It said 8%. Why did you end up with 3%?

Someone asked me a very practical question: I put USDT in flexible savings for a year, the page showed around 8% the whole time, and I ended up with a bit over three. Did the platform short me? It did not. Everything missing is inside the definition of the word “annualised”.

APR and APY are not the same word

The two abbreviations get used interchangeably and describe different things.

• APR is simple interest: take the current daily or weekly rate, multiply by the number of periods in a year, ignore any earnings on earnings.
• APY is compound: assume every payout is reinvested and keeps earning, and compute the full-year result.

For the same underlying rate, APY is always at least APR, and the gap grows with both the rate and the compounding frequency:

Pure arithmetic at 365 daily periods, unrelated to any specific product. At low rates the two are nearly identical; at high rates they diverge wildly — which is why high-yield products prefer to quote APY.

The practical takeaway: when two products show similar numbers, check they are quoting the same measure. One APY against one APR can mean the better-looking one is worse.

The biggest layer: it is a projection, not a promise

Here is what the page is really computing:

It takes the rate right now and assumes it persists for twelve months. In crypto savings that assumption essentially never holds:

• Flexible rates are set by borrowing demand and move daily. Hot market, lots of leverage, borrowing demand up, rate up. Market cools, rate follows it down. The 8% you saw might have lasted a few weeks.
• Promotional rates expire. New products often launch with a subsidy; when it stops, the rate returns to normal. You may simply have looked during the subsidy.
• Tiering makes the displayed rate different from your rate. Many flexible products pay a high rate on a first slice and much less above it. The page usually shows the top tier.

Which answers the opening question: the platform did not short anyone. The rate simply was not 8% for most of the year. Seeing that clearly requires watching what actually lands rather than what the page displays. A personal ledger makes that comparison possible.

Second layer: accrual start and end

Almost irrelevant on long holds, and capable of eating most of the return on short ones.

Common variations: the subscription day does not accrue, the redemption day does not accrue, interest starts the following day, weekends and holidays are treated specially. Each product writes its own.

A hypothetical to show the size of it: a seven-day product quoting 10% annualised, where neither the subscription nor the redemption day accrues, actually accrues five days. You realise roughly 10% × 5 ÷ 7 ≈ 7.1% — nearly a third gone. The shorter the term, the heavier those two days weigh.

Third layer: which asset the yield is paid in

The layer most often ignored entirely, and capable of outweighing all the others combined.

If a product quotes 20% but pays in a token you have never heard of, that 20% is denominated in that token. Its real value depends on two things: what the token is worth now, and what it is worth when you sell.

The reward asset determines how your exposure moves. This column is missing from most product comparisons and has to be dug out of the product description.

In the third case the credibility of the quoted rate depends entirely on that token's liquidity. If the book is thin, realising it may cost meaningful slippage. Which coin the yield is paid in and where the risk hides in high-APR products take the two angles.

Fourth layer: fees and minimums

Only lethal for small balances, but genuinely lethal for them.

Fixed costs to account for include deposit or transfer fees, any subscription or redemption fee, and the withdrawal fee when you eventually take the money out. Many of these are flat amounts rather than percentages, so the smaller the principal, the larger the share they consume.

A rough test: put your expected annual return next to the cost of one withdrawal. If a year of earnings does not comfortably exceed a single withdrawal, the exercise does not make economic sense. Working out where fees eat the yield gives a method you can run yourself.

Stacking the four

A hypothetical to close. Say the page shows 10% APR. Then:

1. The rate was not constant; the weighted average over the year was more like 6%;
2. You subscribed in several tranches and accrual rules cost you a few days, so call it 5.8%;
3. The yield was paid in the deposited asset, and that asset fell, so in fiat terms it shrank;
4. The balance was modest and a withdrawal fee took another bite.

You look at the statement and see three-point-something. Every layer was disclosed and compliant; together they are the gap between the page and the account.

How to use the number

Treat an annualised rate as a comparison tool, never as expected income:

• Comparing two products at the same moment on the same measure — useful.
• Projecting how much you will have in a year — essentially meaningless.
• What is worth recording is what actually arrives: how much, in what asset, at what price. A few months of that will tell you more about your own setup than any published rate.

Working the real rate out backwards

Everything above explains why the page is unreliable. Doing it in reverse is more useful: derive your real rate from the money that actually landed.

• Total received — add up every payout. If the yield was not paid in your base currency, convert at the price on the day it arrived, not today's price, or you will mix price movement into the return.
• Average principal — weight by day if you added or withdrew. The midpoint of opening and closing balances is adequate for a rough read.
• Actual days — calendar days from the first accrual to the last.

The result is the only rate that is genuinely yours: it already contains the rate movement, the accrual rules, the reward asset and whether or not you compounded. Comparing it against the page figure gives you the discount to apply mentally to every advertised number you meet.

Comparing two products fairly

Reading two numbers and picking the bigger one is the standard error. Five things have to be on the same basis first:

Five dimensions to align before comparing. Any mismatch and the two percentages are not on the same scale.

Line all five up and the remaining gap is usually small — which is itself informative: the real spread between comparable products is far narrower than the headline numbers suggest. Better to spend the attention on exit conditions and risk structure.

When compounding actually matters

Compounding gets talked about as though it were magic. In practice it needs three conditions at once — a high enough rate, frequent enough reinvestment and a long enough horizon — and missing any one flattens the effect:

• At single-digit rates, daily compounding beats simple interest by a fraction of a percentage point over a year;
• If the yield is a stablecoin and the principal is not, it does not auto-compound at all and you have to do it manually;
• Over a few months there is not enough time for it to open up.

So for most people, plugging the four leaks above matters more: moving your real rate from 3% to 4% beats turning 3% simple into 3% compound by a wide margin.