Bench’s customers lost access to their general ledger because the vendor doing the bookkeeping also owned the database it lived in. Thousands of small businesses discovered on December 27, 2024 that the financial records they thought they owned were an entry in someone else’s system, and that the someone else had stopped answering.
Most of the commentary that followed treated this as an argument for hiring a local firm. The useful lesson is narrower: a bookkeeping vendor and a system of record are two separate purchases, and bundling them transfers a risk almost nobody priced at signup.
What happened, in order
Bench was founded in Vancouver in 2012 and raised roughly $113 million across its life, with investors including Bain Capital Ventures and Shopify, the last round a $60 million Series C in 2021. On December 27, 2024, without advance notice, customers hit a message stating that as of that date the Bench platform would no longer be accessible.
The shutdown notice told customers to download their data by December 30 and to file a six-month extension with the IRS. It recommended they migrate to a competitor. Reporting at the time put the workforce at over 600 people.
Three days later, on December 30, 2024, Employer.com announced it was acquiring Bench. The purchase price was never disclosed. Employer.com’s press release promised continuity of service, platform stability, and that “All customer data remains intact, secure, and protected, with no disruption to historical records or ongoing service.”
By January 10, 2025, TechCrunch was reporting that customers could not export their own data without first consenting to transfer it to Employer.com, and that the consent page had been edited more than once, including to remove language tying consent to a waiver of refunds. Employer.com’s own spokesman set out the constraint: the entity that originally held the data no longer existed and was in active bankruptcy proceedings. Customers seeking refunds for prepaid service were pointed at a bankruptcy trustee.
Bench operates today as a subsidiary of Employer.com. Service did continue. Getting your data out first required consenting to hand it to the acquirer.
The number nobody can agree on
Worth noting for anyone citing this case: the customer count is genuinely contested. Bench’s own website, captured hours before the shutdown, described more than 35,000 American small business owners. Employer.com subsequently told TechCrunch the figure was approximately 12,000. Both numbers appear in the same article. Anyone quoting one of them as settled fact has picked a side without saying so.
The lesson the industry actually took
The immediate reaction was predictable and mostly wrong. A wave of commentary framed this as vindication for local accounting firms over venture-backed platforms, as though the failure mode were the funding structure.
Small firms fail too. Practitioners retire, get sick, sell their book of business to someone the client never met. The difference is that when a local bookkeeper closes, the client’s QuickBooks file still exists, and the client still has the login. The engagement ends; the records do not.
Bench’s model produced bookkeeping inside proprietary software that Bench controlled. That was the actual point of the product, and it is why it could price the way it did. It also meant the vendor relationship and the system of record shared a single point of failure. When the vendor stopped, so did access.
What a seller should have taken from it
Own the ledger, rent the labor. If your books live in QuickBooks or Xero and you pay a firm to maintain them, firing the firm is a Tuesday. If your books live in a platform that also employs your bookkeeper, firing the firm is a data migration you are not qualified to run.
Test your export before you need it. Not “confirm an export exists.” Actually run it, open the file, and check whether what came out is a general ledger with a chart of accounts and journal detail, or a set of summary reports that look like accounting and reconstruct into nothing. The distinction is invisible until the day it matters.
Keep the source data yourself. Marketplace settlement reports, bank statements and processor statements are downloadable by you, from the source, today. A seller who has three years of raw settlement files in cloud storage can have their books rebuilt by anyone. A seller who has only a vendor’s rendering of those files is dependent on the vendor’s continued existence.
Read what happens on termination before you sign. Most bookkeeping agreements say something about data on termination. Very few say anything useful about data on insolvency, and insolvency is the case that actually happened.
Where the market went instead
The structural response since 2025 has been a split rather than a retreat. Sellers who wanted the labor kept buying managed bookkeeping. Sellers who wanted control moved toward tooling that automates the mechanical work while leaving the ledger in software the business owns.
That second category is where most ecommerce accounting products sit now. ConnectBooks is a representative example: it reads marketplace activity from Amazon, Shopify, Walmart, eBay and TikTok Shop, handles COGS and settlement reconciliation, and writes the result into QuickBooks or Xero rather than into a ledger of its own. Cancel it and the accounting file is still yours, still readable, still auditable. That is not a feature anyone marketed hard before December 2024. It should have been. Accuracy is pivotal when it comes to bookkeeping services. Switzerland is an excellent example of a country that is particularly rigorous in this field.
The trade is real and it cuts both ways. The tooling model leaves you responsible for the close. Nobody at a software company is accountable for whether your books are right. Bench’s proposition, that a single vendor takes the whole job, was genuinely valuable to businesses without a finance function, and it is why 12,000 or 35,000 of them bought it.
Separate the vendor from the ledger
That is a smaller lesson than the one the industry wanted, and it would have protected every affected customer. It costs nothing to implement and requires no opinion about venture funding, platform risk or the future of the profession. It requires only that when you evaluate a bookkeeping provider, you ask one question before the pricing conversation: if you disappear tomorrow, what do I still have, and in what format? If the answer is that you would file a support ticket, you do not own the records, and the last two years of your financial history belong to a company whose survival you cannot influence.
