There is one Secretary of State field worth watching weekly across the operators in a thesis, and almost no PE shop pulls it: the registered agent line. Not the entity name, not the officer list, not the formation date. Owners change registered agents when they hire new corporate counsel or move to a transactional law firm. The pattern shows up months before the banker email does.
It is a sourcing signal. The standard PE workflow treats the fifty SOS portals as a diligence-stage compliance check and misses what they could be telling a partner about the next twelve months of inbound.
The SOS file is a calendar, not a directory
A Secretary of State record for an LLC or corporation is a thin document on its own. Entity name, status, formation date, registered agent, officers and managers in some states, assumed names, periodic reports. Pull it once and you have a snapshot of who they say they are.
Pull it every week for two years and the same file turns into a calendar of every moment the owner had to call a lawyer.
That second thing is the sourcing artifact. A clean ten-year filing history with no agent changes and a single founder still listed as manager is a different target than a five-amendment entity that just swapped its registered agent twice. The firm watching the second pattern gets a call queue. The firm that isn't gets an inbound CIM nine months later and a competitive process.
What each filing type tends to mean
Five fields carry most of the weight. They cluster in patterns that show up months before a deal does, and the firms that miss the patterns are the same firms paying retainers to bankers who already saw them.
Change of registered agent. The most underrated signal on the list. Owners do not change registered agents on a whim. They change them when they hire a new corporate lawyer or move from a generic agent service to a transactional one. A small operator who quietly switches to CT Corporation or Cogency Global is not doing it for fun. The shift typically precedes any banker engagement by months — the legal scaffolding for a sale is the first work, the banker is hired after. The signal is loudest when paired with no other operational changes. The business looks identical from the outside. The legal representation just got more serious.
Officer and manager turnover. Noisier, because operating agreements update for tax, estate, and family-office reasons that have nothing to do with a sale. Watch for two specific shapes: a long-tenured CFO dropping off and being replaced by a "consultant" or "advisor" title, or a new manager whose name maps to a known fractional CFO or M&A advisor. Riveron, BDO, and the major fractional-CFO networks are all searchable in role-history feeds. That pattern is the closest thing to "the owner is now packaging the business" the public record gives you.
Dissolution and intent-to-dissolve filings. A subsidiary going to dissolved status while the parent stays active usually means the owner is cleaning up the org chart for a transaction. Merging two related LLCs into one almost never happens for operational reasons. It happens because a buyer asked.
Name changes and conversion filings. A C-corp converting to an LLC eight months before any other public signal is somebody's tax counsel running a structure analysis for an exit. Renames from a founder-name LLC to a generic operating name are the cosmetic version of the same prep work.
Assumed name and DBA filings. Less useful for exit, more useful for roll-ups. When a target you already know is consolidating files multiple new DBAs in three counties, that is a tuck-in that closed before any press release. The assumed-name index is the closest thing to a real-time deal feed the public record offers.
The states that matter, and the states that don't
Nobody on the SERP says this out loud: SOS data quality varies more than the signals themselves do.
Delaware is the canonical case. A large share of the entities you care about for a U.S. mid-market thesis are domiciled there, and the Division of Corporations entity search exposes a structured form with consistent fields, plus a paid status and document-order layer behind it. The catch is that Delaware's LLC Act does not require member or manager names on the Certificate of Formation, so the file itself almost never carries them. Delaware tells you a lot about formation, amendments, mergers, and registered agents. It tells you almost nothing about who runs the company day-to-day.
California is the opposite. The Secretary of State's bizfileOnline portal exposes officer and agent data on corporations, LLCs, and LPs and refreshes with reasonable latency. The structured fields are there. California does not require periodic reports as often as some peers, so the staleness on any given entity can run a year or more before a refresh forces an update.
New York is workable but stingier. The Department of State's corporation and business entity database exposes the DOS process address (a de facto agent) and principal executive office data. There is no clean bulk-data export. Anyone telling you they have a New York entity feed is scraping it.
Texas is large, important, and frustrating. SOSDirect charges a dollar per search on top of a prefunded client account, the UI is dated, and the structured data behind it is uneven. The signals are there for any operator domiciled in Texas. The cost of getting them out at sourcing-pipeline volume is not trivial. We bake the prefund-and-query loop into the scraper; most firms try it manually for two weeks and quit.
After those four, the curve drops off fast. Most states publish entity searches fine for a one-off lookup and miserable for continuous monitoring. Some still require a mailed form for anything past the search page. A thesis weighted toward the long tail (Wyoming, Nevada, the Dakotas) will hit pages that have not been redesigned since around 2008, and the scraper that handles them looks like a small museum exhibit.
Why cadence beats corpus
Almost every blog post about SOS data treats the corpus as the asset. It isn't. The corpus is downloadable, and most of it is free. The asset is the difference between this week's snapshot and last week's.
A pipeline that pulls the same set of operators every Wednesday and diffs the result is doing something a vendor cannot do for you on a one-off engagement. It produces a stream of events: agent changed, officer added, status flipped to merged, DBA registered. A partner can scan that stream in five minutes on a Thursday morning and pick the three calls worth making that week.
That's the part Corridome builds. A firm running this themselves should prioritize the four states where most LMM activity lives — Delaware, California, New York, Texas — plus Florida and Illinois for sector-specific theses, and accept that the long-tail states need a tolerant scraper and slower cadence. The diffs feed a scoring layer alongside whatever else the firm is watching: LinkedIn role changes, license-renewal windows, permit velocity, web-stack churn. The SOS feed is one input, and a strong one, because the underlying record is a legal document and operators do not file legal documents lightly.
Here's the part that costs firms real money. A firm without a cadence layer can buy the corpus and call it deal sourcing. What it cannot do is see the moment a target changes shape. By the time the change surfaces in a vendor refresh cycle, the banker is hired, the teaser is in three competitors' inboxes, and the entry multiple has moved half a turn. The cost is the proprietary call that closes at a lower multiple, and the head start on an owner who hasn't yet decided to sell. Neither shows up in a vendor's sales deck.
The firms that get value from this do not buy "SOS data." They run a pipeline that watches the right fields on the right operators at the right interval, and they own the result.
