Four companies own the integration map of small business software

Every product profile on this site records the integrations its vendor names. We have now counted all of them: 9,601 integration mentions across 858 product dossiers in 49 categories, resolving to 3,507 distinct destination names. Read one at a time an integration list looks like a feature. Read all 858 side by side and it becomes a map of who small business software is really built to talk to, and the map is far more concentrated than any individual pricing page suggests.

The headline is four names. Zapier is named by 423 products, 49.3 percent of everything we track. HubSpot follows at 305 products, 35.5 percent. Slack is at 282, 32.9 percent. Salesforce is at 269, 31.4 percent. Then the curve falls off a cliff: Shopify is fifth at 141 products, 16.4 percent, less than half the fourth-place name. The median product names nine integrations, and the odds are close to even that Zapier is one of them.

The four gravity wells

Platform	Products naming it	Share of 858
Zapier	423	49.3%
HubSpot	305	35.5%
Slack	282	32.9%
Salesforce	269	31.4%
Shopify	141	16.4%
Google Analytics	130	15.2%
Make	105	12.2%
Mailchimp	100	11.7%
Stripe	92	10.7%
Pipedrive	84	9.8%

Four platforms and four jobs, and the jobs explain the ranking better than market share does. Zapier is the universal adapter, the thing a vendor supports when it cannot afford to build every integration a customer might want. HubSpot is where a lead is expected to land. Slack is where a human is expected to be told. Salesforce is where a larger buyer's record of truth is assumed to live. Almost every product in this database is, structurally, a machine for producing one of those four events: a record, a lead, a notification, or a row someone else's system needs.

Worth noting what is not in that list. Only 26 products name n8n, 3.0 percent, against 423 for Zapier, a sixteen to one gap that has not closed despite n8n's visible popularity among technical teams. The self-hosted and open-source automation tools have won mindshare in a way they have not yet won integration directories, which is a useful reminder that a vendor's integration page is a lagging indicator of where builders actually are.

Most of an integration list is decoration

Here is the number that should change how you read a pricing page. Of the 3,507 distinct integration names we counted, 2,916 are named by exactly one product. That is 83 percent of the vocabulary appearing exactly once. Only ten names are shared by 100 products or more, and the top ten names account for 20.8 percent of all 9,601 mentions.

The distribution is a power law with a very short head and an enormous, almost entirely private tail. And once you know that, the marketing claim that a product has "5,000 integrations" resolves into something much less impressive: it usually means the product connects to Zapier, and Zapier connects to 5,000 things. The vendor built one integration and inherited a catalogue.

The tail also contains a lot of entries that are not integrations in any meaningful sense. Reading through the names that appear once, a substantial share are restatements of the same generic capability: "Webhooks and Data API", "Any processor via webhooks", "CRM systems via webhooks", "JSON webhooks", "REST API and webhooks", "SFTP file transfer". These are not connections to named partners. They are a vendor saying, correctly but unhelpfully, that data can get out if you write some code. A list padded with those is telling you the integration work is yours.

So the practical read is inverted from the intended one. A long integration list is weak evidence. What matters is whether the four or five systems you personally run appear as named, native connections, and everything past that is inventory.

Vendors build for a CRM small businesses do not buy

This is the sharpest thing in the data, and it deserves to be stated plainly. Here is how often each CRM is named as an integration across the 858 tools we track, all of which are, by our own inclusion rule, products a small business can actually adopt.

CRM	Products naming it	Share
HubSpot	305	35.5%
Salesforce	269	31.4%
Pipedrive	84	9.8%
Zoho CRM	33	3.8%
Microsoft Dynamics 365	13	1.5%
Close	9	1.0%
Attio	9	1.0%
Copper	3	0.3%
Insightly	2	0.2%
Nutshell	0	0%
Less Annoying CRM	0	0%

Salesforce is named by 31.4 percent of the small business tools in this database. Salesforce does not have a profile on this site, because it does not pass our inclusion rule: it is not a product a small team buys self-serve without a procurement cycle. The second most integrated destination in small business software is a product our readers are not supposed to be buying.

Meanwhile the CRMs built specifically for small teams are close to invisible to the rest of the ecosystem. We publish full profiles on Nutshell and Less Annoying CRM, both perfectly respectable products with real customers. Not one of the other 857 tools in this database names either of them as an integration. Close and Attio manage nine each, roughly one percent. Pipedrive, the most successful SMB-native CRM by some distance, gets 84, still less than a third of Salesforce's count.

We want to be fair about why. Integration engineering is expensive and vendors build against the largest addressable install base, which usually means the platform their enterprise prospects use rather than the one their smallest customers use. HubSpot's 35.5 percent is partly earned by having a genuinely free CRM tier that small teams adopt in volume, which is exactly why it tops this table rather than Salesforce. None of this is a conspiracy. It is a rational allocation of scarce engineering time, aggregated across hundreds of companies.

But the consequence lands on the buyer regardless of the cause, and it is a real cost that nobody quotes you. If you choose a challenger CRM on the merits, and there are good reasons to, you are also choosing a Zapier subscription, because native support for your CRM will not exist in most of the tools you later buy. That is a recurring bill, an extra failure point, and a latency penalty on every record. Price it into the comparison. The cheaper CRM with the better interface may still be right, but it is not as cheap as the pricing page says.

Zapier's presence is a diagnostic, not a feature

Zapier's overall 49.3 percent hides a much more interesting pattern. Broken out by category, Zapier support is close to universal in some markets and completely absent in others, and the split is not random.

Category	Products naming Zapier
Local SEO	11 of 12
CRM	20 of 22
Hiring and recruiting	7 of 8
AI voice agents	13 of 15
Email service providers	19 of 22
Lead capture	12 of 14
(36 categories in the middle of the range omitted)
Email deliverability	3 of 21
AI SDR	2 of 16
PPC ads	2 of 17
SEO content	2 of 23
GTM analytics	1 of 15
AI content writing	1 of 17
Customer data platforms	0 of 13

The rule that explains this is about the shape of the data, not the sophistication of the vendor. Zapier appears wherever a product emits discrete, low-volume business events: a booking was made, a candidate applied, a review was posted, a form was filled, a call was answered. Those are exactly the payloads a per-task automation tool handles well, and 92 percent of local SEO tools and 91 percent of CRMs live in that world.

Zapier disappears in two distinct situations, and it is worth separating them because they mean different things to a buyer. The first is where the product moves data in bulk, at which point per-task pricing becomes absurd. Not one of the 13 customer data platforms we track names Zapier, and they should not: a CDP is itself the pipe, and routing millions of events through a task-metered intermediary would be both slower and financially ridiculous. The same logic covers GTM analytics at one in fifteen. When a category has no Zapier support, that is often a sign the product is serious about volume.

The second is where the product's output is a human artifact rather than a record. One in seventeen AI content writing tools names Zapier, and two in 23 SEO content tools, because the thing they produce is a draft that a person reads, edits, and pastes somewhere. There is no event to forward. That is not a sign of seriousness about volume, it is a sign that the tool is a terminal in your workflow, and you should check what the export looks like rather than what the integration list says.

A third of the market is an island

Taking the four gravity wells together, 263 of 858 products, 30.7 percent, name none of Zapier, HubSpot, Slack, or Salesforce. Almost a third of small business software has no declared connection to any of the four systems the rest of the market organises itself around.

They cluster hard. In SEO content, 21 of 23 products are islands. In AI content writing, 15 of 17. In email deliverability, 18 of 21. Those three categories are where content and infrastructure live, and both produce outputs that are not records: a page of text, or a state of the world in which your domain authenticates correctly. HubSpot's category breakdown says the same thing from the other direction, with zero of 23 SEO content tools and zero of 21 deliverability tools naming it, against 13 of 14 digital sales rooms and 11 of 13 call tracking products.

Island status is not a defect. Some of the best-run products we cover are islands, and a deliverability monitor has no business writing to your CRM. But it does change the diligence question. For a connected tool you ask whether the integration is native or Zapier-mediated. For an island you ask a blunter question: when this stops being the right tool, how do I get my work out of it, and in what format?

How to read an integration list in ninety seconds

Four tests, in order, and none of them involve counting.

Find your own stack in the list. Not the total, not the logo wall, just the four or five systems you actually run. If they are not named individually, the number at the top of the page is irrelevant to you.

Check whether the named connection is native or brokered. If a product's answer to your CRM is Zapier, you have found a recurring cost and a failure point, not an integration. This matters most for the challenger CRMs above, where brokered is usually the only option available.

Look for Salesforce without Zapier. A hundred products in our database name Salesforce and do not name Zapier, and that combination is a reliable tell that the product's engineering priorities point upmarket, toward buyers with a Salesforce administrator. It does not make the product bad. It makes it a product where your requests will queue behind someone larger.

Count the generic entries. If a meaningful share of the list reads "webhooks", "REST API", or "any tool via webhooks", the vendor has told you the integration work is yours to do. Budget for it, or pick something else. We took apart the automation layer these lists lean on in our comparison of Zapier, Make, and n8n, and the cost of brokering everything through it is real once several tools depend on it.

The short version

We counted 9,601 integrations across 858 products, and the market turns out to be organised around four names. Zapier appears in 49.3 percent of products, HubSpot 35.5, Slack 32.9, Salesforce 31.4, and then the curve collapses. The integration count on a pricing page is close to meaningless, because 83 percent of the 3,507 distinct destinations we found are named by exactly one product and a long list usually means the vendor built one connection to Zapier and inherited a catalogue. The CRM finding is the one with money attached: Salesforce is named by 31.4 percent of small business tools while Nutshell and Less Annoying CRM are named by none, so buying a challenger CRM means buying a permanent automation dependency nobody quotes you. Zapier's presence in a category tells you what shape of data the product handles, universal where tools emit discrete events, absent where they move bulk streams or produce documents for humans. And a third of the market connects to none of the four, which is fine, as long as you have checked how your data gets out.

Every figure here is computed from the integration lists published in our own product dossiers and can be checked against them.