In October 2001 the Good Housekeeping Institute published a short report about a product of ours. It was headlined “Low-Cal Muffins: Too Good To Be True?” and the answer, as the magazine put it, was You bet.
They were right. What follows is how it happened, because the mechanism is more interesting than the mistake, and because almost nobody who makes a packaged food is safe from it.
The product
We bought The Poi Company in May 1997. At the time its only product was poi. Poi is about as simple as a manufactured food gets, and in Hawaiʻi it is legally required to stay that way: by statute, poi is taro and water. That is the whole permitted ingredient list. When we later made a squeeze-pack poi snack with flavourings — chai and others — we needed specific approval to call it what it was, because adding anything at all takes a product outside the definition.
Poi as an ingredient, though, was open country. So we put it into baked goods: English muffins, bagels, cheesecake. The muffins were made for us by Hawaiʻi Star Bakery, who guarded their process closely. We sold them from 1999. I never once saw them produced.
There is a wrinkle worth recording here, because it explains why the muffins existed in the form they did. People could not taste the poi unless they could see it. Given a pale muffin, tasters reported nothing in particular; given a faintly purple one, the same tasters found the poi immediately. So the products were coloured — Warner-Jenkinson purple, the standard commercial carmine. Carmine is made from an insect. Our poi English muffins were, therefore, quietly not vegetarian, which is not a sentence I expected to write about a taro product.
The analysis
A packaged food sold at retail needs a Nutrition Facts panel, and the numbers on it have to come from somewhere. We were not food chemists. We did not make the muffins. So we did what any small manufacturer does: we paid a professional laboratory.
In 1999 our colleague David Reed sent samples to Food Products Laboratory, Inc. of Portland, Oregon — since absorbed into Exova and now part of Element Materials Technology. His transmittal letter said what he was sending and how it should be treated. It discussed a serving size of half a muffin, thirty-five grams, twelve servings per container. Nobody reading it could have thought it was about anything other than muffins.
What came back was a report headed Poi.
I noticed the discrepancy and telephoned in August 2000 to ask why the cover letter said “Poi English Muffin” and the printed report said “Poi.” I was told the description had simply been abbreviated on the printed report. I asked whether they were certain it was the muffins that had been analysed. I was told yes. They offered, unprompted, to reissue the report with the correct product name, and did so.
The number on that report became the number on the package: seventy calories, eighteen grams of carbohydrate.
We printed a great many packages.
What the panel actually said
Here is what we printed, on every bag:
Nutrition Facts. Serving Size 3.21 oz. (90g). Servings Per Container 6.
Calories 70. Calories from Fat 0.
Total Fat 0g. Sodium 25mg. Potassium 150mg. Total Carbohydrate 18g. Dietary Fiber 2g. Protein 0g.
Read it again with fresh eyes and it convicts itself. A ninety-gram muffin made from wheat flour cannot contain zero grams of protein. Flour is roughly a tenth protein by weight; a muffin that size holds something like eight grams of it. Zero is not a rounding error, it is a different food.
And it is a food we can identify. Ninety grams of poi is about eighteen grams of carbohydrate, negligible fat, negligible protein and a useful amount of potassium. Every figure in that box is poi. Not approximately poi — poi.
Which meant that for two years the answer was sitting on the side of every bag we shipped, legible to anyone who happened to wonder how a wheat product could contain no protein at all. Nobody wondered. Nobody reads the nutrition panel on a muffin.
Then we made ourselves visible
In October 2000, following a meeting on the mainland, I faxed Beth Haas at Weight Watchers a photocopy of the muffin bag and an explanation of what taro does in baked goods — it is high in amylopectin, which is why poi keeps a crumb moist. Sending information to an editor who has asked to see it is ordinary trade practice. We were not buying anything.
In its January/February 2001 issue, Weight Watchers magazine described the muffins as “a speedy, low-Points breakfast” and called them “moist & flavorful.” As Good Housekeeping later reported, that “sent dieters to the company’s Web site to place orders, sometimes by the case.”
It did more than that. On the WeightWatchers.com community boards, members began hunting the muffins down for each other — posting which Whole Foods in which Michigan county had them in stock, who had bagged fourteen at a time, which stores would order them in. People were driving across counties for our muffins on the strength of a Points count.
Then we advertised, and the order matters. We would never have gone hunting for a mainland magazine to promote a Hawaiʻi product — the freight alone makes that a poor bet. We bought the page because Weight Watchers had already begun promoting the muffins for us, and the demand was visibly there. In the July/August 2001 issue of Weight Watchers magazine, page 46, we ran a third-page advertisement:
Yummy. Double Yummy.
Poi English Muffins — “Moist & flavorful” — Weight Watchers Magazine, Jan/Feb 2001
Taste it for yourself!
We paid Weight Watchers Publishing Group $3,718.75 for it, net of agency discount, on invoice 1500565. We were quoting the magazine’s own review back to the magazine’s own readers, in the magazine, to sell them a product whose calorie claim was wrong by a factor of two.
In fairness to us, the advertisement itself made no calorie claim. It sold flavour. But nobody in that readership was buying an English muffin for the flavour alone, and we knew exactly why they were coming.
The advertisement ran in July and August. Good Housekeeping’s fax arrived on 7 September.
Good Housekeeping buys two bags
The Good Housekeeping Institute heard about the muffins the same way everyone else did, and set up a taste test. The muffins, they wrote, “won rave reviews — until our staff noted that the other ingredients were similar to those in ordinary muffins, calling the calorie count into question.”
That is a very good piece of noticing. Nothing was wrong with the muffins. What was wrong was the arithmetic: an ordinary muffin made of ordinary flour cannot contain half the calories of an ordinary muffin. The claim was implausible on its face to anyone who thought about it for thirty seconds, and for two years nobody had.
So they bought two bags and sent them to an independent laboratory of their own choosing. The result:
“GHI sent two bags to an independent lab, which found that each muffin actually contains about 160 calories — more than twice the label claim. Instead of 18 grams of carbohydrates per muffin, each has more than 30 grams, the equivalent of two slices of bread. So if you’re dieting, cut each muffin in half.”
They faxed us the piece on 7 September 2001, weeks before the October issue reached newsstands. They were under no obligation whatsoever to do that.
What had actually happened
The panel printed on our muffin bags was the nutrition panel for poi. Not a miscalculation of the muffins — an analysis of an entirely different food.
Confronted with this, the laboratory’s position was that they had been sent poi, and had analysed what they received. Our position was and is that the transmittal letter said muffins, discussed muffins, and specified a serving size of half a muffin.
Both things cannot be true, but it barely matters which is, because the outcome is identical either way: a poi panel on a muffin bag is wrong regardless of who put it there. What I could not accept was the suggestion that the error was ours to have caught. From my letter of 10 September 2001:
“Surely your laboratory knows the difference between a muffin and something that is so obviously not a muffin. Why didn’t someone write to David to say, ‘You forgot to send us the muffins, David. What do you want us to do?’”
And, on the proposal that we should have detected the error by consulting the Federal Register and a reference work called Bowes and Church:
“We are not food experts, and we made this clear when we personally visited your lab in 1999. We do not manufacture the Poi English Muffins, and have never done so. Although we have sold this product since 1999, I have never seen it produced.”
Consider what that proposal actually asks. We had engaged a laboratory because we could not do the analysis ourselves. Their remedy was that we should have independently verified their analysis — by consulting the Federal Register and a food-composition reference — which is to say we should have possessed the very expertise we were paying them for. If we had had it, we would not have needed them.
There was also a simple logical point that I still think is unanswerable. We had no reason to commission an analysis of poi. We had bought the company in 1997; its poi had been analysed years earlier by a different laboratory and we were still using those figures on our poi bags. Poi is cooked taro and water, and by law it cannot be anything else. Nobody pays to re-analyse a two-ingredient product whose recipe is fixed by statute.
As I put it at the time, less temperately: there is no likelihood of confusing sticky, gooey poi with a muffin.
The other side of it
Twenty-five years later I will put the case against us as well, because it exists.
We were financial executives. We bought a poi company because it was a business, not because we understood food chemistry, and we never pretended otherwise. But David Reed, our operations manager at the time, came out of the food industry. In retrospect he should have caught it. A ninety-gram baked good containing no protein at all ought to have stopped somebody with that background in his tracks.
What I would say in his defence, and in ours, is this: we had paid a food laboratory on the mainland precisely so that nobody at our end had to be the expert. We FedExed multiple packages to them so they could check and double-check. They were the experts. They had every control a laboratory is supposed to have — except a control on correctly labelling the analysis before it went out of the door.
And that is the failure entire. Every measurement they made may well have been flawless. It did not matter, because flawless measurements of the wrong food are not a small error. They are worse than no measurement at all, because they arrive looking exactly like a fact, and facts get printed.
None of which shifts the fault, and I want to be exact about why. A missed backstop is not the same thing as the error. Reed might have caught it; they made it. Those are two different failures, and only one of them was being invoiced for. An expert who is paid for expertise does not get to argue that the customer should have supplied it.
What it cost
We scrapped the entire stock of printed packaging and had it printed again. That was the direct cost, and it was substantial for a company of our size. Set beside it the $3,718.75 spent advertising the claim six weeks before it collapsed.
The indirect cost was worse and unrecoverable. A national magazine had told several million readers that our headline claim was false, and a diet magazine’s readers — the people most injured by a wrong calorie count, and the ones who had ordered by the case — had been buying on the strength of it.
We pursued a claim against the laboratory, through their professional indemnity insurer, and instructed an attorney. We recovered nothing.
The part worth taking away
Two things, and neither is about poi.
You can buy expertise. You cannot buy away the liability for it. We paid a professional laboratory precisely because we lacked the competence to do the work ourselves — which is the entire reason such laboratories exist. When the work was wrong, it was our name on the package, our claim in the magazine, and our customers who had been eating two slices of bread while counting Points. The lab’s error became, in every way that mattered commercially, our error. If you outsource a number, you still own it.
An error in printed packaging is dormant until you succeed. It costs nothing and harms nobody while the product is obscure. The moment the product does well enough to be noticed, every claim on it is audited by strangers — and the more successful you are, the better resourced those strangers are. We were caught not because we grew careless but because we grew visible. If the muffins had sold modestly, that panel would probably still be wrong today, and nobody would ever have known.
Which is the uncomfortable version of the lesson. Getting away with it is not the same as being right, and you generally find out which one you were at the worst possible moment.
A word for Good Housekeeping
It would be easy to tell this story with the magazine as the antagonist. It was not.
The Good Housekeeping Institute did, at its own expense, the thing we had paid a laboratory to do and had not received: it tested the actual product and reported the actual number. It gave us advance sight of the piece. Its report is why the packaging was corrected, and why people counting calories stopped being misled by us.
They did us a favour, and they did Weight Watchers’ readers a considerably larger one. That the favour was delivered in a national magazine under the headline Too Good To Be True? is simply the price of having been wrong in public.
Sources: Good Housekeeping Institute Report, October 2001, p. 18; Weight Watchers magazine, January/February 2001; correspondence between The Poi Company, Inc. and Food Products Laboratory, Inc., September–October 2001, held in the company’s files.
Also in this series