The six days after the lesson are where the money goes

Six days after a Spanish lesson that cost me $12, I was trying to say my flight was delayed and I could not find retraso. I had written it down. I had said it out loud in the lesson, twice, and my tutor had nodded.

Gone anyway. And the part that stung: the next lesson opened with my tutor teaching me the same word again. I'd paid twice for it.

The unit nobody prices

Every page about whether tutors are worth it prices the lesson. Fifty minutes of a stranger's attention is easy to price, which is why everyone prices it. What you're buying is a set of words you didn't have before and will still have next month.

So the honest unit is lesson price divided by words retained: words you can still produce a week later, in a sentence, under a little pressure.

Which means I needed a denominator, so I went back through six weeks of lesson notes.

Across 6 lessons my tutor put an average of 14 new items in the chat window, words and set phrases. The ones I'd actually failed to produce mid-sentence ran 5 or 6 a lesson. That's the split I'll divide by, and it turns on the difference between recognizing a word and producing it: a word you understood instantly was never a purchase.

Call it 6 words a lesson that are genuinely new to your mouth. At $16 that's $2.67 a word if you keep all 6. Keep 2 and the same lesson cost you $8 a word. Same tutor, same invoice, a 3x spread in what you took home.

Six lessons is a small sample and it's mine, so treat the 14 as my handwriting. Count your own.

That spread lives outside the lesson, decided in the six days afterward by whether anything happened to those words.

What a lesson actually costs, with the numbers

To divide, you need a numerator. Here are the real ones, from the platforms' own pages.

italki publishes per-lesson bands: community tutors "typically charge between $4-$20 per lesson," professional teachers "typically charge between $10-$40 per lesson." Per lesson, not per hour. Almost every affiliate roundup quietly converts one into the other, whichever way makes the page's argument land.

Preply's own rate table, updated 2 March 2026 and labeled "Average Price on Preply," is per hour and says so: $16/hr for speaking practice, $17 for beginners, $20 for DELE prep, $21 for SIELE prep. Its stated range is $15 to $50. Preply raises the same caution I did: check whether the price buys an hour or 50 minutes.

And prices in a band aren't distributed evenly across it, which is why a bare range is close to useless. On 22 August 2026 I pulled the trial prices for the first 21 Spanish tutors italki showed me:

italki Spanish trial prices, 21 listings, 22 Aug 2026

$12

median trial price

0 of 21

priced under $10

0 of 21

priced at $15 or above

The middle is crowded and the tails are thin. Most people booking Spanish on italki see something near $12 for a trial, and more once they're a regular.

So I'll build the weekly figure in italki's own unit. A $12 trial implies $17 to $24 for a standard booking, depending on where in italki's discount band it sits. Call a regular weekly lesson $16, at the cheap end and inside italki's $4-$20 community band. Preply's $16 is the same digits by coincidence, and I'm not stacking the two: that one is per hour.

$16 a week is $832 a year. That's a fair price for a year of talking to a native speaker, and I think it's worth spending. The question is what fraction of it survives to next spring.

Why the sixth day is the expensive one

Hermann Ebbinghaus put a number on this in 1885, on himself, with nonsense syllables. He learned a list to perfect recall, waited, then measured the effort to relearn it. That measure is savings, and it's generous: it counts memory that won't surface on its own.

His savings figure after 6 days was 25%. Even by the flattering measure, three quarters of the original work was gone, on exactly the interval a weekly lesson leaves open.

The obvious objection: one German man, a pile of syllables, 1885. Murre and Dros re-ran it in 2015 with modern controls, and the curve came back nearly identical.

I wrote about the mechanism in why your brain forgets Tuesday's vocab by Friday. The budget version: a word you met once on Monday and never touched again is, by Sunday, mostly a feeling that you've heard it somewhere. Which was my state with retraso, with a human being waiting for me to finish my sentence.

Where the week goes

lesson day 1 day 3 day 6 next lesson

The fix is old and dull and very well evidenced: pull the word back out of your head a few times, spaced out, rather than rereading it. That's the testing effect. In Karpicke and Roediger's 2008 Swahili study, students who kept being tested recalled about 80% a week later. Students who kept restudying recalled about 36%.

The delay is what makes that gap visible. When Roediger and Karpicke ran the comparison on prose in 2006, the rereaders won after 5 minutes and lost badly after a week. Whatever you do on the evening of the lesson feels like it worked, because everything does.

Run that through the division. Take my 6 keepers from a $16 lesson. At the restudy rate you'd finish the week with 2, so $8 a word. At the retrieval rate you'd finish with 5, so $3.20. Identical lesson. The difference is four or five minutes a day spent pulling words back out instead of highlighting them.

So here is the shape I run, which fits in the gaps of a normal week.

Lesson night, 5 minutes. Write your 6 keepers as blanks: a sentence from the lesson with the word cut out, English meaning underneath. That's the only writing you do all week.

Day 1, day 3, day 6. Cover the answers and say each word out loud before you look, because a spoken guess can't be quietly revised into a pass once the answer is in front of you. Then check. Six words takes a minute.

Day 6 is the rep that counts, landing closest to the interval Ebbinghaus clocked at 25%, and it's your last chance to fix a word before your tutor spends paid lesson time re-teaching it. Skip the other two before you skip that one.

Missing a word costs you nothing. Kornell, Hays and Bjork found failed retrieval attempts still improve later memory, provided you see the answer afterward. What hurts is deciding you basically had it.

I'm mixing two things there, and I'd rather say so than hope you don't notice. The 80 and the 36 are lab percentages from a controlled Swahili list; my 6 is a count from my own notebook. Neither licenses the other, so what the arithmetic gives you is the shape of the gap rather than a figure you can hold me to. Same caveat on Ebbinghaus: nonsense syllables decay faster than real words that hook onto things you already know, so read the 25% as the shape too.

The claim I'll defend: that gap is bigger than the gap between a $10 tutor and a $20 one.

The best line on the internet's most popular page about this

The page currently ranking for this question is TruFluency's "Are Language Tutors Worth It?", a tutoring company writing about whether you should buy tutoring. It concludes yes, in the heading, before the evidence.

I want to give it credit anyway, because buried in its cons section is the sharpest sentence anyone has written about this problem. Inconsistent classes, it says, "will keep you paying, while relearning the same information."

That's the whole thesis, arrived at by instinct, stated in ten words, then abandoned in the next clause. The page names the waste mechanism and walks away.

Where it goes wrong is the section right after, titled "Language Tutors Can Be Expensive," where the expense gets resolved as a benefit: the cost, it argues, "may motivate you to not waste your own money." A reader who arrived worried about wasting money is told the waste is motivational. And on a page about whether a purchase is worth it, no quantity is ever divided by any other. Its only number is the size of the packages it sells.

Who should throw this arithmetic away

If you're booking lessons to get fluent rather than to acquire vocabulary, everything above is noise and you should ignore it.

TruFluency closes on a piano analogy, and on that it's right: you can't learn the instrument without playing it badly for a while. If your bottleneck is that you freeze, the 50 minutes is the product. Book more hours and stop doing arithmetic about it.

Two other people should skip this. If your tutor sets homework you actually do, do that instead, because the thing you'll do beats the thing you won't. And if you'll genuinely build your own cards, Anki is free on desktop and Android and better than paying anybody. Migaku, an Anki-adjacent company, puts card-building at roughly 30 seconds each, 12.5 hours for 1,500 cards. Be honest about whether you'll spend it.

For everyone else, I built Verbamor because I kept losing words like retraso. You record or type what came up in your lesson and it turns those words into fill-in-the-blank cards with native audio and an image, scheduled by FSRS.

The arithmetic, since arithmetic is the point of this page: Native is $129.99 a year, which against $832 of lessons is about 8 more lessons a year spent keeping the other 44. Explorer is $19.99 a month. iOS only, five languages, no free tier. It teaches you nothing new, and it doesn't find you a tutor.

Whatever you use, the test is the same. Six days after your next lesson, before you open your notes, write down every word that came up. Then check.

If the list is short, the problem was never the price of the lesson.

---

Sources

Every load-bearing claim Verbamor makes is traced to its paper on the research page.