Medusa
Cloud Pro → Self-hosted (Free)
Data Loss
Cloud deployment metrics
Feature Loss
Managed hostingManaged PostgreSQLCDN for mediaTeam dashboard
Gotchas
Need to self-manage PostgreSQL and RedisMedia storage needs S3 or equivalentPlugin marketplace items may need manual installation
Cooldown: Immediate after self-host setupRefund: Usage-based, no refund
Medusa is open-source headless commerce. pg_dump your data, deploy to any Node.js host. Cleanest e-commerce downgrade path.
MinIO
Enterprise (SUBNET) → Community (Free)
Data Loss
SUBNET support historyEnterprise audit logs
Feature Loss
SUBNET supportPanic buttonHealth diagnosticsLicense compliance tools
Gotchas
S3-compatible API is identical between editionsEnterprise features are support + tooling, not functionalitySame binary — just no support contractCluster still runs identically after license expires
Cooldown: ImmediateRefund: Annual subscription, no refund
MinIO Community is the same binary as Enterprise. You lose support, not features. Easiest enterprise-to-free downgrade possible.
Cloudflare R2
Paid (over 10GB) → Free Tier
Feature Loss
10GB storage limit1M Class A operations/month free10M Class B operations/month free
Gotchas
Egress remains free at all tiersIf you exceed free tier, billing kicks in automatically — no service interruptionMigrating away is trivial with rclone since egress is free
Cooldown: ImmediateRefund: Pay-as-you-go, no refund needed
Easiest object storage downgrade in the industry. Free egress means leaving R2 is also painless.
Feature Loss
Custom pricing ratesDedicated supportAdvanced fraud rules
Gotchas
Transaction fees increaseCustom rate locks expireVolume discounts removed
Cooldown: End of billing cycleRefund: No refund, switch at cycle end
Clean downgrade. All data preserved. You just pay more per transaction. Easiest payment provider to downgrade.
Payload CMS
Pro (Cloud) → Free (Self-hosted)
Data Loss
Cloud deployment logsTeam management settings
Feature Loss
Managed hostingAutomatic updatesTeam collaboration dashboardPayload Cloud storage CDN
Gotchas
Must set up your own MongoDB/PostgreSQL and S3 storageNo managed deployments — you handle CI/CDPlugins that depend on Payload Cloud API need reconfiguration
Cooldown: Immediate after self-host setupRefund: Usage-based, no refund
Payload is fully open source — self-hosting gives you 100% feature parity. The only effort is DevOps setup. One of the cleanest escape hatches in CMS.
Directus
Professional (Cloud) → Community (Self-hosted)
Data Loss
Cloud analytics dashboardManaged backup history
Feature Loss
Managed hostingAutomatic updatesDirectus Cloud marketplace extensionsPriority support
Gotchas
Self-hosted needs Docker + PostgreSQL setupCloud-only extensions won't work self-hostedMust handle backups yourself
Cooldown: Immediate after migrationRefund: Pro-rated on annual
Open-source core is identical to cloud. Export your SQL database, spin up Docker, restore. Textbook clean downgrade.
n8n Cloud
Starter → Community (Self-hosted Free)
Data Loss
Cloud execution logsWorkflow run history
Feature Loss
Managed hostingBuilt-in authCloud-only credentials storageAutomatic updates
Gotchas
Workflow JSON exports are fully portableSelf-hosted needs Docker + PostgreSQLSome cloud-managed credential types need re-authenticationSustainable Use License — can't offer n8n as a competing SaaS
Cooldown: Immediate after self-host setupRefund: Pro-rated on annual
Workflows export as JSON — import into self-hosted instance directly. Docker setup is trivial. One of the smoothest cloud-to-self-host transitions.
Tawk.to
Hired Agent → Free
Data Loss
Hired agent support history
Feature Loss
Hired human agents removedThat's basically it — core product is free
Gotchas
Tawk.to core chat is completely free — no feature gatingRemoving Tawk.to branding is a paid add-on ($19/mo)Hired Agents is the only paid feature besides white-labelAll chat history, triggers, shortcuts survive on free
Cooldown: ImmediateRefund: Pay-as-you-go, no refund needed
Tawk.to is one of the most generous free tiers in SaaS. Downgrade only means losing hired agents or white-label branding.
Data Loss
Audit log history reducedTeam-only features removed
Feature Loss
HIPAA BAA removedSSO with SAML removedSingle sign-on for dashboard removedAdvanced security featuresRead replicas may be removed
Gotchas
HIPAA workloads must migrate immediately or violate complianceSSO loss may require team to re-create individual accountsDatabase, auth, storage all unaffected
Cooldown: End of billing cycleRefund: Pro-rated on annual
Easy downgrade unless you rely on HIPAA or SAML SSO. Core data and features survive.
Data Loss
Events beyond 1M/mo not ingested
Feature Loss
Group analyticsA/B testing beyond 1M eventsCorrelation analysisPriority support
Gotchas
Events over 1M/mo silently droppedFeature flags still work but limited to 1M evaluationsSession recordings capped at 5k/mo
Cooldown: ImmediateRefund: Usage-based, no refund needed
Generous free tier. Most small-mid products fit under 1M events. Smoothest downgrade in analytics.
Data Loss
Email logs beyond 1 day
Feature Loss
100 emails/day → limitedCustom domains limited to 1Dedicated IPs removedWebhooks limited
Gotchas
100/day limit means transactional emails can fail silently if cap hitLosing dedicated IP may affect deliverabilityAPI key count limited
Cooldown: ImmediateRefund: Usage-based, no refund
Low-volume senders fit fine. 100/day is the hard cap — any production app sending verification emails will hit it.
Flagsmith
Scale → Open Source (Free)
Data Loss
Cloud analytics historyAudit trail on cloud
Feature Loss
Managed hostingRole-based access on cloudChange requests workflowScheduled flagsPriority support
Gotchas
Must self-host — requires Docker and PostgreSQL setupNo managed infrastructure — you handle uptimeFeature parity with open-source is close but some enterprise features missing
Cooldown: Immediate after self-host setupRefund: Usage-based, no refund
One of the best escape hatches — open-source version is fully featured. Self-hosting is the only effort required.
Cal.com
Team → Free (Self-hosted)
Data Loss
Cloud booking analyticsTeam management history
Feature Loss
Managed hostingRound-robin across teamManaged integrationsRouting forms (some)Priority support
Gotchas
Self-hosted needs Next.js + PostgreSQL + Prisma setupSome OAuth integrations need re-registering with your own app keysCodebase moves fast — updates require git merge effort
Cooldown: Immediate after self-host setupRefund: Pro-rated on annual
AGPL-licensed — full source available. Export PostgreSQL, deploy yourself. The fast-moving codebase is the only ongoing maintenance burden.
Windmill
Team (Cloud) → Community (Self-hosted Free)
Data Loss
Cloud execution historyTeam audit logs
Feature Loss
Managed hostingPriority supportCloud-specific integrationsMultiplayer editing
Gotchas
Scripts and flows export as JSON — fully portableSelf-hosted needs Docker ComposeSome enterprise features (SAML, audit) only in Enterprise EditionWorker management is your responsibility
Cooldown: Immediate after self-host setupRefund: Usage-based, no refund
AGPLv3 — full source available. Export scripts/flows, import into self-hosted. Clean migration path for an automation platform.
Ghost
Ghost(Pro) → Self-hosted (Free)
Data Loss
Ghost(Pro) analytics dashboardManaged email sending reputation
Feature Loss
Managed hostingCDNEmail sending infrastructureAuto-updatesManaged backups
Gotchas
Self-hosted Ghost is identical feature-wiseNeed your own SMTP for newsletters (Mailgun, SES)JSON export/import covers all contentTheme and settings fully portable
Cooldown: Set up self-hosted, import JSON — 1-2 hoursRefund: Pro-rated on annual
Ghost is fully open source. Export JSON, import on self-hosted instance. The only real migration cost is setting up email sending.
Doppler
Team → Developer (Free)
Feature Loss
Audit logs reduced to 7 daysTeam members capped at 5Service tokens limitedNo SAML/SSONo advanced access controls
Gotchas
Secrets remain accessible — no data lossIf team has more than 5 members, you'll be blocked from inviting new onesWebhooks count limitedIntegrations like Vercel, AWS, Kubernetes still work
Cooldown: ImmediateRefund: Pro-rated on annual
Smooth downgrade. Secrets stay intact. Only constraint is team size and audit log retention.
Linear
Business → Standard
Feature Loss
Triage automation removedAsks (issue intake) limitedInitiatives feature limitedCustom views per workspace reducedInsights with custom dashboards limited
Gotchas
All issues, comments, attachments preservedCycles and projects unaffectedIntegrations continue workingPer-seat pricing reduces from $14 to $8/user/mo
Cooldown: End of billing cycleRefund: Pro-rated annual
Clean downgrade. Linear preserves all core data. Only advanced workflow features are gated.
Data Loss
Roadmap historyCustom views beyond 5
Feature Loss
Unlimited members → 250 issuesRoadmapsCustom workflowsTriageSLAs
Gotchas
250 issue limit means you must archive old issuesIntegrations limitedNo API access on free
Cooldown: ImmediateRefund: Pro-rated
Small teams survive fine. 250 issue limit is the real constraint — archive aggressively before downgrade.
Data Loss
Audit logsUsage analytics detail
Feature Loss
Team members removed$20 credit → $5 creditHorizontal scalingMultiple environments per servicePrivate networking customizationPriority support
Gotchas
$5 monthly credit may not cover existing workloads — services sleep when exhaustedTeam members lose access immediatelyIf monthly spend exceeds $5, services pause until next billing cycleExecution time limit drops
Cooldown: ImmediateRefund: Usage-based, no refund
Simple downgrade if your workloads fit in $5/mo. Single-developer projects work great. Multi-service setups will likely exceed the hobby credit.
Data Loss
Usage metrics history
Feature Loss
Dedicated IPv4 addressesAutoscaling policiesMulti-org supportPriority supportVolume snapshots
Gotchas
Hobby has $5/mo included usage — anything beyond is billedExisting machines keep running but new ones may hit limitsOnly 3 shared-CPU VMs on hobby
Cooldown: ImmediateRefund: Usage-based, no refund
Docker containers run unchanged. The main constraint is the $5 included usage and 3 VM limit. Single-app projects work fine.
Turso
Scaler → Starter (Free)
Data Loss
Multi-region replica locations beyond 3
Feature Loss
Storage 24GB→9GBDatabases 10k→500Locations 16→3Embedded replicas limitedGroups limited to 1
Gotchas
If exceeding 9GB, must delete data before downgradeDatabase count above 500 will need consolidationFewer edge locations means higher latency for global users
Cooldown: ImmediateRefund: Usage-based, no refund
SQLite data is trivially portable. Generous free tier (9GB, 500 DBs). Downgrade is painless if you're within limits.
Mux
Enterprise → Pay-as-you-go
Data Loss
Custom SLA metricsDedicated encoding pipeline logs
Feature Loss
Volume discounts removedCustom CDN configurationPriority encoding queueDedicated supportCustom player branding
Gotchas
Per-minute pricing increases without volume discountsEncoding priority drops — uploads take longerLarge video libraries become expensive fast without enterprise rates
Cooldown: ImmediateRefund: Usage-based, no refund
Clean downgrade — all videos stay accessible. You just pay more per minute of storage and streaming. No data loss.
Data Loss
Slow log historyUsage analytics beyond 7 days
Feature Loss
256MB storage → 256MB (same)10k commands/day → 10kGlobal replication removedMulti-region limitedMax data size per key drops
Gotchas
10k commands/day is very low for production RedisGlobal replication stops — latency increases for non-primary regionQStash messages per day limited on free
Cooldown: ImmediateRefund: Usage-based, no refund
10k daily commands is the bottleneck. Dev/staging works fine. Production Redis workloads will hit the limit in minutes.
Trigger.dev
Pro → Hobby (Free)
Data Loss
Run history beyond 7 daysUsage analytics
Feature Loss
Concurrent runs 100→5Run duration 5min→30sTeam members removedLog retention 30d→7dSchedules limited
Gotchas
5 concurrent runs means queue backlog under load30s max run duration kills long-running tasksIf you have >5 team members they lose access immediately
Cooldown: ImmediateRefund: Usage-based, no refund
Open-source — self-host for unlimited runs. Hobby tier works for light background jobs. The 30s limit is the real constraint.
Data Loss
Run history beyond 3 daysUsage analytics detail
Feature Loss
Function runs 50k/mo→5k/moConcurrency 50→10Step limit 50→25Branch environmentsPriority support
Gotchas
5k monthly runs is low for productionStep limit of 25 may break complex multi-step functionsBranch environments removed — only production works
Cooldown: ImmediateRefund: Usage-based, no refund
Functions are your code — portable. 5k/mo runs is the constraint. Self-host Inngest Dev Server or migrate to BullMQ for unlimited.
Budibase
Premium → Free (Self-hosted)
Data Loss
Cloud hosting backupsUsage analytics
Feature Loss
Cloud hostingPriority supportCustom brandingEnforceable SSOApp backups managed
Gotchas
Self-hosted is full-featured — almost no feature lossDocker Compose setup is straightforwardMust manage your own backups and updatesCommunity support only on free
Cooldown: Must set up self-hosted instance firstRefund: Pro-rated
One of the best downgrade paths — self-hosted Budibase has near-full feature parity. The open-source version is genuinely usable.
Data Loss
Web Analytics advanced metricsCache Analytics detail
Feature Loss
WAF managed rulesetsImage optimization (Polish, Mirage)Mobile optimizationCloudflare FontsEnhanced HTTP/2 prioritizationPriority email support
Gotchas
Free WAF has limited rulesets — custom rules still workImage optimization removal may affect page speed scoresDNS and CDN still fully functional on freePage Rules limited from 20 to 3
Cooldown: ImmediateRefund: Pro-rated
Cloudflare free is genuinely excellent — DNS, CDN, basic WAF, DDoS protection all included. Pro mainly adds WAF rulesets and image optimization.
OpenAI API
Tier 5 → Tier 1
Feature Loss
Rate limits drop dramatically (Tier 5: 10k RPM → Tier 1: 500 RPM on GPT-4)Token-per-minute limits dropSome models gated behind higher tiers
Gotchas
Tier is auto-determined by usage and account age — you can't manually downgrade, only restrict spendingLower tier hits rate limits during traffic spikesFine-tuned models remain accessibleAPI keys unchanged
Cooldown: Tier resets based on 30-day rolling spendRefund: Pay-as-you-go — no refund concept
Tier downgrade isn't a manual process — it happens automatically with reduced usage. Rate limits are the main constraint.
Sentry
Team → Developer (Free)
Data Loss
Event history limited to 30 daysPerformance data beyond quota
Feature Loss
Multiple members → 1 user onlyDiscover queries limitedDashboards limitedRelease health5k errors/mo → limitedGlobal search
Gotchas
Single user only — team members lose access immediatelyEvents over quota silently droppedCustom dashboards may become read-onlyIntegrations limited on free tier
Cooldown: ImmediateRefund: Usage-based, pro-rated
Solo developers survive fine. The single-user limit makes it unusable for teams. Export your alert rules before downgrading.
Data Loss
Branch history beyond 7 daysCompute usage analytics
Feature Loss
Storage 50GB→512MBCompute 8 CU→0.25 CUBranches 500→10IP allow listsLogical replicationPriority support
Gotchas
Compute scales to zero with ~500ms cold starts512MB storage limit requires data pruning if aboveAutosuspend after 5 minutes of inactivityNo configurable autoscaling range
Cooldown: ImmediateRefund: Usage-based, no refund
Standard Postgres — pg_dump before downgrade and you're safe. Free tier is generous for dev/small projects. Cold starts are the main annoyance.
Postmark
Standard → Free (100 emails/mo)
Data Loss
Message retention drops to 10 daysDetailed analytics history
Feature Loss
10k+ emails/mo → 100/moDedicated IPPriority supportSuppressions management
Gotchas
100 emails/month is extremely low — basically dev/test onlyMessage stream separation still works but with tiny volumeDMARC reports stop
Cooldown: ImmediateRefund: Prepaid credits don't expire
Only viable for testing. 100/mo is not production-ready. But credits don't expire — unused prepaid balance carries over.
Mapbox
Commercial → Pay-as-you-go (Free tier)
Data Loss
Custom style analyticsUsage reports detail
Feature Loss
Volume discountsCustom map styles beyond 3Dedicated supportSLAMapbox Studio advanced features limited
Gotchas
50k free map loads/mo — anything above is billed at $5/1kAttribution must be displayed on free tierGeocoding requests capped at 100k/mo free
Cooldown: ImmediateRefund: Usage-based, no refund
Free tier is generous for small apps. 50k map loads covers most MVPs. Just ensure you display attribution or face enforcement.
Data Loss
Campaign analytics beyond 30 days
Feature Loss
1000 contacts → 100Custom domains removedAPI rate limits dropAudience segments limited
Gotchas
100 contact limit makes free tier effectively a sandboxContacts over 100 stop receiving emailsTransactional emails share the contact limit
Cooldown: ImmediateRefund: Usage-based, no refund
100 contacts on free is dev/testing only. Clean API — easy to switch to Resend or Postmark. Export contacts as CSV before downgrade.
Pulumi Cloud
Team → Individual (Free)
Data Loss
Team activity logsDeployment analytics
Feature Loss
Team members removedRBACAudit logsDrift detectionDeployment history limitedTTL stacks
Gotchas
Only 1 user on free tierStack state is still accessibleSelf-managed backends (S3, Azure Blob) work as alternativePulumi code is yours regardless of tier
Cooldown: ImmediateRefund: Pro-rated
Your IaC code is fully portable. Switch to self-managed backend (S3 bucket) for state and you lose nothing except the dashboard.
Docker Hub
Team → Personal (Free)
Data Loss
Team audit logsVulnerability scan history
Feature Loss
Unlimited private repos → 1Team managementVulnerability scanningAccess tokens per-repo scopingBuild minutesSSO
Gotchas
Only 1 private repo on free — all others become public or inaccessiblePull rate limits apply (100/6h anonymous, 200/6h authenticated)Image auto-builds no longer available even on paidTeam members lose access to private images
Cooldown: ImmediateRefund: Pro-rated
Push critical images to GitHub Container Registry or self-hosted registry before downgrading. The 1 private repo limit is the main constraint.
Appwrite
Pro → Free (Starter)
Data Loss
Bandwidth history beyond limits
Feature Loss
300K monthly active users → 75KBandwidth 300GB → 10GBStorage 150GB → 2GBFunctions executions 3.5M → 750KCustom SMTPAdditional organization members limited
Gotchas
2GB storage is tight if storing user uploads10GB bandwidth low for media-heavy appsSelf-hosting Appwrite is always free with no limitsDatabase and auth features identical on free tier
Cooldown: ImmediateRefund: Pro-rated
Appwrite free tier is generous for most indie projects. Self-host for zero limits. Storage and bandwidth are the main constraints.
imgix
Growth → Free (Basic)
Data Loss
Transformation analyticsSource image cache purge history
Feature Loss
1k master images → limitedCustom domains removedAdvanced transformations restrictedVideo API disabledPriority rendering queue
Gotchas
Existing image URLs stop working if source is disconnectedMust update all image references in your appNo custom subdomain means URL changes everywhere
Cooldown: ImmediateRefund: Usage-based, no refund
Image URLs are the trap — they're baked into your content. Plan URL migration before downgrading or you'll break every image on your site.
Data Loss
Audit log history beyond 90 daysSAML SSO configConcierge support history
Feature Loss
SAML SSO removedDDoS Mitigation level reducedWAF rules limitedDedicated infra removed99.99% SLA → 99.99% on Pro (similar)RBAC simplified
Gotchas
Custom contracts and discounts expireEnterprise-only features (Spend Management, Edge Config v2) become limitedPricing reverts to public pricing — typically 30-50% higher per unitExisting deployments unaffected
Cooldown: End of contract termRefund: Per contract — usually no refund
Possible but expensive. Public pricing on Pro for enterprise-scale workloads can exceed previous Enterprise spend.
1Password
Business → Teams Starter
Feature Loss
Custom roles removed (only Admin/Owner/Member)Activity log retention reducedVIP support removedAdvanced reportingCustom security policies
Gotchas
Vaults and items unaffectedProvisioning via SCIM still works on Teams Starter10 user limit on Teams Starter — must remove members above limitTravel Mode and Watchtower still included
Cooldown: End of billing cycleRefund: Pro-rated annual
Vaults stay safe. Main loss is custom roles and member cap. SMB-friendly downgrade.
Data Loss
Audit logs beyond 7 daysCustom domain configuration
Feature Loss
Custom domainAllowlist/blocklistMulti-sessionMAU limit drops to 10kPriority support
Gotchas
If over 10k MAU, excess users can't sign in until you upgrade or users churn below limitCustom OAuth providers removedBot protection reduced
Cooldown: ImmediateRefund: Pro-rated on annual
Clean if under 10k MAU. The hard cap on users is the blocker — no grace period for overages.
Data Loss
Web analytics detail (sampled on free)Firewall event logs beyond 24h
Feature Loss
WAF custom rulesetsImage optimization (Polish, Mirage)Cache AnalyticsMobile redirectCloudflare Fonts20 page rules → 3
Gotchas
WAF managed rules disabled — security posture drops significantlyImage optimization stops immediatelyCustom error pages revert to Cloudflare defaultsRate limiting rules removed
Cooldown: ImmediateRefund: Pro-rated on annual
Free tier is still excellent for caching and DDoS protection. Main loss is WAF rules — evaluate if your origin can handle unfiltered traffic.
SurrealDB Cloud
Pro → Free (Self-hosted)
Data Loss
Cloud monitoring dashboardsManaged backup snapshots
Feature Loss
Managed hostingAuto-scalingManaged TLSCloud dashboard
Gotchas
SurrealQL queries and schema are fully portableSelf-hosted needs careful storage engine choice (RocksDB vs TiKV)Cloud-managed auth tokens need re-creationMulti-node setup requires TiKV which is complex to operate
Cooldown: Immediate after self-host setupRefund: Usage-based, no refund
Export via surreal export, import into self-hosted. Single-node is easy. Multi-node with TiKV adds significant ops complexity.
Kong Gateway
Enterprise → OSS (Free)
Data Loss
Kong Manager UI settingsVitals analytics dataDev Portal content
Feature Loss
Kong Manager (admin GUI)Vitals monitoringDeveloper PortalRBACWorkspace isolationEnterprise plugins (OIDC, Vault, OPA)
Gotchas
Enterprise-only plugins stop working immediatelyKong Manager replaced by Admin API only (or use Konga)OIDC auth plugin is enterprise — must switch to key-auth or JWTNo workspace isolation on OSS
Cooldown: Must reconfigure, typically 1-2 daysRefund: Annual subscription, no refund
Kong OSS is production-ready. You lose the GUI and enterprise plugins. Most teams can replace enterprise plugins with OSS alternatives.
Loom
Business → Starter (Free)
Data Loss
Videos beyond 25 (oldest archived)Transcript history for removed videosEngagement analytics
Feature Loss
Unlimited videos → 25Video duration 45min → 5minDrawing toolsCustom brandingEngagement insightsCTA buttonsPassword protectionHD recording
Gotchas
Videos beyond 25 are archived, not deleted — upgrade to recover5-minute limit makes it useless for tutorials or demosExisting long videos become unplayable on freeDownloads of your own videos still work
Cooldown: ImmediateRefund: Pro-rated on annual
25 videos at 5 minutes each is extremely limited. Download all videos before downgrading. Consider alternatives like Tella or OBS for recording.
Pinecone
Standard (Pod) → Starter (Serverless)
Data Loss
Pod-specific configurationsIndex history beyond serverless retention
Feature Loss
Dedicated pods replaced by serverlessPerformance characteristics change (pod p99 latency was guaranteed; serverless is best-effort)Some pod-only index types unavailable
Gotchas
Must re-create indexes in serverless formatVectors must be re-uploaded if architecture differsServerless billing is per-query and per-storage — different cost model from pods$50 free credit on Starter for testing
Cooldown: Manual migration requiredRefund: Pro-rated, but pod commitments may have minimums
Not a true downgrade — it's a re-architecture. Serverless changes the performance contract entirely.
Data Loss
Generated audio over 90 daysCustom voice clones above free tier limit
Feature Loss
100k characters/mo → 10k characters/moCustom voice cloning limited (3 voices on free, vs 30 on Creator)Commercial use license restrictions tightenVoice library access limited
Gotchas
Existing cloned voices over the free limit may be inaccessible (not deleted)Free tier audio includes ElevenLabs attribution requirementProfessional voice cloning requires Creator+Audio quality (192kbps) reduced to 128kbps on free
Cooldown: ImmediateRefund: Pro-rated annual
Workable for prototyping. Commercial projects need at least Creator tier for licensing and quality.
Data Loss
Analytics history beyond 24hTeam members removed
Feature Loss
Preview deployments per PRPassword protectionServerless function 10s→5s timeoutBandwidth 1TB→100GB
Gotchas
Commercial use not allowed on HobbyCustom domains limitedConcurrent builds drop to 1Edge config removed
Cooldown: ImmediateRefund: No refund on monthly, pro-rated on annual
Hobby plan prohibits commercial use — that alone blocks most downgrades. Bandwidth limits may kill your site.
Render
Team → Individual (Free)
Data Loss
Team audit logsShared environment groups
Feature Loss
Team members removedDDoS protection downgradedAutoscalingPersistent disks limitedPreview environments limited
Gotchas
Free web services spin down after 15 min inactivityFree PostgreSQL expires after 90 daysBandwidth and build minute limits tighten significantlyNo custom Docker images on free tier
Cooldown: ImmediateRefund: Pro-rated on annual
Free tier is viable for hobby projects but the 90-day PostgreSQL expiry is a trap. Production workloads need at least Starter.
Mixpanel
Growth → Starter (Free)
Data Loss
Data history beyond 90 days (from unlimited)Saved reports by non-admin users
Feature Loss
Unlimited history → 90 daysGroup analyticsData modelingSignal reports20M events/mo → limited
Gotchas
90-day history means losing long-term trend analysisGroup analytics removal breaks B2B product metricsSaved reports from removed features become inaccessible
Cooldown: ImmediateRefund: Pro-rated on annual
The 90-day history cap is the real pain. If you need year-over-year comparisons, export data before downgrading.
Data Loss
Session analyticsFraud detection logs
Feature Loss
MAU 10k → 5kCustom domainsOrganization features limitedAdvanced MFA methodsPriority support
Gotchas
Users above 5k MAU can't authenticate until under limitB2B features (organizations) heavily restrictedRBAC policies may stop working on free tier
Cooldown: ImmediateRefund: Usage-based, pro-rated on annual
Similar to Clerk — the MAU hard cap is the blocker. If under 5k users, downgrade is smooth. Above that, users get locked out.
Data Loss
Branch history beyond 7 daysSearch analytics
Feature Loss
Storage 15GB→750MBRecords 750k→75kFile attachments 2GB→200MBAI questions limitedBranches limited to 3
Gotchas
75k record limit means bulk delete before downgradeFull-text search usage is metered on free tierFile storage cap blocks new uploads but existing files stay temporarily
Cooldown: ImmediateRefund: Usage-based, no refund
Built on PostgreSQL + Elasticsearch so data is exportable. But the 75k record cap is tight — export and prune before switching.
Data Loss
Metrics older than 13 monthsLog data beyond 30 days on free
Feature Loss
10k metrics series → 10k (same)50GB logs → 50GB (same)Synthetic monitoring checks 100k→10kAlerting rules limitedEnterprise plugins removed
Gotchas
Free tier limits look generous but data retention drops significantlyAlerting contact points limited to 5Grafana OnCall features restrictedSome data source plugins are enterprise-only
Cooldown: ImmediateRefund: Usage-based, no refund
Generous free tier. If you're within 10k series and 50GB logs, downgrade is painless. Self-host Grafana + Prometheus for unlimited.
Elastic Cloud
Enterprise → Free (Self-hosted)
Data Loss
Cloud snapshotsElastic Cloud management console data
Feature Loss
Managed hostingMachine learning anomaly detectionSearchable snapshotsCross-cluster replication managedEnterprise security features
Gotchas
Self-hosting Elasticsearch requires significant ops knowledgeSecurity features (RBAC, encryption) need manual configurationKibana dashboards exportable as NDJSON but some features are Platinum-onlyMemory-hungry — plan 4GB+ RAM minimum
Cooldown: Immediate after self-host setupRefund: Usage-based, no refund
AGPL license means self-hosting is legal and full-featured. Export indices via snapshot/restore. The ops burden is the real cost.
Terraform Cloud
Team & Governance → Free
Data Loss
Policy check historyCost estimation historyRun triggers history
Feature Loss
Team management (1 team only)Sentinel policiesCost estimationRun triggers between workspacesSSODrift detectionConcurrent runs limited to 1
Gotchas
Sentinel policies stop enforcing — governance gaps appearSingle concurrent run means queue backlogTeam-based access controls disappearState file access still works but management is limited
Cooldown: ImmediateRefund: Pro-rated on annual
State files are portable. The real loss is governance (Sentinel) and team management. Self-host with OpenTofu + Atlantis for a free alternative.
DigitalOcean
Pro → Basic Droplet
Data Loss
Monitoring alerts historyManaged database backups (if removing managed DB)
Feature Loss
Managed databases auto-failoverPremium AMD/Intel CPUsUptime monitoring advancedPriority supportSLA guarantees reduced
Gotchas
Managed databases require separate migration to self-hostedSpaces CDN bandwidth included dropsPremium Droplets have better network performanceTeam member management simplified on lower tiers
Cooldown: Immediate, but migrate databases firstRefund: Pay-as-you-go, no refund
Basic Droplets are solid. The main loss is managed database features and premium CPU/network. Manual database management is the real cost.
Feature Loss
MAU cap drops to 7,500Custom domains removed (auth0.com URLs only)Enterprise connections removed (LDAP, SAML, AD)Role-based access control limitedActions count reduced
Gotchas
If MAU exceeds 7,500, Auth0 throttles auth — users may be unable to loginCustom domain change breaks existing app config and OAuth callbacksFree tier has Auth0 branding on login pagesUser data preserved
Cooldown: ImmediateRefund: Pro-rated annual
Risky downgrade. MAU spikes can break user login. Custom domain removal requires app reconfiguration.
Feature Loss
MAU cap drops to 10,000Removable Clerk branding becomes mandatoryCustom domains removedAllowlist/blocklist limitedSAML SSO removedMulti-session handling limited
Gotchas
If you exceed 10k MAU, new signups blockedCustom email/SMS templates revert to Clerk defaultsBranding on auth pages and emailsUser data fully preserved
Cooldown: ImmediateRefund: Pro-rated annual
Workable for small projects. SAML and custom branding loss is the main cost. User data preserved.
Data Loss
Message history beyond 90 days hiddenFile storage beyond 5GB oldest deleted
Feature Loss
Guest accountsGroup calls (15→2 people)Custom retention policiesWorkflow builderScreen sharing annotations
Gotchas
Messages still exist but aren't searchable/visibleApp integrations limited to 10Cannot use SSO/SAMLHuddles limited to 2 people
Cooldown: ImmediateRefund: Pro-rated on annual plans
Message history isn't deleted, just hidden — upgrade again and it's back. But 10 app integration limit cripples most teams.
Netlify
Pro → Starter (Free)
Data Loss
Analytics dataForm submissions beyond 100/mo
Feature Loss
Bandwidth 1TB→100GBBuild minutes 25k→300Background functionsPassword protectionRole-based access
Gotchas
Bandwidth overage bills at $55/100GB even on freeForms submissions capped hardSplit testing removedBuild image selection limited
Cooldown: ImmediateRefund: Pro-rated on annual
Watch bandwidth — overages on free plan are expensive. Low-traffic sites survive fine.
Data Loss
Version history beyond 7 daysBulk export limited
Feature Loss
Unlimited blocks → 1000 block limit for teamsUnlimited file uploads → 5MB per fileGuests limited to 10Automations limitedBulk PDF export
Gotchas
1000 block limit doesn't delete content but makes it read-only beyond limit5MB file upload cap blocks new attachments but existing ones remainIf you have >10 guests, excess lose accessCustom automations stop running
Cooldown: ImmediateRefund: Pro-rated on annual
Personal use is fine — free tier is generous for individuals. Teams hit the block limit fast. Export everything important before downgrading.
Data Loss
Deployment history beyond 7 daysScheduled function logs
Feature Loss
Database storage 1GB→512MBBandwidth 5GB→1GBScheduled functions limitedTeam members removedCustom domains
Gotchas
If database exceeds 512MB, writes are blocked until you reduce dataFunction execution limits tighten significantlyNo streaming export on free tierComponents limited
Cooldown: ImmediateRefund: Usage-based, no refund
The 512MB storage cap is the blocker. If your data fits, free tier works. But blocked writes when over limit can break your app silently.
Amplitude
Growth → Starter (Free)
Data Loss
Behavioral cohort definitionsCustom event properties beyond limits
Feature Loss
Unlimited events → 50M/moPredictionsCausal analysisCustom dashboards limited to 5Advanced behavioral cohorts
Gotchas
50M events sounds generous but large apps hit it fastCohort-based targeting in experiments stops workingData retention drops from unlimited to 12 months
Cooldown: ImmediateRefund: Annual contract, pro-rated
Free tier is generous for small products. But losing behavioral cohorts breaks any experiment or targeting workflow.
Data Loss
Transformation logs beyond 7 daysCustom backup configurations
Feature Loss
25 credits → 5 credits/moAdvanced transformationsAuto-tagging AIBackup to S3Custom domainsMulti-CDN
Gotchas
Credits system is opaque — 5 credits covers ~5GB storage OR 5GB bandwidth, not bothExisting transformations keep working but new ones may fail if over creditNamed transformations capped at 3 on free
Cooldown: ImmediateRefund: Pro-rated on annual
Credit-based pricing makes it hard to predict limits. Images stay but heavy transformation usage will exhaust free credits fast.
Data Loss
Data retention drops from 13 months to 8 daysSynthetic monitoring history
Feature Loss
Full-platform users → 1 free userData retention 13mo → 8 daysAlerts unlimited → limitedSynthetics checks 10k → 500/moSAML SSO
Gotchas
Only 1 full-platform user on free — everyone else gets basic access100GB/mo free ingest sounds generous but APM data burns through it fast8-day retention means no historical comparisonInfinite Tracing removed
Cooldown: ImmediateRefund: Usage-based, pro-rated on annual commitment
Generous 100GB free ingest, but 1 full-platform user and 8-day retention make it a solo dev tool. Teams need Pro.
Data Loss
Incident history analyticsPostmortem reports
Feature Loss
Unlimited users → 5 usersEvent intelligenceStakeholder notificationsResponse playsPostmortemsService graphChange events
Gotchas
5-user limit immediately locks out most of the on-call teamEscalation policies simplified — complex routing breaksAnalytics and reporting completely removedOnly 1 on-call schedule on free
Cooldown: ImmediateRefund: Pro-rated on annual
5 users is barely enough for a small team's on-call rotation. Migrate to Grafana OnCall (free, self-hosted) for equivalent alerting.
1Password
Business → Individual
Data Loss
Shared vaults dissolvedActivity logsCustom groups
Feature Loss
Shared vaultsTeam managementSSO integrationCustom rolesAdvanced reportingAdmin console5GB per user → 1GB
Gotchas
Shared vault items must be moved to individual vaults before downgradeTeam members lose access to shared credentials immediatelySSO provider integration breaks — users need master passwordsWatchtower business features removed
Cooldown: ImmediateRefund: Pro-rated on annual
Individual passwords survive. Shared team credentials are the risk — export shared vault to individual vaults before downgrading.
Data Loss
Deployment tracking historyCode insights data
Feature Loss
Unlimited users → 5 usersDeployment permissionsIP allowlistingRequired merge checksBuild minutes 3500 → 50/moSmart mirroringPremium support
Gotchas
5-user limit kills it for any team beyond a startup50 build minutes/month is nearly zero CI/CDPipelines effectively become unusable on freePrivate repos still allowed (unlike GitHub old policy)
Cooldown: ImmediateRefund: Pro-rated on annual
Repos and code survive but the 5-user limit and 50 build minutes make free tier suitable only for personal projects.
Figma
Professional → Starter (Free)
Data Loss
Projects beyond 3 become view-onlyVersion history beyond 30 days
Feature Loss
Unlimited projectsTeam librariesCustom file/user permissionsBranching and merging
Gotchas
Files remain but excess projects become read-onlyTeam library links break for consumersCan't publish components
Cooldown: ImmediateRefund: Pro-rated on annual
Files don't disappear but you lose team libraries — design systems break. Move critical files to the 3 allowed projects first.
Data Loss
Advanced report data granularityCustom report definitionsThird-party calculated shipping rates
Feature Loss
Advanced report builderThird-party calculated shipping15 staff accounts → 2Inventory locations 8 → 4Shipping discounts 84% → 77%International pricing per market
Gotchas
Staff accounts beyond 2 immediately lose accessIf using >4 inventory locations, must consolidate firstInternational pricing reverts to single currencyTransaction fees increase if not using Shopify Payments
Cooldown: Immediate, billed at new rate next cycleRefund: No refund, switch takes effect immediately
Most painful part is staff account limits and losing calculated shipping. If you rely on complex shipping rules, you'll feel this downgrade immediately.
LaunchDarkly
Pro → Starter
Data Loss
Audit log history beyond 14 daysFlag analytics history
Feature Loss
Unlimited flags → 1k flagsCustom rolesScheduled flag changesApproval workflowsFlag triggersExperimentation
Gotchas
If over 1k flags, excess flags are disabled — features may breakApproval workflows stopping means anyone can change flagsSSO removed forces password-based login
Cooldown: ImmediateRefund: Annual contract, no refund
1k flag limit sounds high but grows fast. The real risk is losing approval workflows — accidental flag changes in production.
Data Loss
Email activity beyond 3 daysAdvanced stats history
Feature Loss
100 emails/day limitDedicated IP removedEmail validation APISubuser managementTeammate roles limitedAdvanced statistics
Gotchas
100/day cap means transactional emails fail when exceededLosing dedicated IP affects deliverability reputationSender reputation doesn't transfer to shared IP poolMarketing campaigns volume severely limited
Cooldown: ImmediateRefund: Pro-rated on annual
100 emails/day is fine for development only. Production apps sending verification or notification emails will hit the cap immediately.
Intercom
Advanced → Essential
Data Loss
Conversation history preserved but AI features lostCustom bot flow analytics
Feature Loss
Fin AI Agent limitedCustom bots restrictedAdvanced ticketingWorkload managementTeam inboxes limitedSLA rulesCSAT surveys limited
Gotchas
Fin AI usage drops significantly — customer support quality affectedCustom bot flows may stop workingTeam inbox consolidation requiredSLA tracking disappears
Cooldown: End of billing cycleRefund: Annual contract, no refund
Conversation history is preserved. The main hit is AI and automation features. Essential still works for basic live chat.
CircleCI
Performance → Free
Data Loss
Build artifact historyInsights analytics dataTest splitting history
Feature Loss
Unlimited credits → 6000 credits/moParallelismSelf-hosted runners unlimited → 1Docker layer cachingmacOS/Windows executors limitedPriority support
Gotchas
6000 credits sounds generous but medium Docker builds burn 100+ credits eachNo parallelism on free means slow test suitesmacOS builds extremely credit-expensiveDocker layer caching removal doubles build times
Cooldown: ImmediateRefund: Usage-based, no refund
Free tier works for small open-source projects. Production CI/CD with test parallelism and Docker layer caching needs paid tier.
Zendesk
Suite Professional → Suite Team
Data Loss
Custom analytics reportsSLA achievement history (reports, not data)
Feature Loss
Custom reporting and dashboardsSkills-based routingCommunity forumsMultiple ticket formsConditional fields and formsMultibrand supportCSAT surveys customization
Gotchas
Custom reports disappear — must recreate with limited reportingSkills-based routing fallback to round-robinCommunity forums entirely removed on TeamMacros and triggers survive but conditional fields break
Cooldown: End of billing cycleRefund: Annual contract, no mid-term refund
Ticket data survives but routing intelligence and custom reporting are gutted. Team tier works for basic support but not for scaled operations.
Crisp
Unlimited → Basic (Free)
Data Loss
Chatbot scenario analyticsKnowledge base analyticsCampaign history
Feature Loss
Chatbot scenariosKnowledge baseCampaigns and triggersAudio/video callsMagicMap visitor trackingCanned responsesStatus page
Gotchas
Free tier is basic live chat only — no automationChatbot scenarios stop working immediatelyKnowledge base goes offlineOnly 2 seats on free tier
Cooldown: ImmediateRefund: No refund on monthly
Free Crisp is functional live chat for 2 operators. Automation, knowledge base, and campaigns all require Pro or Unlimited.
Fastly
Professional → Free (Developer)
Data Loss
Historical traffic analyticsReal-time log streaming configs
Feature Loss
Bandwidth limited to free tier capsImage optimizerWeb application firewallDDoS protection advancedReal-time log streamingCustom VCL (limited)99.99% SLA
Gotchas
Free developer tier has strict bandwidth limitsVCL (Varnish Configuration Language) customization limitedNo WAF on free means reduced security postureOrigin shield removed — more requests hit your origin
Cooldown: End of billing cycleRefund: Pay-as-you-go, no refund
Fastly free works for development and testing. Production CDN needs paid tier for WAF, SLA guarantees, and reasonable bandwidth.
HashiCorp Vault
Enterprise (Cloud) → OSS (Self-hosted Free)
Data Loss
Cloud audit logsHCP Vault analyticsManaged snapshots
Feature Loss
Managed hostingMulti-tenancy (namespaces)Sentinel policiesControl groupsMFA enforcement at auth levelReplication (DR + Performance)
Gotchas
Self-hosting Vault is operationally complex — unsealing, HA setupEnterprise namespaces don't exist in OSS — must restructure secretsDR replication is enterprise-only — must build your own backup strategyAuto-unseal with cloud KMS requires enterprise in some configurations
Cooldown: Must set up self-hosted, migrate secrets — 1-2 weeksRefund: Annual contract, no refund
Vault OSS is production-grade but running it is an ops commitment. Secrets are exportable. The real cost is operational: HA, unsealing, backups.
Data Loss
Analytics historyForm submission history beyond limits
Feature Loss
Custom domain removed1000 CMS items → 10Password protectionCustom codeFavicon customizationFramer badge appearsBandwidth 200GB → 1GB
Gotchas
Losing custom domain breaks all existing linksCMS items above 10 unpublishedFree tier is essentially unusable for production sites1GB bandwidth is exhausted quickly on any real traffic
Cooldown: ImmediateRefund: No refund on monthly, pro-rated on annual
Free Framer is a design playground, not a production tool. Export to code or migrate to a static site generator before downgrading.
Calendly
Professional → Free
Data Loss
Workflow automation historyRouting form submissionsAnalytics data
Feature Loss
Unlimited event types → 1Workflows/remindersRound-robin schedulingCollective schedulingStripe/PayPal integrationCustom brandingRedirect after bookingRouting forms
Gotchas
1 event type on free is extremely limitingAll workflow automations (email reminders, follow-ups) stopExisting booking pages with multiple event types breakCal.com (open source) is a near-equivalent alternative
Cooldown: ImmediateRefund: Pro-rated on annual
1 event type makes free Calendly barely usable. Migrate to Cal.com (self-hosted, free, unlimited) for equivalent functionality.
Data Loss
Incident history beyond 6 monthsCustom dashboards
Feature Loss
5 user limit on FreeEmail-only notifications (no SMS or phone)Basic integrations only (limited webhooks)No service dependenciesNo event orchestrationNo global event routing
Gotchas
Existing escalation policies may not work with reduced featuresSMS and phone notifications crucial for on-call — losing them risks incident responseIf team > 5, must remove usersSlack/Teams integrations limited
Cooldown: End of billing cycleRefund: Per contract
Free tier of PagerDuty is too limited for real on-call. Consider Grafana OnCall (free, self-hosted) or incident.io as alternatives.
Data Loss
SAML SSO configAudit log historyAdvanced security findings
Feature Loss
Protected branches rules (limited)Required reviewersCode owners enforcementActions minutes drop to 2000/mo
Gotchas
Private repos stay but lose advanced featuresEnterprise-only API endpoints stop workingGitHub Connect disabled
Cooldown: ImmediateRefund: Pro-rated refund on annual plans
Private repos survive but governance features vanish. CI pipelines may break from reduced Actions minutes.
PlanetScale
Scaler Pro → Hobby
Data Loss
Read replicas removedQuery insights historyBranch deploy requests
Feature Loss
Horizontal shardingRead replicasSOC 2 audit logsSSO5GB storage → 1GBUnlimited branches → 1
Gotchas
Hobby tier was removed in 2024 and brought back — verify current availability1 branch means no branching workflow at all1GB storage requires aggressive data pruningConnection limits drop significantly
Cooldown: Must provision new database, migrateRefund: Usage-based, no refund
PlanetScale's free/hobby tier has been unstable (removed and restored). The 1GB limit and single branch make it barely usable for production.
Data Loss
Audit log historyDirectory sync snapshots
Feature Loss
SSO connections beyond 1Directory syncAdmin portal customizationCustom domainsDedicated support
Gotchas
Enterprise SSO connections disabled — their employees can't log inDirectory sync stopping means user provisioning breaksIf customers depend on SAML login, downgrade breaks their access immediately
Cooldown: ImmediateRefund: Annual contract, pro-rated
If you sold enterprise SSO to customers, downgrading breaks their login flow. Must migrate each customer's connection before downgrade.
Temporal Cloud
Growth → Self-hosted (Free)
Data Loss
Cloud metrics and namespace analyticsManaged retention policies
Feature Loss
Managed hostingMulti-region replicationAuto-scaling workersCloud UIDedicated support
Gotchas
Self-hosting Temporal is complex — needs Cassandra or PostgreSQL + ElasticsearchRunning workflows must drain before migrationWorker code is portable but namespace configs need recreationVisibility store setup is non-trivial
Cooldown: Must drain running workflows, 1-2 weeks migrationRefund: Usage-based, no refund
Worker code is 100% portable. But self-hosting Temporal's server is a significant ops burden — it's a distributed system with multiple dependencies.
Data Loss
Advanced roadmaps dataAutomation audit logsSandbox environment
Feature Loss
Unlimited users → 10 usersAdvanced roadmapsIP allowlistingProject archivingSandbox testingAdmin insights24/7 Premium support
Gotchas
10-user limit means most team members lose access immediatelyAdvanced roadmaps disappear — cross-project planning breaksAutomation rules limited to single-projectStorage drops from unlimited to 2GB
Cooldown: ImmediateRefund: Pro-rated on annual
The 10-user limit kills it for any real team. Export issues as CSV/XML before downgrade. Consider migrating to Linear or Plane instead.
ClickHouse Cloud
Production → Free (Self-hosted)
Data Loss
Cloud backups and snapshotsPerformance monitoring dashboards
Feature Loss
Managed hostingAuto-scalingManaged backupsCloud consoleAWS PrivateLinkDedicated support
Gotchas
Self-hosting ClickHouse requires significant ops knowledgeReplicatedMergeTree setup for HA is complexZooKeeper/ClickHouse Keeper needed for replicationData migration via clickhouse-client INSERT from remote or backup/restore
Cooldown: Must provision self-hosted cluster, migrate data — 1-2 weeksRefund: Usage-based, no refund
ClickHouse is open source — full feature parity self-hosted. But operating a ClickHouse cluster at scale requires dedicated DBA effort. Data export via native format is fast.
KeyCDN
Standard → N/A (No free tier)
Data Loss
Analytics historyZone configurations
Feature Loss
No free tier — all CDN delivery stopsZone pull configurationsOrigin shieldLet's Encrypt SSLReal-time analytics
Gotchas
Pay-as-you-go with minimum $4/month — no true free tierZone aliases and custom hostnames stop resolvingCached content purged after account closureMigrate DNS CNAME records to new CDN first
Cooldown: Immediate on non-paymentRefund: Pay-as-you-go, remaining credit refundable
No free tier. Straightforward migration to any CDN — just update CNAME records and cache settings.
Twilio Flex
Active → Pause / Cancel
Data Loss
Conversation history beyond export windowCustom plugin state
Feature Loss
All Flex UI access removedStudio flows for Flex deactivatedTaskRouter Flex workspace removedActive Voice/Messaging Channels detached
Gotchas
Phone numbers must be ported out (30-60 days) or releasedCustom plugins (React) need separate hostingConversation history must be exported within retention windowTaskRouter SDKs in your code must be migrated to standalone TaskRouter
Cooldown: Per contract termsRefund: Per contract — annual commits typically non-refundable
Twilio Flex is deeply integrated — downgrade requires careful migration plan. Phone number porting is the longest pole.
Data Loss
Database paused after 7 days inactivityStorage files beyond 1GB deleted after 30 days
Feature Loss
Database size 8GB→500MBFile storage 100GB→1GBNo daily backupsEdge Functions limitedNo email templates
Gotchas
Database gets PAUSED if inactiveData exceeding free limits may be deletedBranching disabledPoint-in-time recovery lost
Cooldown: Immediate downgrade, 30-day grace for dataRefund: Pro-rated refund on annual
The pause-on-inactivity kills it. If your DB exceeds 500MB you must trim before downgrade or risk data loss.
AWS
Enterprise Support → Basic (Free)
Data Loss
Trusted Advisor full checks historyTechnical Account Manager notesSupport case history beyond 12 months
Feature Loss
24/7 phone/chat supportTAM (Technical Account Manager)Concierge supportInfrastructure event managementFull Trusted AdvisorThird-party software support
Gotchas
Support response time goes from 15min to no guaranteed SLAReserved Instance purchase recommendations lostOperational reviews discontinuedArchitecture guidance unavailable
Cooldown: End of billing cycle, 30-day notice requiredRefund: No refund, pro-rated for annual commitment
Losing Enterprise Support feels like flying blind. No TAM, no architecture reviews, no rapid incident response. Workloads keep running but you're on your own.
Algolia
Premium → Build (Free)
Data Loss
Analytics historyA/B test resultsQuery suggestions data
Feature Loss
10k records → 10k (same but search units capped)Advanced AI featuresCustom ranking strategiesSLA guaranteesDedicated infrastructurePersonalization
Gotchas
Search units hard-capped — site search stops responding when exhaustedNo query suggestions on free tierAPI key scoping limitedSupport drops to community-only
Cooldown: Immediate, must re-index on new planRefund: Annual contract, early termination fees
Record limit isn't the issue — search unit quotas are. A popular site can exhaust free-tier units in hours. Must have fallback search ready.
CockroachDB
Dedicated → Serverless (Free)
Data Loss
Cluster performance metrics historyBackup schedules removed
Feature Loss
Dedicated hardwareMulti-region by defaultPrivate endpointsVPC peering10GB storage → 5GB50M RU/mo limit
Gotchas
Request Unit model is hard to predict — queries that were free on Dedicated now cost RU5GB storage cap forces data pruningConnection pooling limits dropServerless has cold starts on inactive clusters
Cooldown: Must provision new serverless cluster, migrate dataRefund: Usage-based, no refund
CockroachDB Serverless uses Request Units — a completely different billing model. Your existing queries may burn through free RU in hours.
Data Loss
Analytics on page viewsAutomation historyAdmin insightsTeam calendars
Feature Loss
Unlimited users → 10 usersSpace permissions granularityArchivingTeam calendarsIP allowlistingExternal collaboratorsUnlimited storage → 2GB
Gotchas
10-user cap immediately locks most team members out2GB storage means large attachments must be prunedMacros from paid apps stop renderingPage history preserved but analytics lost
Cooldown: ImmediateRefund: Pro-rated on annual
Same 10-user trap as Jira. Export spaces as XML before downgrade. Most macro-heavy pages will break visually.
Data Loss
Security dashboard findingsCompliance pipeline historyValue stream analytics
Feature Loss
Unlimited CI/CD minutes → 400/moSecurity scanning (SAST, DAST, dependency)Compliance frameworksPortfolio management (epics hierarchy)Audit events streamingVulnerability management
Gotchas
400 CI/CD minutes exhausted fast on active projectsSecurity scanning results disappear from MR widgetsEpics limited to single level on free5GB storage per project on free tierProtected environments removed
Cooldown: ImmediateRefund: Pro-rated on annual
Code and repos survive but security and compliance features vanish. CI/CD minute limit is the biggest operational hit for active teams.
Data Loss
CSAT survey historyTime tracking dataCustom reports
Feature Loss
Unlimited agents → 2 agentsAutomationsSLA managementCustom ticket viewsTime trackingCSAT surveysMultiple products/brandsRound-robin routing
Gotchas
2-agent limit makes it a personal ticketing system onlyAll automation rules stop runningSLA policies removed — response time tracking goneCanned responses limited on free tier
Cooldown: ImmediateRefund: Pro-rated on annual
Free Freshdesk is an email ticketing system for 2 people. Any team operation needs Growth tier minimum.
Data Loss
Metrics history beyond 1 dayAll custom dashboardsAll alerting rules
Feature Loss
5-host limit on free1-day metric retention (vs 15 months)No APM, no logs, no syntheticsNo integrations beyond default agent
Gotchas
Free tier essentially useless for production monitoringExisting dashboards and monitors deleted after grace periodRe-onboarding new monitoring tool takes weeksMulti-year contracts have early termination penalties
Cooldown: End of contract (usually annual or multi-year)Refund: Per contract — typically no refund
Datadog downgrade essentially means migrating to another tool. Free tier is too restrictive for real use.
MongoDB Atlas
Dedicated → Shared (Free)
Data Loss
Continuous backupsPerformance advisor dataCluster configs
Feature Loss
Dedicated clustersVPC peeringEncryption at rest (custom keys)Advanced monitoring512MB storage limit
Gotchas
Must migrate data to smaller cluster512MB limit requires data pruningConnection strings changeNo SLA on shared tier
Cooldown: Must provision new cluster, migrate dataRefund: Pay-as-you-go, no refund
Not really a downgrade — you create a new free cluster and migrate. 512MB limit means serious data pruning required.
Data Loss
Log retention drops to 2 daysAnomaly detection historyBranding customization
Feature Loss
Unlimited social connections → 2MFA with SMSCustom domainsLog streamsOrganizations featureUp to 10k MAU → 7.5kCustom DB connections beyond 1
Gotchas
Social connections beyond 2 are DISABLED — users who signed up via those providers lose accessCustom domain removed breaks existing URLs/tokensActions may stop working if they use Pro-only featuresExisting sessions invalidated on plan change
Cooldown: ImmediateRefund: Pro-rated on annual, none on monthly
Social connection limit is the killer. If users signed up via Google AND GitHub AND another provider, one of those breaks. Plan carefully.
Zapier
Professional → Free
Data Loss
Task history beyond 7 daysAll multi-step Zap configurations disabled
Feature Loss
Unlimited Zaps → 5 active ZapsMulti-step Zaps → single-step only100 tasks/moPremium apps removedFilters, Paths, FormattersWebhooksCustom logic (Code steps)
Gotchas
5 Zaps with single steps is nearly useless for real automation100 tasks/month exhausted by a single active workflowPremium app connections disabled — many popular apps are premiumNo webhooks on free means no real-time triggers
Cooldown: ImmediateRefund: Pro-rated on annual
Free Zapier is a demo, not a product. Migrate to n8n (self-hosted) or Make for equivalent automation without the restrictions.
HubSpot
Professional → Free
Data Loss
Custom reportsWorkflow historyCall recordings
Feature Loss
Marketing automation entirelyCustom reportingA/B testingPredictive lead scoringCustom workflowsSequences1M contacts → 1M (same) but features gated
Gotchas
Workflows stop immediately — all automated emails and tasks haltCustom properties remain but custom reports disappearHubSpot branding added to forms and emailsSequences for sales outreach disabled
Cooldown: ImmediateRefund: Annual contract, no mid-term refund
CRM data stays intact but automation is gutted. The free CRM is useful for contact management but useless for marketing automation.
Data Loss
Audit logsVersion history beyond latestUsage analytics
Feature Loss
Unlimited apps → 5 appsUnlimited end users → 5 end usersSource control (Git)Custom brandingSSO/SAMLGranular permissionsEnvironments (staging/prod)
Gotchas
5-app limit makes free tier a prototype tool only5 end users means no internal tool deploymentGit sync removed — no version control for appsCustom components still work but can't be shared across apps
Cooldown: ImmediateRefund: Annual contract, no refund
Free Retool is a demo environment. Any real internal tool deployment needs paid tier. Consider Budibase or Tooljet as open-source alternatives.
Webflow
Business Site → Free (Starter)
Data Loss
CMS items beyond 50Form submissions beyond 50Bandwidth history
Feature Loss
CMS items 10k → 50Form submissions 2500 → 50Custom code removedPassword protectionWebflow.io subdomain onlyCustom domain removedSite search removed
Gotchas
Losing custom domain breaks all SEO and inbound linksCMS items above 50 are unpublished — content effectively hiddenCustom code injections removed may break site functionalityWebflow branding badge appears
Cooldown: ImmediateRefund: No refund on monthly, pro-rated on annual
Free Webflow is a portfolio/prototype tool. Any business site needs paid. Export HTML/CSS before downgrading if migrating away.
Firebase
Blaze (Pay-as-you-go) → Spark (Free)
Data Loss
Cloud Functions code stays but stops executingScheduled functions haltExtensions disabled
Feature Loss
Cloud Functions entirelyFirestore 1GB→free limitsStorage 5GB→1GBHosting custom domain SSLRemote Config conditions limitedBigQuery export
Gotchas
Cloud Functions stop IMMEDIATELY — any backend logic breaksFirestore reads/writes have daily caps that reset at midnight PTCannot downgrade if any resource exceeds free limits — must delete firstPhone auth verification costs stop being covered
Cooldown: Immediate, must reduce usage below free limits firstRefund: Pay-as-you-go, no refund
If you use Cloud Functions at all, downgrading breaks your backend instantly. Must rewrite server logic before switching. Painful.
Bunny.net
Standard → N/A (No free tier)
Data Loss
All cached contentPull zone configurationsDNS recordsStorage zone data
Feature Loss
No free tier — all services stopCDN deliveryBunny DNSEdge storageStream video hostingImage optimization
Gotchas
Bunny.net has no free tier — downgrade means cancellationAll pull zones stop serving content immediatelyDNS records must be migrated before cancellationStorage zone files must be downloaded first
Cooldown: Must migrate before cancellationRefund: Pay-as-you-go, unused balance refundable
No free tier. Migrate to Cloudflare free (CDN + DNS) or self-host with Nginx caching before canceling.
Data Loss
All metric history beyond 1 dayAll APM tracesAll log dataAll custom dashboards
Feature Loss
APM entirelyLogs entirelyCustom metricsAlerting (beyond 2)15-month retention → 1 day
Gotchas
1-day retention means you effectively lose everythingDashboards saved but useless without dataOnly 5 hosts on freeNo integrations beyond basic
Cooldown: Immediate, data retention drops instantlyRefund: Annual contract, early termination fees
Downgrading Datadog is effectively canceling it. 1-day retention is useless for monitoring. Export data before downgrade.
Splunk Cloud
Enterprise → Free (Community)
Data Loss
All indexed data beyond 500MB/daySaved searches and dashboardsAlert history
Feature Loss
Unlimited data ingestion → 500MB/dayClusteringPremium appsRole-based accessDistributed searchData model acceleration
Gotchas
500MB/day on free is barely enough for a single server's logsSPL queries are Splunk-specific — no portabilityDashboards and saved searches don't transfer to free tierEnterprise apps stop working entirely
Cooldown: Annual contract, early termination fees applyRefund: Annual contract, no mid-term refund
Splunk free is effectively a different product. 500MB/day makes it a dev tool. Export to Elasticsearch/OpenSearch before downgrading.
Dynatrace
Enterprise → Free Trial (no permanent free tier)
Data Loss
All monitoring dataAll dashboardsAll alerting configurationsAll synthetic tests
Feature Loss
Full platform — Dynatrace has no permanent free tierOneAgent monitoringDavis AI engineApplication securityCloud automation
Gotchas
There is no free tier — downgrade means cancellationOneAgent instrumentation becomes dead code in your appsAll custom dashboards and alerting rules are lostDavis AI anomaly baselines are not exportable
Cooldown: Annual contract, 90-day notice requiredRefund: Annual contract, no refund
Dynatrace has no free tier. Downgrading means leaving entirely. Export dashboards as JSON, recreate monitoring in Grafana/Prometheus stack.
Twilio
Pay-as-you-go → Free Trial
Data Loss
Call/SMS logsPhone number ownershipMessaging service configs
Feature Loss
Cannot downgrade — Twilio has no free tier post-trialPhone numbers releasedAll API access revokedMessaging services deleted
Gotchas
Twilio has no free tier — only a trial with $15 creditPhone numbers are released and cannot be reclaimedAll active phone numbers start failing immediatelyVerify API stops working — user verification breaks
Cooldown: Immediate — phone numbers released within 30 daysRefund: Pay-as-you-go, no refund
No downgrade path exists. Canceling Twilio means losing phone numbers and breaking all SMS/voice integrations. Port numbers out first.