Data quality
Same product. Different batches. Different numbers.
When you ICP-MS-test a packaged food, you are testing one bag, from one factory run, at one moment in time. The same brand on the same shelf can vary 10× or more between lots, depending on source country, harvest, drought, processing, and supply chain. A single test is real. It is also a snapshot of one batch.
So when two results on "the same product" disagree, the most common reason is they were not actually the same product, just the same brand in different bags. Sometimes it is also a chain-of-custody issue, when one of the labs received the sample after it had been opened and repacked. Either way, a single ICP-MS number is a snapshot of one lot, not a brand verdict. This page is the running list of cases in our database where the snapshots disagree, with both numbers side by side, the math, and what we still don't know.
1,293
food entries in our calculator
338
single-sample entries (sourced from one ICP-MS test, no replication)
2
contested pairs in our database (two different labs, same product, different result)
$100
total cost to nominate a batch for the Lead Lottery (one Purity Labs ICP-MS panel + one sealed retail unit). 4 to 5× cheaper than the same test through a single-tester GoFundMe.
First read. Parts per billion is not a dose.
Before any of these batch-variability cases makes sense, look at the math. The headline ppb number tells you almost nothing about how much lead reaches you in a day. Serving size does most of the work.
Two products with very different ppb numbers can deliver wildly different doses, in either direction. A "low" 1 ppb in tap water at 1 liter per day delivers 1 µg of lead. 690 ppb in black pepper at 0.4 grams per serving delivers 0.28 µg. Same logic, opposite direction, applies to every entry on the calculator. The cases below matter because the disputed batch numbers, IF accurate, would land in serious-dose territory at typical servings, and a single-batch ICP-MS result can't tell you whether the next bag from the same shelf is similar.
Canned tomato soup (FDA TDS, one cup)
Heavy serving + tiny ppb
20ppb Pb
×
240gper cup
=
4.8µg / cup
218% of the FDA child IRL (2.2 µg/day) on a single bowl. FDA Total Diet Study median value for canned condensed tomato soup. The number you don't think about does the most damage.
McCormick Black Pepper
Tiny serving + alarming ppb
690ppb Pb
×
0.4gper serving
=
0.28µg / serving
13% of the FDA child IRL. Headline ppb looks scary; actual dose is small because the serving is small.
Philadelphia Cream Cheese (LSM 120.6 ppb)
Why the contested cases matter
120.6ppb Pb
×
28g(1 oz / bagel)
=
3.4µg / bagel
155% of FDA child IRL per bagel IF the 120.6 ppb is accurate. Our retest of the same product came back nondetect. That gap, between "yes a public-health event" and "ordinary cream cheese," is exactly why methodology matters.
Black pepper looks scary because 690 ppb is a loud number. Canned soup at 20 ppb sounds harmless. The math runs the other way: the soup delivers 17× more lead per serving than the pepper does, because you eat 600× more soup than pepper by weight. Most lead exposure is invisible because it's spread across the things you don't think about. The contested cream cheese number is in a bad zone because cream cheese has a normal serving and the disputed ppb is high enough to clear the FDA child reference level on a single bagel. That's why we need to know whether the number is right.
Case 1. Philadelphia Original Cream Cheese.
When the gap between two numbers is bigger than batch variability can explain. Cream cheese is pasteurized milk from US dairies, not a high-variance commodity. So when one lab reports 120 ppb and another reports nondetect, lot variation is unlikely to be the answer.
In April 2025, Lead Safe Mama sent a sample of Philadelphia Original Cream Cheese to Purity Laboratories for ICP-MS analysis. The reported result was 120.6 ppb lead, published with a graphic citing 2,412% of the proposed action level for kids. We sent an unopened block of the same product (same retail format, similar timeframe) to Purity Labs for ICP-MS. Our result was nondetect, below 2 ppb.
Two batches of the same cream cheese from US dairies should not produce results 60× apart. Dairy is a relatively homogeneous matrix compared to cereal, spices, or root vegetables. So this case is one where ordinary batch variability is not a sufficient explanation, and chain-of-custody differences (sample anonymization, transfer, repacking before mailing) become the more likely story.
Philadelphia Original Cream Cheese · Two batches, two answers
LSM · 1 sample
120.6 ppb Pb
DetectLead · 1 sample
ND < 2 ppb Pb
What we think happened
Both numbers cannot describe the same cream cheese. The Purity Labs report carries LSM's standard footnote disclosing that products are opened in the LSM kitchen, transferred into a different container, relabeled with an internal lot number, and then mailed to the lab. Every step in that chain introduces lead-contamination opportunity at the sub-ppb scale. The FDA Total Diet Study and our own retest, both with intact retail packaging, both came back nondetect. The most likely explanation is contamination introduced in the home-kitchen transfer, not in the cream cheese.
We are not asserting bad faith. ICP-MS is sensitive enough that a few micrograms of household dust can shift a 28-gram sample from nondetect to 120 ppb. That's why accredited heavy-metals labs require samples to arrive in sealed retail packaging. The methodology is what we contest, not the lab.
Case 2. 365 Whole Foods Organic Morning O's.
This is what batch variability looks like in a high-variance matrix. Cereal grains absorb cadmium and lead from soil at uneven rates depending on farm, harvest year, drought, fertilizer, and rotation. Same brand. Different lot. Different number. Even the original tester saw it.
In September 2024, Lead Safe Mama published a single-sample ICP-MS result of 362 ppb lead on 365 Whole Foods Market Organic Morning O's, with cadmium and arsenic also detected. A separate group of community members independently funded a retest of a different box of the same product. Here is what Lead Safe Mama said about the result, verbatim from the post:
"Another member of the Lead Safe Mama community had a different box of this product tested (independently, raising funds for the testing from a group of parents) and their test results were different (showing positives for Cadmium, as anticipated). In both cases, the testing results showed levels indicative that the product is not safe for young children to consume, and while the exact numbers between cereal boxes were different (for whatever reason, batch specific concerns or otherwise), the overall outcome was the same." Lead Safe Mama, Oct 2024 post (last updated Dec 10, 2024).
And in the same post, on what she planned to do next:
"Our intention with this product is to test it again (from another batch) to see what the range of test results might be."
So we are not adding the framing here. Lead Safe Mama acknowledges that two boxes of the same product produced different exact numbers, and that the next step is a third box to map "the range of test results." That is batch-to-batch variability, in her own words. The 362 ppb headline number remains in circulation as if it were the brand value. It is one box's value.
The retest's actual numbers (lead, cadmium, arsenic in ppb) were not published in the post or in any indexed follow-up we could find. The community retest funded the work but did not result in public numeric values, only the qualitative description above. So the headline number circulates and the retest does not.
365 Organic Morning O's · Two boxes, two outcomes, one published number
LSM box 1 · published
362 ppb Pb
Box 2 · community-funded retest
numbers withheld
Status
We display the 362 ppb number on the calculator with a tooltip pointing here. If you have a sealed box, mail it (eric@detectlead.com) and we will run the panel at Purity Labs on our dime and publish the result side by side with the original. If the retest stays elevated, it stays elevated. If it comes back nondetect, that's a public correction worth making.
Full disclosure on the framing.
I bought a $50,000 GFAAS (graphite-furnace atomic absorption) instrument in summer 2024 to do food testing in-house. Two months later I had walked away from the method. Wet-acid digestion is finicky in ways that are not about how careful you are. Pure trace-metal-grade nitric acid runs $1,000 per liter. Acid-washed vessels have to be kept fingerprint-free. A speck of household dust contaminates a 28-gram digest at sub-ppb levels. And a single slice of bread is not homogeneous, the lead is concentrated unevenly, so two pinches from the same loaf give different numbers before the lab even sees the sample. More on my food testing journey →
Case 3. Raw Baby Carrot. The FDA actually ran this experiment.
Most consumer-product testing programs publish one number per item. The FDA ran the same item 27 times. Here is what that looks like for one product.
The tooltip on the Raw Baby Carrot entry in the calculator reads, verbatim from the FDA monitoring data we sourced:
"FDA tested 27 samples of Raw Baby Carrot for lead and found 23 had no lead, 4 had lead. The highest concentration of lead found was 36 ppb and the average across all samples was 4.9 ppb. USDA Serving Size: 85 g (2020)."
Same product. Same brand category. Same lab method. Across 27 different bags, the lead values ranged from nondetect on 23 of them to 36 ppb on the worst single bag. The "average" of 4.9 ppb is the number that ends up on the calculator. It is also a number that does not represent any actual carrot you would buy. 85% of the time you would test ND. Roughly 1 in 7 bags would test elevated. None of them are "the carrot."
This is what batch variability actually looks like in real food when somebody bothers to run the experiment. It is exactly why a single-bag ICP-MS result on any packaged food is not a brand verdict.
If you ran the test on a typical bag
23 of 27 bags · ND
ND< 1 ppb
×
85g1 serving
=
0.0µg
Headline if a single-test program ran this bag: "FDA Carrots Are Lead-Free."
The number on the calculator
Average of all 27
4.9ppb avg
×
85g1 serving
=
0.42µg
19% of FDA child IRL. The averaged number hides everything that matters.
If you ran the test on the bad bag
1 of 27 bags · the worst
36ppb Pb
×
85g1 serving
=
3.06µg
139% of FDA child IRL. Headline if this was the only bag run: "FDA Carrots Test Positive for 1,800% of Action Level."
The takeaway
If a single-sample program had pulled the worst bag, the headline would be "FDA Carrots Test Positive for 36 ppb of Lead." If the same program had pulled any of the 23 nondetect bags, the headline would be "FDA Carrots Are Lead-Free." Same product. Same year. Same store. Different bags. The single-sample number is real, but it tells you about the bag the lab tested. It does not tell you about the next one.
This is the entire reason we keep this page. Not because labs make mistakes (they mostly don't), but because food doesn't have one number, and reporting it as if it does makes the public worse at thinking about food risk, not better.
What a single batch can't tell you.
338 of 1,293 entries in our calculator come from a single ICP-MS test, mostly via Lead Safe Mama. The lab method is fine. The single-batch sampling plan is the weak link. One test result tells you what was in the specific bag that reached the lab. It does not tell you:
- Whether the next bag from the same shelf is similar.
- Whether the brand is "safe" or "unsafe" in general.
- What the distribution of levels looks like across lots, source countries, harvests, or seasons.
- Whether the tested batch is typical or an outlier.
- Whether a different lab, or a lab that received the original sealed package, would find the same number.
Brand consistency is a marketing claim, not a chemistry one. Salt is mined in different mines. Cereal grain is grown in different fields. Spices are sourced quarterly from whichever country has the best price that quarter. The FDA Total Diet Study has been running since 1961 and reports food heavy metal results as distributions across multi-region multi-year purchases, with means, 75th percentile, and 90th percentile values. That's how you ask a single-batch question and not get a single-batch answer. We use TDS values where available. For everything else, we tag confidence honestly.
How we tag confidence.
Every entry on the calculator gets one of these four labels in its tooltip. We are still backfilling the tags across the database. If you see one missing, email and we'll fix it.
| Tier | Source pattern | What it means |
|---|---|---|
| A | FDA Total Diet Study, USDA monitoring, multi-year multi-region federal data | Multi-sample, multi-region, distribution-aware. The most defensible numbers we have. |
| B | Multi-source independent agreement, or our own ICP-MS panel where two or more independent labs agree | Reproducibility met. Different lab, similar number. Still subject to lot variability. |
| C | Single sample, no contesting result. Most LSM-sourced rows fall here. | Directional only. Treat as a flag for follow-up, not a definitive brand verdict. |
| D | Contested. A retest disagrees, or the original tester acknowledged the value as unexpected. | Hold. We display the result, but the tooltip points here. Send us a sealed sample if you have one. |
Want a specific batch retested? Nominate it for the Lead Lottery.
This is our crowdfunded retest model, and it's a direct alternative to the GoFundMe campaigns where a single tester charges 4 to 5× the actual lab cost to run one bag. Anyone can nominate a product. When chip-ins reach $100 (the actual cost of one Purity Labs ICP-MS panel plus a sealed retail unit), we buy the unit, ship it sealed, and publish the result side by side with whatever number is currently on record. No anonymization, no markup, no theatrics.
The Lead Lottery →Or email eric@detectlead.com · 631-461-1838.
What you bring home from work.
Take-home lead is the secondary-vector your kid ends up wearing on their hands.
Stained glass, fishing weights, ammunition reloading, antique restoration, battery work, and construction on pre-1978 homes all involve direct lead handling. Boots off at the door, work clothes laundered separately, no eating in the work area. Lead dust is the ride-along.
Free tools we maintain.
All four are free, no signup. Built and updated by us. Anyone, anytime, in your kitchen or your house or wherever you happen to be.
Baby food lot lookup
18,124 baby food lots indexed by brand, flavor, and lot number. Search and find your specific batch.
Prenatal supplement guide
Which prenatals tested clean. Which didn't. The math, the brand list, the alternatives.
Universal food calculator
1,343 foods. Pick what you eat. Get your µg/day. The only number that matters for daily exposure.
Environmental exposure screener
Pre-1978 housing density, soil lead, water risk. Type your ZIP, see the local picture.
"Lead is bad" — the primary sources.
- CDC: There is no safe blood lead level for children.
- FDA: Closer to Zero action plan — interim reference level 2.2 µg/day for children, 8.8 µg/day for adults.
- EPA: EPA drinking water lead guidance; SDWA "lead-free" definition was tightened to 0.25% in 2014.
- AAP: Prevention of Childhood Lead Toxicity (2016 policy statement).
- CDC: Blood lead reference value — 3.5 µg/dL since 2021.
Find it. Deal with it. Don't let lead weigh you down.