Work backward from what still exists. Inventory every address and account, pull the records that can disappear, then fill what remains using a method you write down. The mistake I see most often is treating this as arithmetic. Reconstruction is an evidence exercise, graded on whether your method was reasonable, consistent, and documented.
Part of our guide: Crypto Taxes.
The short version
- Rank your sources by independence and permanence. A transaction hash outlives every company that touched it; an exchange CSV holds more detail and disappears with the account.
- The chain never carries a price. Every valuation comes from an off-chain source you have to name and keep naming.
- The standard is a reasonable method, applied consistently and documented, the same footing the IRS uses in its basis-allocation safe harbor (Rev. Proc. 2024-28).
- Reporting zero basis is a deliberate choice. It overstates gain, and I would reach it only after documenting a search that failed.
- Form 1099-DA turns a private reconstruction into a reconciliation. Brokers report gross proceeds for transactions effected on or after Jan. 1, 2025, and basis on certain transactions effected on or after Jan. 1, 2026 (IRS).
Rank the evidence before you gather it
Digital assets are property for U.S. tax purposes, so basis works as it does elsewhere: it is “the amount of your investment in property for tax purposes,” and Publication 551 requires that you “keep accurate records of all items that affect the basis of property” (IRS). Years later you are rebuilding that record from survivors, and ranking them is the first decision.
On-chain data is the strongest thing you hold. A hash outlives every company that touched it, and any stranger with a block explorer can confirm it. It proves a quantity moved between two addresses at a block time, and nothing more.
Venue records carry price, date, quantity, fee, and counterparty in one row. They are also mortal, living on a company’s servers until the account closes or the venue fails.
Fiat rails (bank and card statements, wire confirmations) corroborate only the dollar leg, and because they are not crypto-specific they usually outlive the venue they paid. Your own files are weakest alone, since nobody independent produced them, though each row gains weight the moment it ties to a hash or a bank line.
Strength and durability run in opposite directions here. The most permanent record carries the least information, and the richest record is the one most likely to be gone. That mismatch dictates the order of work.
Inventory addresses before you pull anything
You cannot request records for an account you have forgotten, so the address list caps the quality of everything downstream.
Walk every seed phrase, hardware device, browser extension, and custodial login you have used. One seed can produce many addresses across many chains, and older software used different derivation paths, so pull the full address set instead of the one address you remember. Then let the chain extend the list: each address history names counterparties, and old deposit addresses surface venues you had forgotten. Email is the best index most people have, since defunct venues leave withdrawal confirmations behind long after the site is gone.
The IRS expects identification at the unit level, either by “the specific unit’s unique digital identifier such as a private key, public key, and address,” or by records covering all units of that asset in one account, wallet, or address (IRS FAQs). Both routes assume you know which wallets exist. The crypto estate data room checklist is the same sweep built for another purpose and works as a starting list.
Fill the gaps with a method you can name
Rev. Proc. 2024-28 is worth reading even if you never claim its safe harbor, because it shows what the IRS treats as adequate. It permits reliance on any reasonable allocation of unused basis, then conditions the whole thing on records:
“The taxpayer must be able to identify and maintain records sufficient to show the number of units of unused basis, the original cost basis of each such unit of unused basis, and the acquisition date of the digital asset unit to which the unused basis was originally attached.”
Rev. Proc. 2024-28, § 4.02(5)
Cost and acquisition date, per unit. That is the target, and it ranks your fallbacks. Direct evidence first: a hash, a venue confirmation, and a bank line that agree. Where the day is missing but the window is known, price it from a dated source inside that window and record which source. Where only a range exists, pick a convention that does not flatter you, since the least favorable price in a window is defensible and the most favorable one invites a question. Zero basis sits at the bottom, overstating gain and overpaying tax, and I would accept it only where the failed search is written down and the position is too small to justify more work.
Consistency is the part people skip: one method per asset type, applied across the whole record, with every exception noted. Then write a basis memo, one page covering what you searched, what you found, what is missing, the method you chose and why, and the price source. Date it. That memo converts a spreadsheet into evidence, and it costs an hour while the search is fresh.
Do this before a broker forces the comparison. Assets transferred in from an outside wallet often reach a broker with no basis attached, and Form 8949 handles that by letting you “enter the correct basis in column (e)” (Instructions for Form 8949). Everything supporting column (e) is on you. The volume is worst for entities, where staking and stablecoin activity multiply the events feeding the entity’s return.
What I actually see
The reconstruction covers the largest position and stops. People attack the holding that matters most, which usually has the best paper trail already, and leave a dozen small positions on a venue that shut down. Those are the entries with nothing behind them, and the ones an examiner asks about.
The work gets redone from scratch by every new preparer. Nobody wrote the method down, so the next CPA rebuilds it and reaches a different figure for the same coins. I have seen one wallet produce three basis totals across four years of filings because the reasoning was never recorded.
The valuation source drifts. Early years priced from one aggregator because the old software used it, later years from the exchange’s own fill, with nothing marking where the switch happened. Each half is reasonable. Together they are unexplainable.
Here is the check I would run first. Pick one asset. Open a block explorer and list every incoming transfer to every address you control for it, oldest first. Beside each, write where the units came from and what you paid. Total the units and compare against what you hold today. If the two disagree, you are missing addresses, and no amount of price research will fix it. Run this before paying anyone to point software at your data, because software returns a confident number from an incomplete address set and gives you no way to see the hole.
Where this goes wrong
The file arrives at a number nobody can retrace.
The failures repeat. Prices drawn from several sources with no record of which applied where. Internal moves between the owner’s own wallets landing in the disposition column, so gain is overstated and the overpayment is voluntary. A memo nobody wrote, leaving a method that exists only in the head of whoever left the engagement. Software output accepted as a finding because it arrived formatted, when it only ever saw the addresses it was given. And the expensive one: an exchange that closed years ago holding the only priced record of the earliest buys, discovered in March with a deadline in the way.
The decision rule
- Inventory every address and account first, including seeds you no longer use and venues that no longer exist.
- Pull the perishable sources this week: trade history, deposit and withdrawal statements, and account exports from every venue you can still reach.
- Anchor each acquisition to a hash or a statement, and record on the row which one supports it.
- Name one price source per asset type and apply it to every gap you fill with it.
- Reconcile units before prices, since a unit count that does not tie means the address list is incomplete.
- Reconstruct per wallet, because the accounting now runs wallet by wallet and a pooled total will not survive the split.
- Write the basis memo the same day, covering the search, the findings, the gaps, and the chosen method.
- Have a CPA review the method before the numbers. The method gets tested first, and correcting it early costs far less than repricing a finished file.
Where a position has no surviving evidence at all, keep it in the memo and state what you searched, so the gap stays visible. Reconstructions fail review when nobody can tell which figures were documented and which were supplied.
Where this sits
Reconstruction sits underneath tax reporting, where the numbers surface, and on top of custody, where the evidence lives. When an entity or a trust holds the assets there is an extra link, because the entity’s acquisition and the contributor’s original purchase both need evidence. Families rebuilding a record after a death face the hardest version, and the common mistakes are worth reading in that light.
This question crosses three desks. Custody controls the raw data, a CPA owns the method and the filing, and an attorney matters once an entity or trust document sits in the acquisition chain. Each does their own part competently. The join is where it fails: the CPA assumes the address list is complete, the custodian answers only what it was asked, and nobody owns the question of whether something is missing entirely. Decide who owns that before the work starts. If you would rather have that ownership settled before the reconstruction begins than discovered halfway through it, crypto tax records and coordination is where my firm starts.
Sources
- IRS, Digital assets
- IRS, Revenue Procedure 2024-28, safe harbor for allocating unused basis of digital assets
- IRS, Publication 551, Basis of Assets
- IRS, Instructions for Form 8949
- IRS, About Form 1099-DA, Digital Asset Proceeds From Broker Transactions
- IRS, Frequently asked questions on virtual currency transactions
- 26 CFR § 1.1012-1, Basis of property
Related
- Crypto tax reporting for LLCs
- Crypto estate data room checklist
- Can a Wyoming LLC stake crypto?
- Can an LLC hold stablecoins?
- Crypto custody for LLCs
- Crypto tax and records
Last updated: 3 August 2026.
This article is general education, not legal, tax, or investment advice. Reconstructed basis depends on your own records, the assets involved, and the years at issue. Talk to a qualified CPA about your own situation.
