had a “degradant” vanish once diluent matched mobile phase. If it’s constant across loads and has a distinct UV spectrum, treat it as real. Did they report mass balance or just area%?

That shoulder would bug me too—sending calm-lab-vibes for your next run! 🤞

One thing I’d add: did the independent lab run PDA/UV peak purity across that main peak? A shoulder can look scary, but if

My trendline says that shoulder is begging for its own spreadsheet tab.

I’d ask for the relative retention time of that 11.3 min shoulder vs the main peak. My own logs separate RRT from area%; a stable RRT ~1.05 has usually meant a related species, while a drifting RRT is what makes me suspicious of on-column degradation. Also, did the independent lab run a blank/vehicle injection with the same mobile phase? Different mobile phase pH can shift shoulders independently of purity. If their main peak resolution is <1.5, that 97.8% number gets fuzzy fast.