How to Audit Your Tracker Against the Real Numbers

Most people check their tracker against their memory, decide the number feels roughly right, and move on. That works until the day it matters, at which point they discover the figure has been eight percent wrong for a year and nobody noticed because eight percent is inside the range that markets move anyway.

An audit is the alternative. It is not complicated and it does not take long, but it has to be done in a specific order, because the single most common mistake is checking the tracker against itself.

Where does ground truth actually come from?

Not from your tracker, and not from another tracker either.

Ground truth is the balance reported by whatever actually holds the asset. For an exchange, that is the account page on the exchange itself, which is the venue's own ledger and the only authoritative record of what it owes you. For self-custody, that is the address page on a block explorer such as Etherscan or mempool.space, which reads the chain directly. These are primary sources. Everything else in your life, including every portfolio app you own, is derived from them.

Comparing two trackers to each other tells you almost nothing. When they agree, they may both be reading the same connection with the same missing sub-account. When they disagree, you cannot tell which is closer without going to the source. This is why the first step of an audit happens with the tracker closed.

One caution: an exchange balance and a block explorer measure different objects. Coins on an exchange are an entry in that venue's internal ledger, held in pooled addresses that say nothing about your share. Only compare explorer data against addresses you hold the keys for.

Collect the numbers before you open the app, and write them down somewhere you can see them all at once. Include the timestamp, because prices move while you work and a five-minute gap between reading one balance and another is enough to muddy a comparison of values.

What order should I reconcile in?

Order matters because early steps eliminate whole categories of error, and doing this backwards means chasing individual line items that a structural fix would have cleared in one move.

1. Count your venues. List every exchange account, every wallet, every chain, and every cold storage address you have ever used. Then compare that list against what your tracker contains. A missing venue is the largest single source of error and the easiest to forget, because it is invisible by definition. The hardware wallet you set up two years ago and have not touched since is the classic case.
2. Check the scope of each connection. An exchange API key usually defaults to the spot wallet alone. Funds in earn products, futures margin, staking positions or sub-accounts sit outside that scope and never appear, so the tracker reports a total that is precisely and honestly too low. Read-only API keys covers what each permission grants and what none of them should ever grant.
3. Remove duplicates before comparing anything. Two connections to one account, one address watched twice under two labels, or a manual entry shadowing a synced position. These inflate the total and will send you looking for a phantom surplus. Duplicate transactions after connecting an exchange covers finding them by identifier rather than by label.
4. Filter spam. Hide unrecognised low-value tokens before you compare, or a manufactured price on an airdropped token will swamp everything else. Spam and dust tokens explains why some of them carry five-figure valuations and why you should never try to sell one.
5. Account for committed assets. Staked positions, liquidity pools, lending receipts and locked allocations do not appear as the asset you deposited. Where your staked and pooled assets went covers each type, and for DeFi users this step usually accounts for most of the remaining gap.
6. Compare quantities, asset by asset, largest first. Not values. A quantity is a fact and should match exactly. Start with your biggest holding, because an error there dominates the total and a rounding difference on a small position is not worth your attention.
7. Then compare values, and only then. If quantities match and values do not, the difference is a pricing question rather than a bookkeeping one, which is a different problem with a different answer.

Work top to bottom and stop when the gap closes. Most audits resolve at step two or step five.

How big does a gap have to be before it's a real error?

This is the question that saves the most time, because chasing noise is how people give up on reconciling entirely.

For a portfolio of liquid assets, a difference under about one percent between your tracker and the sum of your ground truth figures is normal. It comes from refresh timing and from different price sources being sampled seconds apart. Nothing is broken. Do not go looking.

That tolerance widens sharply with what you hold. If a meaningful share of your portfolio sits in small-cap tokens, two honest sources can differ by ten percent or more on those positions alone, because thin markets genuinely do not have one price. The mechanism is set out in why the same coin shows two different prices. No reconciliation will close that gap, because there is nothing to close it to.

A gap above a couple of percent on a liquid portfolio is a real error with a specific cause, and it will be one of the seven things in the list above. A gap that is a suspiciously round fraction of the total, exactly half or exactly a third, points at duplication. A gap that appeared overnight without any transaction points at a price feed rather than a balance, or at a rename or migration of the kind described in what happens when a coin gets delisted.

Quantities are different, and this is worth being strict about. A quantity mismatch is never drift. If your tracker says 0.4213 BTC and the exchange says 0.4198 BTC, something specific happened: a fee, a partial fill, a withdrawal that did not import. Chase it, however small it looks.

The last check is performance rather than balance. Quantities can be perfect while profit and loss is nonsense, because a transfer imported without an acquisition price makes the tracker assume zero and report the whole position as gain. Scan your holdings for anything showing an impossible percentage return and repair the basis, as cost basis vs. market value describes.

How often is this worth doing?

Full audits are event-driven, not calendar-driven. Run one when you connect a new account, when you move funds between venues, when a token you hold is renamed or migrates, and before you pull figures for tax. Between those events there is nothing to find.

The quick version is a monthly glance: does the total look plausible, does the account list still contain everything, has anything stopped updating its price. Ninety seconds. The point is not to verify the number but to notice a connection that silently expired, which happens when an exchange rotates or invalidates a key and the tracker keeps showing the last balance it read.

Passive monitoring beats both. A wallet transaction alert on an address you hold long-term tells you when a balance changes, which is the only time a watched wallet needs your attention. Those advanced alert types sit on Pro at $8.99 a month and Pro+ at $14.99, alongside volume, market cap and pump triggers; simple price and percentage alerts are on the free tier.

Export a snapshot each time you finish an audit and keep it. A reconciled record with a date on it is worth a great deal at tax time, and it gives you something to diff against next time rather than starting from memory. Exporting your data covers the formats.

One caveat about doing this in The Crypto App: the depth is on mobile. Its web experience is an early preview of four areas, so plan on reconciling from your phone, where connections, manual positions and account grouping actually live.

Do this properly once and the number stops being something you interpret. After the first pass, the discipline that keeps it true is small: reconcile the moment you add a venue, rather than at the point when the figure has already been used for something.