How we review
There are no stars on this site, no scores out of ten, and no "editor's choice" badges for sale. There are ranked picks with published reasons, and this page is the rulebook those picks are made under, so you can hold them (and us) to it.
The criteria, in order
1. Three-year cost. Formation plus registered agent charges across three years at renewal prices. This single rule reverses most of the industry's marketing order, because the category prices its door at $0 and its renewals at two to four times the honest rate.
2. The public record. Whose address goes on your filings, how documents are handled when they arrive, and whether the service's model treats your data as an asset to protect or a list to monetize.
3. Treatment at friction. Support reachability, cancellation friction, complaint patterns in public records (BBB files, review themes), and whether the checkout respects a "no."
4. Honesty of the price itself. A service whose renewal is stated on day one starts ahead of a cheaper service whose renewal is discovered in year two.
| Criterion | What we measure | What fails it |
|---|---|---|
| Three-year cost | Formation plus registered agent at renewal prices, added in the order the bills arrive | A cheap year one priced back in years two and three; renewals that are not published at all |
| The public record | Whose address appears on the public filing, how legal documents are scanned and forwarded, whether customer data is a product | The founder's home address on the record by default; document handling that waits for a weekly scan batch |
| Treatment at friction | Support reachability, cancellation friction, complaint patterns in BBB files and public reviews | Checkout flows that treat "no" as a speed bump; cancellations that require a phone call and a retention script |
| Honesty of the price | Whether the renewal is stated on day one, in the open, at a figure that does not move | "Free first year" framing with the renewal in fine print, or no renewal figure published anywhere |
The worked example: why the order flips
The two tracked picks show the criteria doing their work. On year-one price, Bizee wins walking away: $0 formation against Northwest's $39, and an agent at $119 against $125. Every listicle in the industry stops there. Run the same two services across three years, the horizon this page exists to enforce. Bizee's Basic route charges the agent from year one (the free agent year applies to its paid tiers), so the agent alone totals $357. Northwest's formation route includes the first agent year with the filing, so formation plus agent totals $39 plus two paid agent years: $289. The $0 door costs $68 more than the $39 door across the period a business actually lives in, and the reversal comes entirely from pricing the renewals instead of the welcome offer. That is the whole method: same published figures everyone else has, added up over a longer honest horizon.
The example also shows what the method is not. It is not a claim that Bizee is a bad service; Bizee is our #2 pick, recommended on state pages where its shape fits, and a founder who will switch agents after year one can rationally take the $0 door. It is a claim that a ranking which never reaches year two is not a ranking of the product being bought, because an LLC is not a one-year purchase. The agent renews every year the company exists. Price the stream, not the door.
What would change a pick
Three events, all mechanical. A renewal price change at either pick re-runs the three-year arithmetic and re-orders the ranking if the arithmetic says so. A privacy regression, an agent whose filings start exposing customer addresses or whose document handling degrades, demotes on criterion two regardless of price. A pattern of friction in the public complaint record, sustained across a recheck cycle rather than a single bad month, demotes on criterion three. Changes land in the configuration, propagate to every page in one build, and are recorded in the updates log with the reason. A pick that cannot lose its position is not a pick; it is an advertisement, and this site already has a page that explains which parts of it are advertising.
The re-review cadence
Reviews are re-read on the same cycle as the state dataset, because a review is a price claim with a date on it. Published prices, program terms and the complaint record are checked together; a price move that does not change the ranking is still updated in the review's table, because a stale table in a review about pricing honesty would be its own refutation. Where a provider's terms cannot be re-verified (an unpublished renewal stays unpublished), the review says so again, louder. The standard this page sets is the one the site is judged by: if a figure in any review disagrees with the provider's current published price, the provider's page is right and ours is wrong until the recheck lands, and the correction joins the public log like any other.
One winner per intent, from one source
Every page on this site renders its picks from a single configuration: the verdict cards, the comparison tables and the body verdicts cannot disagree, because they are the same data. Northwest Registered Agent is the #1 pick for formation and for agent service; Bizee is #2. If that ever changes, it changes in the config, everywhere at once, and the updates log will say when and why.
The conflicts, stated plainly
We earn affiliate commissions from both tracked picks (rates are on the disclosure page). We also link to providers we earn nothing from, and we rank one of them (LegalZoom) off our lists entirely on price and friction, in a category where its program pays publishers well. Our founder uses the #1 pick for his own LLC. We consider that an alignment worth disclosing rather than a bias worth hiding: the pick predates the commissions in his case; he was a customer first.
What we never do: accept payment for placement, publish ratings we cannot defend, invent testing, or write a usage claim beyond the founder's own. Where a service is reviewed on published facts alone, the review says so.
Criteria current as of November 17, 2026.