Secretary of state search: exact, starts-with and contains

A register does not know what you meant. It compares the characters you typed against the characters it stores, under matching rules written for a filing clerk, not a researcher.

Almost every frustrating search is a collision between those two things. Almost every fix changes the string, not the assumption that the business is real.

This page is about the matching itself: how a secretary of state search compares names, how to build queries that survive a register's literalism, how to triage a result list, and why a confident wrong match is the expensive outcome rather than an empty one.

What to take away

The three matching behaviors

Search boxes on filing office sites generally behave in one of three ways, and the same query produces very different results under each.

Exact match returns only records whose name is character-identical to what you typed. A missing period after "Inc" is enough to return nothing.

Starts-with returns records whose name begins with your string. Useful, and it punishes anyone who types a leading article. Searching "The Bridge" finds a company registered as "The Bridge Works", while searching "Bridge Works" does not.

Contains returns records with your string anywhere in the name. It is the most forgiving and the most likely to bury you in results.

None of the three modes is approximate matching. Each one compares characters, so none of them knows that "Jon" and "John" are the same person or that a company was renamed last year.

You usually cannot tell which mode is running by looking at the box. When an office publishes the rule, it sits on the search page itself or in a search tips note, so reading it once for a state you use often beats any amount of retyping. Offices do at least name their tools: New York's Department of State runs its Corporation and Business Entity Database, and the page that carries that search box says so plainly.

Where it does not say, test it: search a fragment from the middle of a name you have already found. If the record comes back, the search does contain matching.

A fourth behavior exists and is worth knowing about even though few entity registers offer it. Some systems search the text inside filed documents rather than the names attached to them, as the SEC's full-text search does. That answers a different question entirely, and results from it should never be mixed into a name search without saying which produced what.

Build a query set, not a query

Stop thinking about the query as one string to get right. Think of it as a small set of transformations of the name you were given, run in descending order of likely success.

That last transformation is usually the winner. A rare root word searched alone under the most forgiving mode the office offers will find the entity no matter how its suffix, punctuation or word order is stored. The cost is a longer result list, which is a problem you can solve by reading. A nil result is not.

Reading a result list

Result lists come back in three shapes, and each has its own failure.

Nothing. Work the query set above first. If it is still empty after that, the assumption to change is which index or which state you are in, not the spelling, and the reasons are set out in the secretary of state search directory.

Too many. Do not lengthen the name. Narrow with a different field: city, formation year, entity type, status, or the agent. A second field cuts a list faster than a longer string, and it does not risk excluding the record you want by adding a word the office does not hold.

Exactly one. This is the dangerous one. A single hit on a name you half-recognize feels like the end of the search, and it is the point at which people stop checking. Confirm it against something the counterparty gave you: the formation state, the city, the entity type, or best of all the file number on their own paperwork. A register that issues one number per entity makes that trivial, which is why a fifty-register system, where each state assigns its own number, routinely does not.

The asymmetry that should govern all of this

The two ways a search can go wrong are not equally expensive.

A false negative announces itself: the list is empty, you can see it, and you go back to the query set. A false positive does not. It hands you a clean-looking file, a plausible screenshot, and a status field that belongs to someone else. Every extra minute spent confirming a single hit is bought against that second failure, which is where the real money is.

This is why similar names in one state deserve more caution than they get. Two entities can hold nearly identical names on the same register, formed years apart by unrelated people, and the register is content with that as long as the office considered them distinguishable when it accepted the filing.

Stop searching and start asking

There is a point where searching harder is the wrong move, and it arrives earlier than most people think. If two rounds of the query set have not produced a confident match, the cheapest remaining step is to ask.

Ask for three things, in writing, from the counterparty: the exact legal name, the state of formation, and the file or entity number. Ask for them on something they produce anyway, such as a quote or a draft agreement. All three take them seconds to supply, and a reluctance to supply them is itself information.

Once you have the number, use it for everything afterwards. It survives the rebrand, the typo and the suffix, which is the argument made at length in entity number lookup.

Keep the search pointed at the entity

Name searching gets close to people quickly, since officer and agent names sit on the same records, so keep the exercise where it belongs.

The question is which legal entity you are contracting with and whether its record is in order, and entity fields alone answer that. The wider routine is set out in company verification.

Common questions

The office's search found nothing, but a commercial site shows the company. Who is right?

The office, about its own register. The commercial site is showing a copy, possibly of a different state's record, possibly old. Take the name and state it shows you, then run that against the office directly and see what comes back.

Should I search every state?

Only if you have no idea where the entity was formed, and then only until you find it. Start with the state of formation if you know it, then the state where the work happens, then anywhere the business has an obvious footprint. Sweeping fifty registers is a sign you are missing one fact you could have asked for.

The register shows two entities with almost the same name in the same state. Which is my counterparty?

Neither, yet: resolve it with the file number from their own document, or the formation date or city.

Choosing the more plausible of two records puts the wrong party on a signature page. The register and the status field tell you nothing to distinguish them, a limit described in registration status.