Corrections
10 corrections · newest first
When we get a fact wrong, we fix it on the page where you read it, and we write it down here. We do not quietly edit and move on.
This page lists things we said that were not true, and things we showed as real that were not. It does not list spelling fixes or rewording.
Shahmaran is written by one person with the help of AI, and every module is read by a human before it goes up. That catches a lot. It does not catch everything, and this page is the honest record of what it missed. If you find something wrong, tell us — you will end up on this page, credited unless you would rather not be.
· How interest and hidden fees quietly drain you — subscriptions
We said: “In a widely-cited survey, people spent about $133 a month on subscriptions but first guessed a much smaller number.” Elsewhere we said $133 and $219 were totals from “different surveys”.
What is true: $133 is not what anyone spent. It is the gap. One survey found people guessed about $86 a month and were really paying about $219 — $133 more than they thought. Both numbers come from the same survey.
Why it happened: A number about a difference was written as a number about a total. That is the same mistake as the Dunning–Kruger one below, in a different subject — which is why we now keep a list of the statistics that break this way.
Found by Our own fact-check · C+R Research, subscription service statistics
· How interest and hidden fees quietly drain you — your card bill
We said: A sentence about the three-year payoff box, printed as a quote from the Consumer Financial Protection Bureau.
What is true: The CFPB never wrote it. It was a reworded version of the title of one of their pages — and that title is a member of the public asking a question, not the agency answering one. The quote is gone. The fact it carried was right, and is now said plainly in our own words.
Why it happened: A page title was mistaken for a statement by the organisation whose page it was.
Found by Our own fact-check
· How interest and hidden fees quietly drain you — small accounts
We said: “Nearly 40% of U.S. households keep less than $400 in easy cash.”
What is true: The Federal Reserve figure counts something else: about 4 in 10 U.S. adults say they could not cover a surprise $400 bill from cash or savings. It is adults, not households, and it is about covering a bill, not about how much cash is in the account.
Why it happened: The number was real and the denominator was wrong — households substituted for adults, and “could not cover a $400 expense” turned into “keeps less than $400”.
Found by Our own fact-check · Federal Reserve, Economic Well-Being of U.S. Households
· Learning how to learn — mixing up practice
We said: “Two classes of 7th graders practiced math. One blocked their practice. One mixed it up.” Then: “The mixed group scored 72%. The blocked group scored 38%.”
What is true: There were no two groups. The same students did both — they mixed the practice for some kinds of problems and blocked it for others, then took one test covering both. The scores are right. The setup was not.
Why it happened: A study where every student does both conditions was described as two sets of children being compared. It makes the finding sound like these kids beat those kids, when it is actually one way of practising beating another inside the same child.
Found by Our own fact-check · Rohrer, Dedrick & Burgess 2014
· Learning how to learn — what the numbers show
We said: “Blocked practice scored 37%.”
What is true: The study reports 38%. The figure 37% does not appear in it.
Why it happened: A number was copied down wrong. Nothing more interesting than that.
Found by Our own fact-check · Rohrer, Dedrick, Hartwig & Cheung 2020
· Learning how to learn — sleep
We said: “People who skipped the nap showed no gain at all.”
What is true: We could not find a study that says this, and it is stronger than what nap research generally finds. The sentence is removed. The screen still says people who napped after learning remembered more, which is supported.
Why it happened: A claim with no source attached to it survived because the sentence beside it was true.
Found by Our own fact-check
· Reading a food label without being tricked — serving sizes
We said: “One study measured ready-to-eat foods. Serving sizes ran from 18 grams up to 100 grams.”
What is true: That was 1,070 products in one supermarket in Melbourne, Australia, under Australian rules. U.S. rules set a reference amount for each kind of food, so the spread an American shopper meets is smaller. Two foods still rarely use the same serving, so the advice is unchanged.
Why it happened: A finding from one shop in one country was written as a fact about food labels generally.
Found by Our own fact-check · Serving size study, PMC
· Cognitive biases — the Dunning–Kruger screens
We said: That people who know the least “end up sounding most certain”, that beginners “often feel very sure”, and that with more learning “that sureness drops”.
What is true: The study found the opposite about confidence: the top scorers rated themselves higher than the bottom scorers did. What it measured was not who feels most sure — it was how far a person's guess about themselves sits from the truth. That gap is widest at the bottom. People in the lowest quarter scored in the 10th percentile and guessed they were in the 67th.
Why it happened: A number about the size of an error was rewritten as a number about the level of confidence. Two different things, and the swap made the screens say something the research does not.
Found by Our own fact-check · Kruger & Dunning 1999, Journal of Personality and Social Psychology
· The home screen, signed out
We said: “12,480 practicing”, “384 completed today”, and a “Trending Financial +18%” row.
What is true: None of those numbers were real. They came from a design mock and were never connected to anything. They are gone, replaced by counts of what is actually in the app — modules published, areas, total sections — worked out fresh each time the page loads.
Why it happened: Placeholder figures from a prototype were built into the real screen and then not removed. Nothing distinguished them from live data, on screen or in the code.
Found by Nika
· The welcome screen, first run
We said: That the wellness check was “Eight questions”.
What is true: The quick wellness check is sixteen questions — two for each of the eight areas. The screen now works the number out from the questions themselves, so it cannot drift again.
Why it happened: The count was typed into the copy by hand instead of being read from the question set, so when one changed the other did not.
Found by Our own review