Most lists of companies likely to churn from a competitor’s product are built from a single fact: the account currently uses that competitor. That is a targeting filter, not a timing signal, which is why those lists tend to perform about as well as a cold firmographic pull. A possible platform switch leaves a two-sided evidence trail instead: technology detections shift, and the roles a company posts shift with them. This guide shows how to read hiring and technology signals together, how to score what you find, and the four places the pattern will lie to you.
TLDR:
- A platform switch shows up as a pattern, not a record: a new technology detection appearing while an incumbent detection goes quiet, alongside job postings that describe migration work in the same window.
- A missing technology detection is never proof of churn on its own, because script changes, recrawl gaps, and signature changes produce exactly the same gap as a real replacement.
- Job postings are the strongest corroborating evidence available, because employers describe migration work in plain language: data migration, deduplication, multi-org consolidation, and hybrid architecture.
- Score the evidence in tiers instead of filtering on a single boolean. Only the top tier, two independent signals moving the same direction in the same quarter, justifies interrupting a rep.
- PredictLeads Technology Detections draws on five independent sources including job descriptions, across 1.5B+ detections since 2018 and 50,000+ tracked technologies, which is why a posting and a detection can corroborate each other inside one dataset.
What a Platform Switch Signal Is (and What It Is Not)
A platform switch signal is a time-ordered pattern across two or more independent data sources that is consistent with a company replacing one tool with another. It is a pattern, never a single record. The minimum useful version is a new technology detection appearing in a category while the incumbent tool in that same category stops being detected, with a job posting in the same window describing the work of moving between them.
What the signal is not matters more than the definition, because this is where displacement programs lose credibility with their own sales team:
- Not a cancellation record. No public source publishes contract terminations. Public data shows evidence of use, and the absence of that evidence, nothing about the paperwork.
- Not a confirmed decision. Evaluations get abandoned, pilots get shelved, and migrations get paused for a budget cycle. The pattern tells you a conversation is probably happening inside the account.
- Not proof of which vendor won. Two tools in the same category frequently coexist for a full renewal cycle or permanently, split across business units.
- Not personal data. Every signal described here is company-level and drawn from public sources: websites, career pages, DNS records, and job descriptions.
Read this as the timing-focused companion to the broader playbook on technographic data and competitor displacement, which covers list construction, segmentation, and messaging. This post deals only with the narrow question of whether an account looks like it is moving.
Why One Missing Detection Is Not Evidence of Churn
A technology that is no longer detected has not necessarily been removed. Every technology detection carries first_seen_at and last_seen_at, so you can see exactly when evidence of a tool last appeared for a company. What the timestamp cannot tell you is why the evidence stopped. The correct phrasing in your CRM field, your alert copy, and your rep’s talk track is “not detected since [date],” never “dropped” or “canceled.”
| Possible cause | What it looks like in the data | How to test it |
|---|---|---|
| Actual replacement | Several sources go quiet within a short period, and a competing technology in the same category gets a new first_seen_at. |
Check whether a same-category detection appeared, then look for hiring that describes the move. |
| Script or tag change | A single website-based source goes quiet while job-description evidence continues. | Compare source_count before and after, and check detection_source_type. |
| Recrawl gap | Everything for that company goes quiet at once, across unrelated technologies. | Look at the company’s other detections. A company-wide gap is a crawl artifact, not a stack change. |
| Signature change | A vendor changes how its product identifies itself, and the gap appears across many companies on the same date. | Check whether the drop-off is account-specific or population-wide for that technology. |
Two fields keep you honest. source_count tells you how many independent sources contributed to a detection, so a tool evidenced by one script tag disappears the moment that tag changes, while a tool evidenced by four sources going quiet together is a materially different event. Extended Technology Detections adds detection_source and detection_source_type, which resolves to values like script_src, cookies, headers, or keyword, so you can inspect what the evidence actually was before you decide what its absence means. If you want the longer treatment of this problem, see the guide on technographic data accuracy.
The Two Signals That Corroborate Each Other
Technology Detections and Job Openings corroborate each other because they are collected from different places and fail in different ways. Detections come from website script tags, DNS records, IP ranges, cookies, and job descriptions. A website redesign breaks the first of those and leaves the rest untouched. A hiring freeze quiets the job signal and leaves the website evidence intact. When both move in the same direction inside the same quarter, the odds that you are looking at a crawl artifact drop sharply.
Job descriptions being one of the five detection sources is the structural reason this works. Enterprise tools that never touch the front end of a website, the ones that matter most in displacement plays, are frequently detected through hiring rather than through page markup. Those detections are flagged with behind_firewall, and the connection between the two datasets is covered in more depth in job openings data and technographics.
| Signal | Evidence of | Cannot tell you alone |
|---|---|---|
New detection (first_seen_at inside the window) |
New or newly visible use of a technology | Whether anything was replaced, or whether the tool is simply newly visible |
Aging detection (last_seen_at going stale) |
Evidence of the tool is no longer appearing | Whether the tool was removed, hidden, or just missed on recrawl |
| Posting with migration language | Planned or in-flight work on the stack | Which direction the migration runs, or whether it will finish |
| Posting listing the incumbent as a required skill | The employer wants that skill on the team | Whether the tool is in production, in evaluation, or supporting a client |
News Event (integrates_with, partners_with) |
A publicly announced vendor relationship | The scope of deployment, or what it displaces |
What Migration Language Looks Like in a Job Posting
Migration work gets described plainly in job postings, because an employer has to describe the work in order to fill the role. That plain description is the most readable corroborating evidence available to a GTM team, and it is public.
Two live examples from the dataset show the shape of it. A Senior Salesforce Administrator posting found on September 10, 2026 lists data migration, deduplication, and multi-org consolidation among its responsibilities, and carries Agentforce, Salesforce’s own AI agent capability, among its tagged skills. A separate Salesforce Administrator posting found on September 9, 2026 describes a hybrid Government Cloud and Commercial environment with migration and compliance responsibility. Neither posting establishes that its employer is leaving any particular vendor, and neither should be read that way. What they establish is the vocabulary you are pattern matching against.
Four phrase families do most of the work:
- Movement: data migration, cutover, legacy system, sunset, decommission, “migrate from.”
- Consolidation: multi-org consolidation, deduplication, merging instances, single source of truth, “consolidate onto.”
- Coexistence: hybrid architecture, dual-run, integration between two named systems, phased rollout.
- New capability: a named feature that only exists in one vendor’s product line, which tells you what the team is being staffed to operate.
Seniority sharpens this considerably. A posting at manager or director level that names a migration usually means the program has a budget owner, while a junior posting listing the same tool as a skill more often means business as usual. Because every job opening carries an O*NET occupation code, you can hold the role definition constant while you compare across employers of very different sizes, which is the same mechanic used to find companies hiring for a specific role by O*NET code.
A Live Example of the Pattern, and What It Does Not Prove
Here is the mechanic on a real record, pulled on September 11, 2026. Ramona Optics, Inc. (ramonaoptics.com) has a Salesforce Technology Detection with a first_seen_at of September 11, 2026, meaning the evidence is brand new as of today, and it has an open Sales role classified under O*NET code 41-4011.00, Sales Representatives, Wholesale and Manufacturing, Technical and Scientific Products. A newly detected sales technology co-occurring with an open sales hire is the co-occurrence shape you are looking for, and it is visible the same day it appears.
Now the discipline. This does not establish that Ramona Optics is replacing anything, leaving any vendor, or running a migration. It establishes two facts with timestamps and nothing more. To move a record like this from interesting to actionable, you would need the other half of the pattern: an incumbent tool in the same category that is no longer detected, ideally across more than one source, in the same window. Absent that, the honest label is “new detection, watching,” and the honest outreach angle is the new adoption itself rather than an imagined competitor’s loss. Treat anything less as a guess you are asking a rep to say out loud to a prospect.
How Much Platform Movement Shows Up in a Single Quarter
The volume of technology movement in a 90-day window is larger than most teams assume, which is why a scoring model matters more than a search query. Between June 13 and September 11, 2026, an estimated 238,919 US companies picked up a newly first-seen Salesforce Technology Detection, and an estimated 768,336 US companies picked up a newly first-seen HubSpot Technology Detection.
Read those figures for what they are. A first-seen detection means the evidence is new to that company’s record, which can mean a new adoption, a newly visible deployment, a newly published careers page, or a tool that finally surfaced through a second source. It does not mean a quarter of a million companies switched vendors. The useful takeaway is different and more practical: the adoption layer moves at six figures per quarter per major platform, so a switch-watch list built on new detections alone will be far too large to work, and the filtering has to come from the corroborating hiring signal. If you need the base mechanic first, start with how to find companies using HubSpot or Salesforce.
A Five-Step Workflow to Build a Switch-Watch List
1. Resolve the incumbent technology
Start from the catalog rather than a string match. GET /technologies with fuzzy_name returns the canonical technology record, including its categories and parent categories, so you know which other tools count as same-category alternatives. This is the step teams skip, and it is why displacement lists end up mixing a CRM with a marketing automation tool.
2. Pull the detection population and sort by recency
Use GET /discover/technologies/{technology_id_or_fuzzy_name}/technology_detections to pull companies with evidence of the technology, ordered by first_seen_at descending. For accounts you already track, GET /companies/{domain}/technology_detections returns the full detection history for one company, which is what you need for tenure. The endpoint reference sits in the PredictLeads API documentation, and the discovery mechanic is walked through in how to find every company using a technology by name.
3. Check hiring inside the same window
For each candidate, call GET /companies/{domain}/job_openings with first_seen_at_from set to the start of your window and with_description_only set to true, then search the descriptions for the phrase families above. To work the other direction and start from the roles, GET /discover/job_openings filters by onet_codes, location, and seniority. Job openings refresh approximately every 36 hours, so this half of the pattern is close to current.
4. Score the evidence into tiers
| Tier | Evidence required | What it justifies |
|---|---|---|
| Tier A | Incumbent not detected since a date inside the window, a new same-category detection, and an open role describing migration work. | Route to a rep now. Open on the work described in the posting, not on the vendor. |
| Tier B | Any two of the three, most often a migration posting plus an aging incumbent detection with no new same-category detection yet. | Add a webhook and a nurture sequence. Re-score when the third signal lands. |
| Tier C | One signal only, such as a quiet detection or a posting that merely lists the incumbent as a required skill. | No outreach. Keep it on the watch list so you can see if it develops. |
5. Follow the accounts and let the signals come to you
Polling a list every morning is wasted budget. POST /companies/{domain}/follow costs 0 credits, and webhooks then push TechnologyDetectionsDataset and JobOpeningsDataset events as they are found, so a Tier B account promotes itself to Tier A the day the missing signal appears. Delivery options across API, flat files, webhooks, and MCP are compared in the guide to company data delivery methods, and AI agents can run the same five steps conversationally through the PredictLeads MCP server.
Four False Positives That Wreck a Displacement List
Every one of these will put a company on your list that has no intention of switching anything, and three of the four are invisible unless you look at the source of the detection.
- Client work. Agencies, consultancies, system integrators, and resellers post migration roles constantly, because migrating other companies is the service they sell. Filter these out on the company record before scoring, not after.
- A required skill is not a deployment. A job description listing a tool tells you the employer wants that skill on the team. It may reflect current use, planned adoption, a migration in either direction, or simply a hiring manager copying a template.
- Infrastructure change dressed as churn. A content management system swap, a tag manager cleanup, or a move behind a content delivery network can quiet several website-based detections at once while nothing in the actual stack changed.
- Category coexistence. Large organizations frequently hold two same-category tools across business units for years. A new detection alongside an old one is often an addition, not a replacement, and the multi-org consolidation language in the postings above is exactly what addressing that state sounds like.
The same discipline applies to the hiring side of the pattern. A surge in postings can reflect a reorganization, a backfill wave, or a new office rather than a platform program, which is why it pays to read hiring in context using the approach in monitoring competitor hiring spikes with job data.
How PredictLeads Supports Switch Detection
PredictLeads is a company intelligence and technographic data provider, not a platform, so what it gives you is the evidence layer the scoring model above runs on. Three properties of the data do the heavy lifting for this specific use case.
- Multi-source detection. Technology Detections spans 1.5B+ detections since 2018 across 95M+ domains and 50,000+ tracked technologies, assembled from website script tags, DNS records, IP ranges, cookies, and job descriptions. Because job descriptions are one of the five sources, enterprise tools that never appear in page markup still get detected, flagged with
behind_firewall. - Source transparency. Every detection links back to the artifact it came from, whether that is a subpage URL, a job opening URL, or a DNS record. When a rep asks why an account is on the list, you can show them the sentence in the posting instead of a score.
- Point-in-time history. Every record carries
first_seen_atandlast_seen_at, which is what makes tenure, quiet periods, and co-occurrence windows computable rather than anecdotal. Job Openings holds 279.2M+ records since 2018 across 2.9M+ websites, with 10.2M active openings at any time.
Adjacent datasets make the tiers sharper. News Events covers 10M+ signals since 2016 across 37 categories, including integrates_with, partners_with, and ends_partnership_with, which occasionally confirms in public what the detections only suggest. Financing Events holds 210,800+ events since 2016, and fresh capital is a common precondition for a replacement program. Connections covers 371.8M+ relationships since 2019 sourced from customer, partner, integration, and case study pages, so you can see whether the incumbent still publicly names the account as a customer. Website Evolution tracks 776M+ subpages since 2021, which is where a quietly updated integrations page tends to show up first. The wider signal set is laid out in the guide to hiring and tech stack signals for lead generation.
Delivery is API, flat files, webhooks, and MCP, so the same evidence can drive a real-time alert, a nightly warehouse job, or an AI agent’s lookup. On compliance, PredictLeads collects only publicly available information, holds no personally identifiable information in its company intelligence datasets, is SOC 2 Type II certified, and is GDPR and CCPA compliant. For displacement work that touches a CRM, that distinction matters: these are company-level signals, not contact records. Timing the outreach once a Tier A account appears is covered in the guide to sales trigger events and outbound timing.
Final Thoughts on Finding Companies Switching Platforms
The teams that win displacement deals are not the ones with the longest list of accounts using a competitor. They are the ones who can tell the difference between an account that has used a tool quietly for four years and an account whose evidence changed last month while it was hiring someone to run a migration. That difference is entirely a matter of timestamps and corroboration, and both are available in public data.
Hold the line on language as you operationalize it. “Not detected since August,” “posted a role describing a consolidation,” and “two signals moved in the same quarter” are all defensible statements. “They are leaving vendor X” is not, and a rep who says it to a prospect who is not leaving vendor X has burned the account and the program. Build the tiers, write the hedged language into the alert template itself, and let the evidence carry the conversation. For the broader competitive picture around these signals, see the practical guide to modern competitor research using digital signals.
Ready to see this in your own data?
Get 100 free API requests when you create an account – no credit card, no sales call.
Frequently Asked Questions
How do you find companies likely to churn from a competitor’s product in 2026?
You look for a corroborated pattern rather than a single fact. The pattern is an incumbent technology that is no longer detected, a new detection in the same technology category, and a job posting in the same window that describes migration, consolidation, or cutover work. Any one of those alone is noise at scale, because an estimated 238,919 US companies picked up a newly first-seen Salesforce detection between June 13 and September 11, 2026 alone. PredictLeads Technology Detections and Job Openings both carry first_seen_at and last_seen_at, which is what lets you require that the signals land in the same window.
Can technology detection data prove that a company canceled a vendor contract?
No, and treating it that way is the fastest route to an embarrassing sales conversation. Technology Detections provide evidence of which technologies a company uses or has recently used, drawn from public sources. A detection that stops appearing means the evidence stopped, which can result from a script change, a recrawl gap, a vendor signature change, or an actual replacement. Use “not detected since [date]” in every field and template, and check source_count and detection_source_type before drawing a conclusion.
What job posting language suggests a company is migrating platforms?
Four phrase families carry most of the signal: movement language such as data migration, cutover, legacy system, and sunset; consolidation language such as multi-org consolidation, deduplication, and single source of truth; coexistence language such as hybrid architecture and dual-run; and named vendor-specific capabilities that indicate which product the team is being staffed to operate. A live Senior Salesforce Administrator posting found on September 10, 2026 combined data migration, deduplication, and multi-org consolidation in a single description, which is the shape to pattern match against. This is corroborating evidence of work being planned, not confirmation that any particular vendor is being replaced.
How long should a detection be quiet before you treat it as a possible switch?
There is no universal threshold, and any provider quoting one is guessing. Set the window against the detection cadence for that specific technology and company: high-traffic websites are crawled multiple times daily, while a tool evidenced only through job descriptions may go months between refreshes simply because the company is not hiring. The practical rule is to compare the quiet period against that company’s other detections, since a company-wide gap is a crawl artifact rather than a stack change. PredictLeads exposes last_seen_at on every detection so you can calibrate the threshold yourself instead of inheriting one.
What is the difference between competitive displacement signals and customer churn signals?
They point in opposite directions and serve different teams. Competitive displacement signals are about a prospect possibly leaving someone else’s product, and they are read from external public evidence: technology detections, hiring language, and news events. Customer churn signals are about your own accounts possibly leaving you, and the strongest inputs there are your product usage and support data, with public signals playing a supporting role. The data and the fields overlap, but the action does not: one routes to new business, the other to customer success.