BuiltWith Alternatives Tested: Seven Sources, One Question, Answers 8.5x Apart

We asked seven BuiltWith alternatives one question, the way a buyer would before an outbound campaign: how many companies use HubSpot? We used only what each site shows a logged-out visitor, on 5 September 2026, and then asked the same thing about Klaviyo and PostHog. The answers for HubSpot ran from 117,653 to 1,001,024. BuiltWith itself did not answer at all; its trends pages served an image verification challenge on three different routes, while its pricing page loaded first time.

The free HubSpot count runs from 117,653 to 1,001,024

Each row is what the source printed. The unit column matters more than the number: three sources count sites or pages, two count companies, one counts a share of a named list.

We expected the company counts to sit inside the site counts, since one company can run several sites. Enlyft's 1,001,024 companies is nearly double Wappalyzer's 540,000 websites, so at least one of the two is counting something other than what its label says.

For PostHog the same two sources are 101x apart

HubSpot is old and everywhere, which flatters every counting method. PostHog is four years into most of these indexes, and the spread opens up.

Wappalyzer has PostHog on more websites than Klaviyo. Every other source has it the other way round, by 2x to 10x. We do not know which is right, and nothing on the pages lets a reader work it out.

Two of seven print what the number is out of, and two print a date

A count is only useful for sizing a campaign if you know the pool it came from. Wappalyzer says it "tracks 540,000 live websites using HubSpot" and TheirStack says "we have data on 229,488 companies and users that use HubSpot". Neither page says how many websites or companies it checked, so neither number can be turned into a share of anything.

Enlyft prints a share, "about 2.1%" of a category it calls Enterprise Marketing Management, next to its 1,001,024 companies. Divide one by the other and the category would hold roughly 48 million companies. The category size is not on the page, and PostHog is scored in a different category, so the shares cannot be compared with each other.

TheirStack's HubSpot page lists customers by continent under the headline: 42,146, 24,258, 7,227, 1,948, 1,573 and 464. Those add up to 77,616. The headline says 229,488. The page does not say where the other two thirds are.

Installmap prints the denominator and the date on the same page as the share: "4,885 of 4,969 Inc 5000 domains returned a usable record and form the denominator", crawled 4 September 2026, with the split between rendered crawl, plain fetch and no record. PublicWWW prints its index size and date on the homepage, 524,927,560 pages on 3 September 2026, though the result page itself only says "117653 web pages in 0.01 s".

The number changes more by list than by vendor

The biggest swing in the whole test was not between vendors. It was inside one source on one day. Installmap crawls named lists of companies and prints the share of each list, so the same tool can be read against several populations at once.

HubSpot goes from one in four Inc 5000 companies to one in forty in the Fortune 500. PostHog is on 12.8% of YC company sites and 1.8% of the Inc 5000. An ICP that says "companies using HubSpot" describes a pool that is ten times bigger or smaller depending on which of those lists your buyers resemble, and none of the global counts above will tell you which. Cells marked n/a fall under Installmap's publishing floor.

What we do with this on a client build

We pick the population before we pick the vendor. If the client sells to funded software companies, the YC and unicorn shares are the sizing numbers, and a global "540,000 websites" is noise. If the client sells to mid-market services firms, the Inc 5000 cut by industry and state is closer. Installmap gives those shares free, with the base printed, so the sizing step costs nothing.

Then we buy or build the list from whichever vendor exports the shape we need. That is a different decision, and it is mostly about price, export limits and whether contacts come attached.

The failure we see most often is a team buying an export of every company running a tool and finding out mid-campaign that the pool was a tenth of what the vendor count implied, because their buyers live in one list and the count was over the whole web. The sizing step above would have told them before they paid.

Our stack-based lists start from companies using Stripe, companies using Snowflake and SaaS companies, and get filtered to a client's ICP before anything is sent. The wider set is on the email lists page.

How we ran it

All fetches on 5 September 2026 through a rotating proxy, with a rendered browser request where a plain fetch returned a challenge or a JavaScript shell. No accounts, no logins, nothing bought. BuiltWith's three trends pages returned the same verification page on a datacenter proxy, a rendered request and a residential IP, and we did not try to solve it. SimilarTech failed at the connection on four routes. NerdyData shows no count to a logged-out visitor. Installmap has no page for some list and tool pairs; those show as n/a above. Where a vendor did not print something, the cell says so rather than guessing.