Dora NodaOn February 6, 2026, Heroku's chief product officer published a short post titled "An Update on Heroku" that made official what the industry had whispered for years: the platform that invented git-push deploy is transitioning to a sustaining engineering model. No new features. No new Enterprise contracts for new customers. The Register's headline was characteristically blunt: "Salesforce puts Heroku out to PaaSture."
Within weeks, every vendor with a PaaS to sell had published a Heroku-alternatives ranking — and every single one converges on its author's own platform as the answer. DigitalOcean's top-11 list leads with App Platform. Back4App's ranked comparison puts Back4App at number one. Railway's top-five table has Railway in the first column.
None of them are lying about Heroku's frozen roadmap. All of them are selling you something. This post is a reader's guide to extracting migration signal from pages designed to convert you — plus the one cost comparison none of those pages include.
The 5-check audit card: use it on any comparison page
Before you trust any vendor-authored ranking, run these five checks. They take ten minutes and they work on every "X vs Y" page, not just Heroku guides:
- Pricing date. Every dollar figure needs an "as of" date. PaaS pricing moves quarterly; a table quoting 2024 Heroku dyno prices against 2026 prices for the author's own platform is rigged by staleness. If the page has no date on its numbers, treat every number as marketing.
- The workload sized. A comparison is meaningless until you know the workload: how many web/worker processes, how much RAM, what database tier, what request rate. A page that sizes the author's platform at a hobby tier and the competitor at production specs is comparing two different workloads and hoping you don't notice.
- Who's missing from the table. The most honest signal on any comparison page is the omitted row. No self-hosted option? No flat-rate VPS? No mention of the author's closest direct competitor? Absences are editorial decisions — read them as such.
- The conversion tell. Find the call to action. "Migrate in one click," a signup form above the comparison table, affiliate-style "best for" badges that always point at the author — these don't invalidate the facts on the page, but they tell you exactly which conclusion the page was built to produce.
- Ops-cost honesty. Does the page price anyone's labor? A managed platform's premium is largely ops labor you don't hire; a self-hosted option's discount is ops labor you do. A page that prices only the meter and never the human is telling half the cost story — whichever half flatters the author.
Keep this card open. Now let's run it against three real pages.
The audit applied: three real rankings, scored live
DigitalOcean's "Top 11 Heroku Alternatives for Cloud Deployment in 2026" is the most comprehensive of the three: App Platform, Render, Fly.io, Upsun, Vercel, Netlify, AWS Elastic Beanstalk, Google App Engine, Azure App Service, Coolify, and Dokku. Credit where due — Coolify and Dokku making the list means self-hosting gets at least a nod, which clears check 3 better than most vendor pages.
But the ordering does the selling: DigitalOcean's own App Platform sits at the top of a list hosted on digitalocean.com, and the page's workload framing stays at the level of feature checklists rather than a sized, priced workload (check 2 fails). The conversion tell (check 4) is the domain itself. Verdict: useful as a landscape map, useless as a price comparison.
Back4App's "Best Heroku Alternatives in 2026 — Compared & Ranked" is technically accurate about the news peg — it correctly cites the February 2026 sustaining-engineering announcement, the end of new Enterprise contracts, and the classic migration drivers (no free tier since 2022, AWS-only regions, the 30-second request timeout, daily dyno restarts).
Then it ranks Back4App itself number one, followed by Sevalla, Render, Fly.io, and DigitalOcean App Platform. A self-ranked number one fails check 4 so loudly it drowns out the page's genuinely correct Heroku facts. The "ranked" framing also implies a methodology that is never shown — no scoring rubric, no workload, no prices with dates (checks 1 and 2 fail together). Verdict: trust its description of why teams leave Heroku; ignore its ordering entirely.
Bejamas' "Heroku alternatives: what to move to, and what to skip" is the interesting outlier. Bejamas is an agency, not a hosting vendor — it has no platform to sell you, which defuses check 4 at the source. Its framing is also the most technically serious of the three: long-running web services, background workers, cron jobs, and a managed Postgres are what Heroku was actually good at.
The guide evaluates destinations against that specific workload shape rather than a generic feature grid. That is check 2 done right. The bias here is subtler — an agency benefits when migrations look complex enough to need hired help — but as a reader you can calibrate for that far more easily than for a vendor ranking itself first. Verdict: the closest thing to an honest broker in this batch, and the right template for how these guides should be written.
Railway's "Top five Heroku alternatives" table and the G2 review-aggregator lists follow the same pattern at lower effort: Railway first on Railway's own blog, and G2's top-10 led by AWS Lambda and DigitalOcean on ratings volume rather than workload fit. Same audit card, same tells.
The cost table none of them print
Here is the question no vendor ranking asks: what does the same workload cost on a flat-rate box you own instead of the next metered PaaS? Let's size one typical production workload — 2 web processes plus 1 background worker, a production Postgres, and Redis, serving roughly 500 requests per minute — and price it across the platforms, using each vendor's published list prices as of mid-2026:
| Platform | Configuration | ~Monthly cost |
|---|---|---|
| Heroku | 2× Standard-1X web ($25 each) + 1× Standard-1X worker ($25) + Postgres Standard-0 ($50) + Redis Mini ($15) | ~$140 |
| Render | 3× Starter services ($7 each) + Postgres Basic ($6–7) + Key Value Starter ($10) + workspace | ~$50–60 |
| Railway | Hobby/Pro base + usage (~$20/vCPU, ~$10/GB RAM) for the same shape | ~$25–40 |
| Fly.io | 3× shared-cpu Machines + Postgres + Redis/Upstash at this scale | ~$20–35 |
| Hetzner (own the box) | One CX32-class cloud server (4 vCPU, 8 GB RAM) + IPv4, Postgres + Redis self-managed, 20 TB traffic included | ~$10 |
The metered-PaaS column comes from independent 2026 benchmarks that priced this exact workload shape across all three platforms; the Heroku column is straight list-price arithmetic any team can reproduce from the pricing page. The Hetzner row is the one no vendor prints: a single flat-rate server at roughly a tenth of the Heroku bill, with 20 TB of included transfer that dwarfs what this workload will ever use.
Now the sensitivity analysis, because a single data point is exactly the trick check 2 warns about. Two variables move this table. Bandwidth barely matters here: at 500 requests per minute of API traffic you are nowhere near any platform's included transfer, and Hetzner's 20 TB ceiling means egress never enters the conversation.
Ops labor is the variable that actually reorders the ranking — self-managing Postgres backups, failover, and OS patching on a $10 box costs engineering hours that the metered platforms absorb into their price. For a team with zero ops capacity, Railway's ~$30 or Render's ~$55 is genuinely cheaper than $10 plus a pager rotation. For a team that already operates infrastructure — or runs a platform that automates it — the $10 row is real money, roughly $1,500 a year saved against Heroku for this one workload, scaling linearly with every additional service.
There is a second sensitivity worth naming: the table above is the startup tier of this workload. Push it to growth scale (more replicas, bigger database, Redis cluster) and the metered bills roughly triple while the flat-rate box merely steps up one server size. The own-a-box advantage widens with scale — which is precisely why vendor-authored pages size their comparisons at the hobby tier where the gap looks smallest.
What the rankings get right anyway
Fairness matters, because a reader who dismisses every vendor page as pure propaganda will miss the real signal buried in the sales pitch. Three things the rankings consistently get right:
First, the news peg is real. Heroku's sustaining-engineering announcement is not FUD — it is a published statement from Heroku's own CPO, and "no new features, no new Enterprise contracts" is exactly the kind of roadmap freeze that should trigger a migration evaluation for any team planning multi-year bets.
Second, the technical critiques are accurate. The 30-second request timeout, daily dyno restarts, AWS-only regions, and sleeping behavior are genuine operational constraints that competing platforms genuinely don't share. When Back4App or DigitalOcean lists these as migration drivers, believe the list — just don't believe that the author's platform is the only escape.
Third, for teams that will never self-host, the top picks are genuinely reasonable destinations. Render's Heroku-like git-push experience with preview environments, Railway's usage-based pricing for spiky workloads, Fly.io's global edge footprint — these are real differentiators matched to real workload shapes. The rankings' sin is presenting a sales conclusion as a research conclusion, not recommending bad platforms.
Each page serves a reader profile: DigitalOcean's list serves the browser still mapping the landscape; Back4App's serves the comparison shopper who wants a ranked answer without doing the math; Bejamas' serves the team that already knows its workload shape and wants a serious destination analysis. Know which reader you are before you let any of them choose for you.
Migrate once, on your own numbers
The gold rush of 2026 Heroku rankings will keep paying dividends to its authors for years — every "should we leave Heroku" thread now surfaces a comparison page with a signup form attached. Your defense is procedural, not tribal: run the five-check audit card on any page that ranks its author's platform first, price your actual workload instead of the page's implied one, and always add the row none of them print — the flat-rate box you own.
Do that, and the migration decision stops being a referendum on vendor marketing and becomes what it should have been all along: arithmetic about your workloads, your scale, and your team's ops capacity. Migrate once, on your own numbers, to a platform whose pricing you verified yourself — not one that verified itself.
Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own. Star the repo on GitHub or deploy your first app today.
