Introduction
Part 4 of 4 — Left Hook's Partner Experience Review. Being a marketplace partner is eight stages of work: decide, join, build, certify, launch, get demand, earn, and survive change. Most "best marketplace" rankings measure only the build. We score them separately, because they don't correlate.
- Developer experience across 41 marketplaces — what building there costs you
- MCP server maturity across the same 41 — whether they're ready for AI agents
- Partner-program transparency across 20 ecosystems — what they'll tell you, and what they'll do with you
- The best partner programs for app marketplaces (this post) — the ranked table, and what predicts a good one
This is the table people keep asking us for: if you're a software company deciding where to list an app, which marketplace partner programs are actually good to be in?
Two things make this different from the listicles. First, "good" here means legible — can you read the terms, compute your tier, and know what you'll be paid — because that's the part you can verify from outside, and it turns out to be the part most programs fail. Second, we tested two hypotheses about why some programs are better, and one of them died.
Short version of the finding: what predicts a well-documented app partner program is not the platform's age, its size, or the quality of its API. It's whether the marketplace is a profit centre or a checkbox. That difference is worth 1.7 points on a 5-point scale and it's statistically significant. Nothing else we tested is.
The Table
24 of the 41 marketplaces from our developer-experience study, scored on the five transparency checks from Part 3: is the program guide public, are requirements numeric, is the reward numeric, are rules of engagement published, and are changes dated. Yes = 1, Partial = 0.5, No = 0. Ranked by program score, then by DX.
"Marketplace" in the last column is our call on whether the marketplace is a core business line (a P&L someone owns, with its own economics) or a side channel (an integrations directory attached to the main product).
| # | Platform | Program | DX | Published rate | Category | Marketplace |
|---|---|---|---|---|---|---|
| 1 | Shopify | 5.0 | 8.5 | 0% to $1M lifetime, then 15% (+2.9% processing) | eCommerce | core |
| 2 | Salesforce | 4.5 | 7.5 | 15% of net subscription revenue; $999 review, ~$150/yr listing | CRM | core |
| 3 | Monday.com | 4.0 | 8.0 | 0% to $200K lifetime, then 15% | CRM | core |
| 4 | Atlassian | 3.5 | 7.0 | Forge 0% to $1M; benefits valued $2.5K–$300K | Collaboration | core |
| 5 | Zapier | 3.0 | 9.0 | — | Automation | core |
| 6 | Google Cloud | 3.0 | 7.5 | 3% / 2% ($1–10M) / 1.5% ($10M+) | Cloud | core |
| 7 | Azure | 3.0 | 7.0 | 3%, 10% reduced for co-sell | Cloud | core |
| 8 | AWS | 2.5 | 6.5 | 3% SaaS, 20% AMI | Cloud | core |
| 9 | GitHub | 2.0 | 8.5 | 5% (95/5) | DevTools | side |
| 10 | Zoom | 2.0 | 7.0 | 15% | Collaboration | side |
| 11 | WooCommerce | 2.0 | 6.0 | 30% (70/30) | eCommerce | side |
| 12 | Microsoft Teams | 2.0 | 6.0 | 3% | Collaboration | core |
| 13 | HubSpot | 1.5 | 8.5 | — | CRM | core |
| 14 | Workato | 1.5 | 7.0 | — ($10K/yr entry) | Automation | side |
| 15 | Stripe | 1.0 | 9.0 | — | eCommerce | side |
| 16 | Slack | 1.0 | 9.0 | — | Collaboration | side |
| 17 | Heroku | 1.0 | 8.5 | — | Cloud | side |
| 18 | Square | 1.0 | 7.5 | — (deferred to PIMA) | eCommerce | side |
| 19 | Vercel | 1.0 | 7.5 | — | Cloud | side |
| 20 | Netlify | 1.0 | 7.5 | — | Cloud | side |
| 21 | DigitalOcean | 1.0 | 6.0 | — | Cloud | side |
| 22 | Datadog | 0.5 | 7.5 | — (guide form-gated) | DevTools | side |
| 23 | Magento / Adobe Commerce | 0.5 | 4.0 | — (must sign PIMA) | eCommerce | side |
| 24 | Make | 0.0 | 5.5 | — | Automation | core |
If you want the one-line answer: Shopify, Salesforce, monday.com and Atlassian are the four app-marketplace programs where you can read the deal before you sign it. Everything below rank 8 requires a conversation to find out what you'll be paid.
A note on the two that will surprise people. Salesforce AppExchange lands second, which is not where its reputation puts it — the security review is genuinely expensive ($999 per paid-app submission, 6–9 weeks, and roughly half of first-time submissions fail) but all of it is published, including the 15% rate, the ~$150 annual listing fee, and the rules of engagement. Expensive and legible beats cheap and unknowable if you're building a business case. And Make scores 0.0 despite the marketplace being core to its business — the one clear exception to the pattern below, and the reason we're calling it a tendency rather than a law.
Are certain categories better at partnerships?
Mostly we can't tell yet, and I'd rather say that than publish a league table built on n=3.
| Category | Scored | Of | Avg | Verdict |
|---|---|---|---|---|
| CRM, Sales & Marketing | 3 | 8 | 3.33 | too thin (38%) |
| Collaboration & Productivity | 4 | 8 | 2.12 | too thin (50%) |
| eCommerce & Payments | 5 | 7 | 1.90 | borderline (71%) |
| Cloud & Infrastructure | 7 | 7 | 1.79 | full coverage |
| Data, Analytics, Automation & DevTools | 5 | 11 | 1.40 | too thin (45%) |
That CRM tops the list on 3 of 8 platforms is an artifact of which three we scored (Salesforce 4.5 and monday.com 4.0 are in; five others aren't). Quoting it as "CRM is best at partnerships" would be exactly the sin we just audited our own data for.
And the five categories above are themselves a coarse cut. Our own curated taxonomy runs to 138 software categories covering 1,305 ranked products — CRM and eCommerce sit alongside ATS, CLM, digital asset management, MSP-PSA, RMM, spend management, EDR, SIEM, password managers, backup and 120-odd others. Measured against that, this table covers the category leader in 14 of 138 categories, about 10%. The 41-platform set we started from is skewed toward large horizontal platforms, which is precisely where marketplace programs are most mature — so a category comparison drawn from it would be biased upward before we even started counting.
That's not a reason to distrust the 24 rows; each one is individually checkable. It's a reason not to generalise from them to "which categories are good at partnerships." Doing that properly means computing marketplace coverage per peer set across all 138 categories, which is a data job rather than a reading job, and it's the next thing we're building.
One category we can answer, because coverage is complete: cloud and infrastructure is systematically opaque about app economics. All seven are scored, and the pattern is clean — the four developer-platform players (Heroku, Vercel, Netlify, DigitalOcean) all score 1.0, every one of them declining to publish revenue terms at all, while only the three hyperscalers publish a rate (2.5–3.0). If you're listing on an infra marketplace that isn't AWS, Azure or GCP, assume the economics are a negotiation and price the uncertainty in.
The other honest observation is that category averages hide bimodality even where coverage is decent. eCommerce averages 1.90 across five platforms, but that's Shopify at 5.0 and Magento at 0.5 in the same bucket. There is no "eCommerce standard." There's Shopify, and there's everyone else.
Is it the leaders, or the newcomers?
Neither. We tested both, and the interesting result is that the obvious framing fails.
Age as a binary split does not hold up. Platforms founded before 2010 average 2.45; those founded 2010 or later average 1.64. That looks like a story, but at n=24 it's not statistically significant (Welch t=1.42, p=0.18). Treating founding year as a continuous variable does reach significance (r=-0.49, p=0.015), so there's some signal — but it's the weaker of the two effects and, as below, it's mostly downstream of something else.
Business model is the variable that actually separates them, and it's significant:
| n | Avg program score | |
|---|---|---|
| Marketplace is a core business line | 11 | 2.91 |
| Marketplace is a side channel | 13 | 1.19 |
Welch t=3.82, p=0.0023. That's a 1.72-point gap on a 5-point scale — more than double the age gap, at a tenth the p-value.
The mechanism is unglamorous and convincing. You write the rules down when the rules start costing you money. Once a marketplace has thousands of partners, real channel conflict, and a revenue line an executive is accountable for, ambiguity becomes expensive: you need published tiers so partners stop asking, published rules of engagement so deals stop getting fought over, and a published rate so finance can forecast. A platform whose marketplace is a directory attached to the main product never hits that forcing function, so the terms stay in a PDF someone emails you.
Age correlates because older companies are more likely to have crossed that threshold — not because maturity makes companies more forthcoming. The counterexamples make this concrete: monday.com (founded 2012) scores 4.0 while HubSpot (2006) scores 1.5. Microsoft Teams (2017) beats Stripe (2010) and Heroku (2007). Age doesn't sort these; business model does.
What this means practically: before you evaluate a program, work out whether the marketplace is a P&L or a checkbox. If it's a checkbox, expect an unpublished rate, expect to negotiate, and expect the terms to change without a changelog — not because the company is badly run, but because nobody there is accountable for your economics.
The best builders are not the best partners
The strongest single pattern in the table is a disconnect. Look at rows 15–17: Stripe, Slack and Heroku carry DX scores of 9.0, 9.0 and 8.5 — three of the five best developer experiences in the entire 41-platform study — and program scores of 1.0.
Across all 24, program score and DX score are statistically unrelated (r=+0.27, p=0.21). That confirms what we argued in Part 3, now with the numbers behind it: great API design and a legible partner program are independent investments, made by different teams, on different timelines, for different reasons. Stripe's docs are an industry benchmark and its app-marketplace economics are unpublished. Both facts are stable and neither predicts the other.
The practical version: do not use a DX ranking to pick a partnership, and do not read a great API as evidence of a serious ecosystem commitment. They're separate diligence questions. Ours are now separate posts for exactly that reason.
What isn't scored yet
17 of the 41 have no app-side program score, and the gaps are not random — they're concentrated in the categories where we'd most like to compare:
- CRM, Sales & Marketing: Zendesk, Pipedrive, Freshworks, Dynamics 365, Zoho
- Collaboration & Productivity: Notion, Figma, Asana, Airtable
- Data / DevTools: Twilio, Okta, Intercom, Snowflake, ServiceNow, Databricks
- eCommerce: BigCommerce, PayPal
For several of these we found no published app revenue terms at all, which likely means a low score — but absence of evidence in a search is not evidence of absence, so they're recorded as unscored rather than zero.
Closing those 17 would complete this table. It would still not answer the category question, because that needs the other 116 category leaders too — Mailchimp, QuickBooks, Gusto, DocuSign, Calendly, Greenhouse, Segment, Tableau, CrowdStrike, Splunk, ConnectWise PSA, NinjaOne and the rest. That's the work we're pointing our own tooling at rather than reading program pages one at a time.
How to use this
Rank by legibility, then negotiate the rest. A published rate is worth more than a favourable rumoured one, because you can model it and you can hold them to it.
Check the business model before the program page. Core or checkbox is the single most predictive thing you can establish in five minutes, and it tells you what kind of relationship you're entering.
Price the review, not just the rate. Salesforce at 15% plus $999 and 6–9 weeks with a ~50% first-pass failure rate may still beat an unpublished rate elsewhere. The rate is the visible cost; the gate is the one that slips quarters.
Find the retention or health floor. Across Parts 3 and 4 the requirement that moved most in 2026 wasn't revenue, it was retention and integration health — HubSpot's 80% GRR for Elite, Zapier's health score, monday.com's 88–90% GDR. Those bind harder than the take rate.
Score both sides if you're both. Atlassian is among the most transparent ecosystems for app vendors and among the least for service partners. One vendor, opposite answers.
Method, and where to argue with us
Program scores are the five binary checks from Part 3, applied to the app/ISV-side program only (service-partner programs are scored separately in Part 3). Yes = 1, Partial = 0.5, No or not-found = 0. No weighting — so if you disagree with a row, you can point at the specific check.
DX scores are from Part 1. Category groupings are Part 1's. Founding years are public record. The core/side classification is our judgement, and it's the one input a reasonable person could reallocate — we'd note that moving any single platform between core and side does not overturn a p=0.0023 result, but moving several would, and we'd want to hear the argument.
Statistics: Welch's t-test for two-group comparisons, Pearson correlation for continuous ones, n=24 throughout. At this n these are directional findings about a specific 24-platform sample, not laws of ecosystems. The category comparison is explicitly not reportable at current coverage, and we've said which cells are thin rather than averaging over the gap.
Where we'd most like to be corrected: any program in the "—" column of the table that does publish its rate somewhere we didn't find. Send us the link and we'll rescore it.
Program terms verified against primary sources — vendor developer documentation, program guides, partner agreements and published policy documents — in August 2026. Several programs in this table changed materially in the first half of 2026; verify current terms before committing. Scores measure documentation transparency, not program quality or partner outcomes.