The Rival Ecosystem-Move Read: Co-Sell, Built, Listed Only, or Lapsed
Score any single change on a rival's partner or marketplace page on five dimensions and land on one dated verdict another analyst reproduces blind.
Key takeaways
- A Microsoft Azure IP co-sell eligible badge requires at least $100,000 of Azure consumed revenue or Marketplace billed sales over the trailing 12 months. Co-sell ready is a separate, lower status that does not carry that threshold.
- A HubSpot Certified badge requires at least 60 active installs, each a unique unaffiliated production account with successful activity in the last 30 days. That makes it the closest public proof that an integration is actually used.
- When a product is removed from AWS Marketplace, it disappears from search but its direct URL still resolves, so check whether a listing is still live through search and browse, never through a stored link.
- At most 31.3% of URLs are archived more than once a month, so report a rival's first-seen date as a window between two captures, not as a single day.
- In Refolk's index, only 7 of 788 US alliance managers (0.9%) name AWS in their headline. Finding a named profile is strong evidence of staffing, and not finding one is weak evidence of anything.
- Apply the verdict rules in a fixed order with Lapsed first, and have a second analyst grade blind. The job is done when both analysts agree on the verdict and the confidence.
A rival just added a partner logo, an integration tile, a marketplace listing, or a new partner tier, and leadership wants to know whether it is a real channel or product commitment or just a page edit. This framework is for the strategy, research, and talent-intelligence analysts who make that call. You score one change on five evidence dimensions and give it one of four verdicts, with a first-seen date window and a stated confidence, so two analysts grading the same change blind reach the same answer.
Why a page-change alert is not a verdict
A page-change alert proves that something on a rival's partner page moved. It says nothing about what the move commits the rival to. The job is to turn each alert into one sentence you can defend: what changed, what it proves, since when, and how sure you are.
Most public advice on how to track competitor integrations and partnerships ends at "monitor the integrations page and set an alert." That gets you volume, not judgement. One page-monitoring vendor reports 1,395 active monitors on partner pages detecting 409 changes in 30 days, or about 0.29 alerts per monitor per month. Spread that across every rival's integrations directory, partner directory, tier page, and marketplace listings, and the research team ends up forwarding logo adds that mean nothing.
The fix is to grade each change instead of passing it on. Every change gets exactly one of four verdicts:
| Verdict | What it means | What you brief |
|---|---|---|
| Co-Sell Commitment | Audited co-sell tier, or joint selling backed by named staff | A channel move |
| Built Integration | Shipped, maintained connector, or an audited install base | A product investment |
| Listed Only | At most an entry gate passed, with no build or joint selling found | A page change, logged |
| Lapsed | Gone from discovery, badge lost, or tile removed | A retreat, dated |
Three terms come up throughout. A listing is an entry on a platform marketplace or on the rival's own integrations page. A tier or badge is a status a platform grants only after it has checked something, such as revenue, installs, or pipeline. First-seen is the earliest point you can show the item existed, and it is always a window, never a single day.
What does each marketplace tier actually prove?
An audited badge proves that a published threshold was met. A logo on the rival's own site proves only that someone edited a page. None of the four major platforms below lets anyone publish a public listing entirely on their own, so even an entry listing shows that a review took place. That entry review is shallow compared with what the top tier demands.
| Platform | Entry gate | Top tier | Top-tier threshold |
|---|---|---|---|
| Microsoft | Published offer, partner profile | Azure IP co-sell eligible | $100K consumed revenue or billed sales, trailing 12 months |
| AWS | GA Marketplace listing | ISV Accelerate | 15 qualified + 5 launched opportunities, $2,000 revenue, 1 person trained |
| HubSpot | 3 active installs, manual review | Certified App | 60 active installs, 6 months listed |
| Salesforce | Security review | Not retrieved | Not retrieved |
A logo certifies an edit. A badge certifies a threshold that someone else audited.
Microsoft: three statuses, often confused
Microsoft publishes three statuses, and analysts most often overstate the middle one.
- In market: the solution is linked to a live Marketplace offer, but it has not met the co-sell-ready requirements.
- Co-sell ready: the solution meets Microsoft's co-sell-ready requirements. This is a readiness status, not the tier that audits revenue.
- Azure IP co-sell eligible: at least $100,000 of Azure Consumed Revenue or Marketplace Billed Sales over the trailing 12 months, plus Marketplace policies met, primary platforming on Azure, and technical validation passed. New offers must be transactable on the Marketplace, and Azure credits or ACO do not count toward the $100,000.
Two things follow. First, IP co-sell eligible is the strongest public commercial signal in this guide, because it means six figures of trailing Azure revenue that Microsoft has checked. Second, apps and add-ins for Microsoft 365 (Teams, Outlook, Excel) cannot get co-sell ready status. A rival's Teams app without the badge is therefore not a negative signal, because the badge was never available to it.
AWS: verify before you lean on it
The ISV Accelerate criteria here come from secondary sources, not from AWS's own pages. According to those sources, the tier requires:
- Validated or Differentiated status
- A GA Marketplace product and Payee Central
- At least 5 launched and 15 qualified ACE opportunities in 12 months
- At least one person trained in Co-Selling with AWS
- At least $2,000 in recognised AWS revenue
Members reportedly also pay lower Marketplace fees (1.5% against 3%), which gives rivals a reason to chase the tier even when the partnership itself is thin. Check AWS's current partner documentation before you brief ISV Accelerate as audited.
HubSpot: installs, audited
The HubSpot Ecosystem Quality team reviews every listing submission by hand. Certification needs at least 6 months listed and at least 60 active installs, and certified apps must be recertified every two years. HubSpot defines an active install as a unique production account, not affiliated with the developer, that has shown successful app activity in the last 30 days. That makes the Certified badge the closest thing any platform publishes to proof that an integration is used, not just built.
The threshold is time-sensitive. HubSpot has raised it before (from 6 to 60 active installs) and is changing its submission flow to check eligibility against active installs rather than total installs. Re-read the requirements page before you quote a number.
Salesforce AppExchange: a slow gate
An app has to pass the AppExchange Security Review before it can be listed publicly. Estimates of how long the review takes vary from 6 to 9 weeks in one source to 8 to 16 weeks in another. The useful conclusion: a new public AppExchange listing means the rival put weeks of engineering in before the listing appeared. I did not retrieve the top-tier criteria for this guide, and secondary sources quote the review fee inconsistently, so treat both as unverified.
What each layer of evidence proves
- Logo or tile on the rival's pageProves an edit, nothing more
- Platform listingPassed a manual review or security gate
- Audited badge or tierMet a revenue, pipeline, or install threshold
- Build artifactsSomeone ships and maintains the connector
- Named staff and joint customersSomeone is paid to make it sell
The five evidence dimensions and how to score them
Score every change 0, 1, or 2 on five dimensions: Tier, Build, Go-to-market, People, and Liveness. Then apply four verdict rules in a fixed order. The fixed order is what makes two analysts reach the same verdict, and it is the core of any defensible competitor partner ecosystem analysis.
| Dimension | Score 0 | Score 1 | Score 2 |
|---|---|---|---|
| Tier | Logo on rival's page only | Entry listing, or co-sell ready | IP co-sell eligible, ISV Accelerate, or HubSpot Certified |
| Build | No artifacts found | One artifact type, or none dated in 90 days | Two or more types dated in 90 days |
| Go-to-market | Logo only | One of: press release, case study, co-branded page | Joint case study plus page or release |
| People | None found | Unnamed alliance staff, or a job post names the partner | Named staff tied to the partner |
| Liveness | Gone from discovery, badge lost, tile removed | Present, but claimed status unconfirmed | In search and browse, badge intact |
What each dimension proves, and how it lies
Tier.
- Proves: the platform audited something, whether revenue, pipeline, installs, or security.
- How it lies: co-sell ready gets read as IP co-sell eligible, or a badge is assumed to last forever. A HubSpot app that loses certification stays listed, just without the badge.
Build. The artifacts to look for are a connector repository with recent commits, versioned docs, changelog entries, and support articles.
- Proves: someone ships and maintains the integration.
- How it lies: the connector belongs to the partner or a third party, not the rival, or the "docs" are one marketing paragraph. No platform or study publishes a list of artifacts that proves an integration is built. The share of directory entries that are third-party-built or "coming soon" is not established either. So record who owns each artifact, not just whether it exists.
Go-to-market. The artifacts to look for are a joint press release, a joint case study, and a co-branded landing page. One widely read monitoring guide treats a logo backed by all three as a real alliance.
- Proves: both companies put their names on selling together. A joint case study also proves at least one shared customer.
- How it lies: the press release comes from the rival alone, or the "case study" is just a quote on a logo wall.
People.
- Proves: someone is paid to make the partnership sell.
- How it lies: absence is read as proof of no staffing. Named-partner headlines are rare (see the people section below), so a 0 here means "not found", never "not staffed".
Liveness.
- Proves: buyers can still find the item where they actually look.
- How it lies: you check the stored direct URL. A product removed from AWS Marketplace drops out of search, but its details page still loads.
The verdict rules, in order
- Lapsed if Liveness is 0. This overrides everything else. A delisted co-sell is a retreat, not a commitment.
- Co-Sell Commitment if Tier is 2 through a co-sell tier (IP co-sell eligible or ISV Accelerate), or if Go-to-market is 2 and People is at least 1.
- Built Integration if Build is 2, or if Tier is 2 through HubSpot Certified.
- Listed Only for everything else.
Confidence
- High: the verdict rests on an audited platform status or on two independently dated artifacts, and the first-seen window is 31 days or narrower.
- Medium: only one of those two conditions holds.
- Low: neither holds, or the verdict depends on a dimension that was scored from an absence.
Where a change lands before the liveness check
How do you date when a partner logo first appeared?
Bracket the first-seen date between the last archived capture without the item and the first capture with it, and report that window. A single date is false precision. A window with both ends cited is defensible.
The tool is the Wayback Machine's CDX server, which serves the index the Wayback Machine itself uses to look up captures. CDX timestamps record capture time in UTC, to the second. Adding collapse=digest drops consecutive captures with identical content, so the output is a list of the moments the page actually changed.
https://web.archive.org/cdx/search/cdx?url=rival-domain.com/integrations&fl=timestamp,statuscode&collapse=digest Read the output as: Window start = last timestamp whose capture does NOT contain the item Window end = first timestamp whose capture DOES contain the item Record both in UTC, plus the status code of each capture.
Replace the URL with the exact rival page you are dating, one page per query.
Then open the captures on either side of the change and confirm that the item itself changed. A new digest means something on the page changed, not necessarily the tile you care about.
How wide your window will be
How tightly you can date a change depends on how often the archive captured the page. The best published estimate of archive coverage across URIs is:
| Archived copies | Share of URIs |
|---|---|
| At least 1 | 35% to 90% |
| 2 to 5 | 17% to 49% |
| 6 to 10 | 1% to 8% |
| More than 10 | 8% to 63% |
| Captured more than monthly | 14.6% to 31.3% |
At most 31.3% of URIs are captured more than once a month, so for most pages your window will be a month or wider. The archive also crawls popular sites more often and more completely, so mid-tier rivals usually get wider brackets than market leaders. Capture frequency for mid-tier vendor pages specifically is not established publicly. To check it for a given page, count the captures in your own CDX output over the past year.
When the archive makes an item look newer than it is
Two effects push first-seen later than the truth:
- JavaScript-rendered tiles may be missing from snapshots. In one validation study of archived pages, roughly 40% to 51% of live tracking-pixel installations did not appear in the archived snapshots. Integration tiles loaded by script can disappear the same way.
- Sites that block the archive's crawlers are covered less well. A rival that blocks crawlers will look as if it changes less often than it really does.
The check: search the archived HTML for the tile's markup. If the tile markup never shows up in any snapshot, including recent snapshots taken while the item was definitely live, the archive cannot date this item. Say so in the record and fall back on your own capture timestamp.
What the date cannot tell you
How far an ecosystem change runs ahead of or behind a public announcement is not established publicly. One monitoring vendor claims its alerts let analysts act before an announcement lands, but it gives no figure. Do not brief a lead time. Record both the first-seen window and the announcement date, and let the gap build up in your own log.
Reading people evidence without over-reading it
Named alliance staff are the strongest public sign that a partnership has real resources behind it. They are also rare, so finding one counts for a lot and not finding one counts for almost nothing. No published threshold separates a staffed partnership from a nominal one. The only documented platform minimum is AWS's requirement that at least one person complete the Co-Selling with AWS learning module.
In Refolk's index of professional profiles, a single query for the titles Alliance Manager, Alliances Manager, Strategic Alliance Manager, and Partner Alliance Manager returns:
| Segment | Profiles | Derived |
|---|---|---|
| US, alliance titles | 788 | Baseline |
| UK, alliance titles | 143 | US pool is 5.5x the UK |
| US, alliance titles naming AWS in headline | 7 | 0.9% of US |
How rare partner-named alliance staff are
- 788US alliance-titled profiles
Four alliance title variants
- 7Headline names AWS
0.9% of the pool
- 2At Arctic Wolf
The only employer with more than one
Three things follow for competitor channel partner analysis:
- Partner-named headlines are specific. People name a partner in their headline when they are dedicated to that partner. Arctic Wolf alone accounts for 2 of the 7 AWS-named US profiles, while Okta, OpenText, Fortreum, Vertical Relevance, and Aziro have 1 each. Finding even one such profile is a strong People score.
- Set thresholds per country. The US pool is 5.5 times the UK pool. The UK sample also leans toward systems integrators (Capgemini, EPAM Systems, IBM), next to software firms such as Zendesk, SAP Concur, and Gearset. Two named alliance staff at a UK rival is a different signal from two at a US rival.
- Titles are noisy. The same title list picks up life-sciences alliance roles in the US (Thermo Fisher, New England Biolabs) and off-target titles in the UK, such as "Online Tutor" and oil trading. Filter by industry before you count anything.
Treat these counts as a snapshot, not as validated thresholds. Re-run the query before you rely on them.
Doing this with keyword search across profile sites means guessing every title variant, then filtering out biotech by hand. With Refolk you describe the people you want in plain English and get matching profiles back across LinkedIn, GitHub, and the open web. That also covers the Build check when the question is who maintains a rival's connector repository.
Check job posts as well as profiles. An open alliance or partner role at either company that names the partner scores People 1, even when no current employee names it.
The procedure: grading one ecosystem change
Grading one change takes eight steps and roughly two hours of analyst time, most of it spent on archive dating and build checks. It also needs a blind second grade and a re-check 90 days later. Use the same procedure for an integrations tile, a partner-directory entry, a new partner tier, or a competitor marketplace listing change.
Grade one rival ecosystem change
- Capture the diffSave the live page and a dated screenshot, about 10 minutes. Done when the change is logged with its URL and a UTC timestamp.
- Bracket first-seen in the archiveRun a CDX query with collapse=digest and record the last capture without the item and the first capture with it, 20 to 30 minutes. Done when you have a date window with both ends cited.
- Classify the listing tierRead the badge or status (In market, co-sell ready, IP co-sell eligible, HubSpot Certified, AppExchange listed), about 15 minutes. Done when Tier is scored against a named published threshold.
- Check the build evidenceLook for a connector repository with recent commits, versioned docs, changelog entries, and support articles, about 30 minutes. Done when each is marked present or absent, with a date and an owner.
- Check go-to-market evidenceSearch both companies' sites for a joint press release, a joint case study, and a co-branded landing page. Done when each is marked present or absent and Go-to-market is scored.
- Check people evidenceQuery Refolk's index for alliance or partner roles naming the partner at either company, and check both companies' job posts, about 15 minutes. Done when People is scored, with any 0 recorded as not found.
- Score and assign a verdictApply the four verdict rules in order, Lapsed first, and set confidence, while a second analyst grades the same change blind. Done when both agree on verdict and confidence.
- Re-check after 90 daysReturn to the item through platform search and browse, not the stored URL, and re-score Liveness. Done when the item is confirmed live or regraded as Lapsed with a lapse window.
A few notes on running it:
- Blind means blind. Give the second analyst only the page URL and the logged UTC timestamp, not your scores or your draft sentence. If you disagree, compare dimension by dimension. The disagreement is almost always one score, and usually Tier or Liveness.
- Search both sides. Build and go-to-market evidence often sits on the partner's site, not the rival's. A partner that lists the rival on its own directory is independent corroboration.
- Stop early when the answer is clear. If Liveness is 0 at step 3, you still date the item, but the build and people checks only add context to a Lapsed verdict.
How this read goes wrong
Most wrong verdicts come from two errors: treating a resolving URL or a listed app as live, and treating a single archive capture as an exact date. Each failure mode below has a check that catches it.
- Lapse missed on AWS. The direct URL still resolves after a product is delisted, so the item looks live. Check: search and browse the marketplace instead of loading the stored URL.
- Lapse missed on HubSpot. A decertified app keeps its listing and only loses the badge. Check: look for the badge itself, not just whether the app is listed.
- First-seen dated too early or too late. JavaScript-rendered tiles may be missing from snapshots, and sites that block the archive's crawlers are covered less well, so an item can look newer than it is. Check: confirm the archived HTML contains the tile markup, and report a window, not a single date.
- Co-sell status overstated. Co-sell ready gets treated as IP co-sell eligible, which turns a readiness status into a claimed six-figure revenue line. Check: Microsoft's exact status label, word for word.
- Conflicting definitions of an install. One public guide says sandboxes, real customers, and beta installs all count, which contradicts HubSpot's rule of unaffiliated production accounts with activity in the last 30 days. Check: HubSpot's own docs win, every time.
- Unverified criteria quoted as fact. Secondary sources quote the Salesforce security-review fee as both $999 and $2,700, and the AWS ISV Accelerate criteria in this guide come only from secondary sources. Check: mark both as unverified in the brief until you have confirmed them against the platform.
- A missing name read as no staffing. With only 7 of 788 US alliance profiles naming AWS, a missing name is not evidence that nobody works the partnership. Check: search job posts at both companies, and record People 0 as "not found".
- Title noise inflating counts. Alliance titles pick up biotech and off-target roles. Check: filter by industry before counting, and spot-check a sample of the profiles by hand.
- Someone else's connector credited to the rival. The integration exists, but the partner or a third party built it. Check: record the owner of every build artifact. A partner-built connector supports the Tier and Go-to-market scores, not the rival's Build score.
Keeping the read current
Re-check every graded item at 90 days, through platform search and browse rather than the stored URL, because platforms are built to let lapses happen quietly. A verdict is a dated claim, not a permanent label.
Why 90 days works as the cadence:
- AWS: sellers must support existing customers for at least 90 days after removal, and AWS can disable a seller's ability to offer on the Marketplace with 30 days' written notice.
- HubSpot: apps with fewer than three active installs may be removed, and certified apps must recertify every two years.
- Everything else: no general purge cadence for partner directories is established publicly. Treat 90 days as a working cadence, not a platform rule.
On monitoring frequency, the sources disagree: one recommends checking the integrations directory daily, another bi-weekly, and a third lists partner pages as weekly. My view is that weekly is enough for most partner and tier pages, because the archive bracket is usually a month or wider anyway. Keep daily checks for the few rivals whose integration moves leadership acts on within days. Whatever the frequency, every alert to monitor a competitor partner directory should end in a verdict record, not a forwarded diff.
Platform thresholds change too. Microsoft added the transactable-offer requirement for new offers, and HubSpot has raised its install bar before. Re-read the four platforms' requirement pages each quarter and re-score any item whose Tier score depended on a threshold that moved. When you re-check People, re-run the same Refolk query you used the first time, so changes in the count reflect the market and not your wording.
Keep each verdict history rather than overwriting it. A rival that moves from Listed Only to Built Integration to Co-Sell Commitment over three re-checks is telling you something no single verdict shows.
Change: [what changed, one line] Page: [URL] | Logged: [UTC timestamp] First-seen window: [last capture without, UTC] to [first capture with, UTC] Scores: Tier _ | Build _ | Go-to-market _ | People _ | Liveness _ Verdict: [Co-Sell Commitment / Built Integration / Listed Only / Lapsed] Confidence: [High / Medium / Low] because [one clause] Unverified inputs: [any secondary-source criteria or fees relied on] Second grader: [name], blind, [agreed / resolved by: ...] Re-check due: [logged date + 90 days] Brief sentence: [Rival] shows a [verdict] with [partner], based on [strongest evidence], first seen between [date] and [date], [confidence] confidence.
One record per change; paste the sentence into the leadership brief unchanged.
Before you brief the verdict
- The change is logged with its URL and a UTC timestamp.
- First-seen is a window, with both capture timestamps cited.
- The archived HTML was checked for the tile markup, not just a digest change.
- The listing status is mapped to a published threshold, and co-sell ready is not scored as IP co-sell eligible.
- Every build artifact is marked present or absent, with a date and an owner.
- Liveness was checked through platform search and browse, not the stored URL.
- Any People score of 0 is recorded as not found, and job posts at both companies were checked.
- A second analyst graded blind, and the verdict and confidence match.
- Every criterion or fee taken from secondary sources is flagged as unverified in the brief.
- A 90-day re-check is on the calendar.
Questions practitioners ask
How do I tell if a competitor's partnership announcement is real?
Look for evidence that a platform or the partner had to sign off. A real partnership usually shows some of these: an audited tier such as Azure IP co-sell eligible or HubSpot Certified, shipped build artifacts, a joint case study with a co-branded page or press release, and named alliance staff. A logo on the rival's own page only proves that someone edited it. Score each dimension separately and let the verdict rules decide, not the press release.
How often should I monitor a competitor's partner directory?
Weekly is enough for most partner and tier pages. Monitoring sources disagree and recommend daily, weekly, or bi-weekly checks. The real limit is the archive, though: at most 31.3% of URLs are captured more than once a month, so your first-seen window is often a month wide whatever you do. Check daily only for the rivals whose integration moves leadership acts on within days. Re-check every graded item at 90 days.
Is Microsoft co-sell ready the same as IP co-sell eligible?
No. Azure IP co-sell eligible is the higher status. It requires at least $100,000 of Azure Consumed Revenue or Marketplace Billed Sales over the trailing 12 months, a transactable offer, primary platforming on Azure, and technical validation, and Azure credits do not count toward the $100,000. Scoring co-sell ready as the top tier is a common way to overstate a rival's commitment. Microsoft 365 apps cannot get co-sell ready status at all.
Why does a delisted AWS Marketplace product still load?
AWS designs removal to be quiet. A removed product drops out of search and discovery tools, but its product details page can still be reached by direct URL, and sellers must support existing customers for at least 90 days. A bookmark or page monitor pointed at the stored URL will keep reporting the product as live. To confirm it is really live, search and browse the marketplace the way a buyer would.
Can the Wayback Machine give me the exact date a partner logo appeared?
No. It gives you a window. Query the CDX index with collapse=digest, take the last capture without the item and the first capture with it, and report both. These windows are often wide: between 35% and 90% of URLs have any archived copy at all, and at most 31.3% are captured more than once a month. Tiles drawn by JavaScript may also be missing from snapshots, which makes items look newer than they are.
Does finding no alliance manager for a partner mean the partnership is unstaffed?
No. In Refolk's index, only 7 of 788 US alliance-titled profiles name AWS in their headline, so most people who work a partnership never name it publicly. Score People as 0 meaning 'not found', check both companies' job posts, and never let an absence on its own push a verdict down. A named profile is strong evidence. A missing one is weak evidence.
Try it on the search you came here for
Stop building boolean strings. Just describe the person.
Type one sentence. I plan the search, read GitHub, public LinkedIn and Crunchbase records, and the open web as it is right now, and hand back a ranked list with the reason next to every name.
01Describe them
One plain sentence. Role, city, stack, stage, whatever matters to you.
02I read the web live
GitHub, public LinkedIn and Crunchbase records, the open web. Not a database that went stale last quarter.
03You read the shortlist
Ranked, with the reasoning under every name. Open a profile, ask a follow-up, narrow it down.
- Staff backend engineers in NYC who shipped Rust in production
- Series A fintechs in SF under 50 people, growing headcount this year
- Maintainers of fast-growing Rust web frameworks on GitHub
- No boolean, no filters, no seat to buy. One box.
- Read at search time, so a profile updated yesterday counts today.
- Every step visible as it runs, every name with its reason.
500 free credits on sign-up. No card, no demo call. See real searches.