Don't judge your DCA profit by the coin's price change
In a 10-round example the panel shows a return of +4.91%, while bitcoin is down 31.10% over the same stretch. DCA money goes in bit by bit and each payment stays invested for a different length of time, so that percentage can't be lined up against a price change. Average cost, simple return and money-weighted annualized return each answer a different question.
- The four numbers on a DCA panel each do one job
- Ten rounds and $5,000 laid out in full
- +4.91% and −31.10% aren't the same kind of percentage
- Average cost is the number you actually control
- An annualized +10.92% doesn't mean you made 10.92% in a year
- Which DCA numbers are worth checking, and how often?
Follow the 10-round example below and the panel's return will read +4.91%. Measured from the month-end price before the first purchase, bitcoin went from $114,048.93 to $78,581.29, which is −31.10%. The two numbers point in opposite directions, and neither is miscalculated.
The whole piece uses one example. Starting in October 2025, buy $500 of bitcoin at the close at the end of each month for 10 rounds, the last one at the end of July 2026, and value the holdings at the end-of-August 2026 close. Prices are BTCUSDT monthly closes (UTC) from Binance's public market data API.
The four numbers on a DCA panel each do one job
The table below is a generic layout of the four figures a DCA plan's panel puts side by side. Platforms label them a little differently, and for how your platform defines return, go by its own help page.
| Generic label | What it is | The question it answers |
|---|---|---|
| Amount invested | Everything the plan has debited so far | How much have I put in? |
| Portfolio value | Your coins valued at the current price | What is it worth now? |
| Total PnL | The second minus the first, as an amount | How much am I up or down on paper? |
| Return | Total PnL ÷ amount invested, as a percentage | For every $1 I put in, what is it worth now? |
Only the last of the four is a percentage, and that's where all the trouble starts: it looks exactly like a price change, so it invites a straight comparison with one. The two percentages don't share a denominator.
Ten rounds and $5,000 laid out in full
Each round is $500, filled at that month's closing price, and buys the following amount of bitcoin.
| Month | Month-end close (USDT) | BTC bought with $500 |
|---|---|---|
| 2025-10 | 109,608.01 | 0.00456171 |
| 2025-11 | 90,360.00 | 0.00553342 |
| 2025-12 | 87,648.22 | 0.00570462 |
| 2026-01 | 78,741.09 | 0.00634992 |
| 2026-02 | 66,973.26 | 0.00746567 |
| 2026-03 | 68,284.48 | 0.00732231 |
| 2026-04 | 76,346.57 | 0.00654908 |
| 2026-05 | 73,674.39 | 0.00678662 |
| 2026-06 | 58,624.71 | 0.00852883 |
| 2026-07 | 62,887.88 | 0.00795066 |
Over 10 rounds you put in $5,000 and end up holding 0.06675284 BTC. At the end-of-August 2026 close of 78,581.29, that's worth $5,245.52: a total PnL of +$245.52 and a return of 5,245.52 ÷ 5,000 − 1 = +4.91%. The table shows amounts to 8 decimal places, while the market value uses the unrounded amounts, so adding up the table and multiplying will leave you a few cents off. Those four values are what the four panel figures show.
One detail about the valuation: it doesn't use the latest price at the time of writing. September's monthly candle hadn't closed yet, so the screen showed a live price that would be a different number a few days later. Treat a number that's still moving as a close, and when you pull the API data yourself later, nothing will match. So every purchase price is the close on the last day of the month, and the valuation uses the August 2026 monthly candle, which has already closed.
This calculation also leaves out fees and spreads. A real account would come out slightly lower, and fills wouldn't land exactly on the month-end close.
+4.91% and −31.10% aren't the same kind of percentage
Bitcoin's −31.10% rests on one firm assumption: all the money goes in at once at the end of September 2025 and stays until the end of August 2026, 11 months. One sum of money, one starting point.
The $5,000 in the DCA plan has no single starting point. The first $500 was in for 10 months, the last $500 for just 1 month, and the eight in between each for their own stretch. Squeeze 10 starting points into one percentage and set it against a single-start price change, and you're comparing two different things.
The amounts bought make the difference plainer. The June 2026 round's $500 bought 0.00852883 BTC; the October 2025 round bought only 0.00456171, close to half as much. In the cheaper rounds the same money brought back more bitcoin, which pulled the average cost of these 10 rounds below the valuation price.
Now the other side. The first purchase was at 109,608.01 against 78,581.29 at valuation, so that round on its own is −28.31%, a loss. The total is still positive because that round is only a tenth of the money. Most of the other nine were bought lower, so most of the money went in when prices were low and stayed in for a shorter time.
One more comparison: the same $5,000 put in all at once at the end of October 2025 would be worth $3,584.65, −28.31%.
It's tempting to conclude from this that DCA is the better deal. Don't. This stretch fell first and then recovered a little, which happened to favour buying in installments; over ten months of steady gains the first purchase would have been the best one, and the conclusion would flip. This example shows how the three numbers are calculated and read. It isn't a backtest and proves nothing about which approach is better. Choosing between DCA and a lump sum is a separate decision.
Average cost is the number you actually control
Average cost = total invested ÷ total held = 5,000 ÷ 0.06675284 = $74,903.18 per bitcoin. The valuation price of 78,581.29 is 4.91% above it.
Matching the simple return above exactly is no coincidence. The two formulas are the same thing: (current price × amount held) ÷ (average cost × amount held) − 1, the amounts cancel out, and you're left with current price ÷ average cost − 1. Watching the average cost and watching the return percentage give you exactly the same information.
Since they're equivalent, I'd rather watch the average cost, because it comes with a unit. 74,903 is a price you can draw as a horizontal line on the chart: price above the line is a paper gain, below it a paper loss, no conversion needed. A percentage can't do that, and it tempts you to compare it with other percentages.
More to the point, you draw half of that line yourself. The price belongs to the market, but how much you put in each round, when it's debited and whether the debit goes through decide how much bitcoin you pile up at each price. The job of DCA is to keep that line from hugging the highs. When you're setting the plan up, the site's DCA calculator can help: enter the amount per round, the frequency, how long you plan to keep going and a worst-case drawdown, and it works out your total invested and your exposure in the worst case. Bear in mind that it models the plan, not the average cost you've actually achieved; for that, go by your own trade history.
An annualized +10.92% doesn't mean you made 10.92% in a year
Put each round's date and amount, plus the final market value, into XIRR as cash flows, and these 10 rounds come out at +10.92%. Same account, and annualizing more than doubles the figure, without a single extra cent or coin in it.
The difference is all about time. XIRR is a money-weighted annualized return: it measures how efficiently each payment worked over the time it was actually invested, then scales that to what a full year at the same pace would give. Of this $5,000, only the first $500 was in for 10 months, and the last round for just 1. Money that gains 2% in a month annualizes to more than twenty percent, not because it really earned that much but because the formula assumes it keeps going at that pace.
So what it can answer is which of two plans, measured the same way, uses money more efficiently. It can't tell you how much extra money you've made over the year; for that, look at total PnL as an amount, +$245.52. Also, the shorter the period, the more annualizing magnifies the result, so three months into a DCA plan, don't take the figure seriously whether it looks great or terrible.
Which DCA numbers are worth checking, and how often?
- Whether each round went through as planned. This is the one thing that needs fixing right away once you spot it. When a round is missing, check which version of Binance recurring buy the plan runs on, whether the execution time has arrived (its minute is the minute you created the plan), and whether your balance covered that round.
- Average cost. One price, one line, easy to read whichever way the market moves, and it records the decisions you've made along the way.
- Total invested as a share of your position cap. This number only goes up, while the cap is one you set at the start. As it gets close, decide in advance where you'll stop instead of noticing only when a debit fails. The cap itself comes from working out how much to put into crypto.
- Total PnL as an amount. A percentage can mislead depending on how it's measured; an amount can't.
As for frequency, look once per debit cycle: two or three days after each debit date, confirm the round went through and note down your average cost. Open the app several times a day and your decisions easily drift away from the rules you set at the start: two days down and you want to add more, three days up and you want to take profit. Checking more often won't lower your average cost; it only makes your fingers itch.
You can look at the return percentage too. Just don't compare it with the coin's price change, or with percentages other people post. You don't know when their money went in, and that's the most important half of the information in that number.
Frequently asked
If I changed the amount per round partway through, do these three numbers still work?
Yes. Average cost and simple return only use two running totals, total invested and total amount held, so how much each round took and how many rounds there were don't affect the formulas. XIRR already works from each payment's actual date and actual amount, so a change in amount is no problem for it. What stops working is a description like “I put in $500 a month”: once you've changed the amount, what represents your plan is the total invested, not the current round's amount.
If I stopped the debits for a few months, will the annualized return come out lower?
Not because of the pause itself. Months with no cash flow don't enter the XIRR calculation, you don't need to add a zero for them, and the time weighting of the rounds you did invest stays the same. The change shows up elsewhere: you bought nothing at the prices of those paused months. If prices were low then, you missed a chance to pull your average cost down; if they were high, you avoided them. A pause affects how much you hold, not the annualizing formula.
After selling part of my holdings, do I need to recalculate my average cost?
That depends on which question you want answered. If you want to know what each remaining coin cost you, then on a moving weighted average basis a sale doesn't change the per-coin average; what changes is the absolute size of the total invested and the total held. If you want to know whether the money made a profit overall, the sale proceeds go into XIRR as a positive cash flow, and leaving them out understates the return. Which basis a panel uses differs between platforms, so if its help page doesn't say, don't assume it matches the way you calculate by hand.