When Highlight.io’s lights went off on February 28, 2026, a few thousand teams woke up to a problem they hadn’t planned for. The post-acquisition email said: migrate your SDK to LaunchDarkly Observability before March 1, or lose your ingest. That’s a fine outcome if LaunchDarkly is where you wanted to end up. It’s decidedly less fine if you signed up for an open-source observability tool and now find yourself staring down a feature-flag platform’s pricing page.
The Highlight shutdown is the cleanest forcing function the error-tracking market has had in years. Six weeks later, plenty of teams are still mid-migration. Meanwhile Sentry’s high-traffic tiers keep inching up, PostHog has quietly shipped a usable error-tracking module, and the self-hosted contingent — Bugsink, SigNoz, GlitchTip — is having its loudest moment since Sentry went source-available back in 2019.
This is the post I wish I’d had three weeks ago when a friend asked where to migrate his Highlight setup. Five trackers, what they actually do, what they actually cost, and the picks I’d make today by stack. Pricing here is current as of early May 2026 — verify the vendor pages before you commit a budget, because every one of them tweaks something quarterly.
The five bets, one sentence each
Sentry is the incumbent and still the most polished error tracker money can buy — at a price that gets uncomfortable past a few hundred thousand events a month.
PostHog has folded error tracking into its product-analytics suite and gives you 100K events a month free, which is the most generous starter floor on this list.
Bugsink is a Sentry-SDK-compatible drop-in you can self-host on a $5 Hetzner box and forget about, with an EU-hosted SaaS option for teams that don’t want to operate it.
SigNoz is the OpenTelemetry-native option — logs, traces, metrics, and errors in one app, self-hostable or as a fully-managed cloud product.
Datadog Error Tracking is the right choice for exactly one team: the one already paying Datadog enough that adding error tracking is rounding error.
I don’t think there’s a single winner here. There’s a single winner per situation, which is what I’ll get to.
What “error tracking” even means in 2026
The labels have drifted. “Error tracking” used to mean stack traces with grouping. Now every vendor on this list will sell you some subset of:
- Stack trace capture, source-map resolution, breadcrumbs
- Release tracking and regression detection
- Session replay (rage clicks, the works)
- Performance and transaction tracking
- Distributed tracing
- Profiling
- Log search
Sentry has the most mature implementation of the first four. PostHog ships replay and analytics linked to errors but has a thinner stack-trace UI. SigNoz leans into traces and logs, with errors as a happy side-effect of OTel ingestion. Bugsink stays narrow on purpose — it’s an error tracker, not an APM. Datadog will sell you all of it as long as you keep signing renewal terms.
If you only care about the first bullet, you don’t actually need most of these tools. You probably need Bugsink or GlitchTip and an afternoon. If you care about replays and analytics linked to errors, the calculus changes a lot.
The pricing math, done honestly
Here’s the comparison I keep getting asked for. Numbers verified against the vendor pricing pages in early May 2026 — check them before signing anything.
At ~100K events/month — small startup, side project, internal tool
| Tool | Cost | What you give up |
|---|---|---|
| Sentry | $80/mo (Business) or $26/mo on Team | Nothing serious |
| PostHog | $0 (under 100K free tier) | Nothing |
| Bugsink Cloud | $16/mo on the 75K plan | Less polish than Sentry |
| Bugsink self-hosted | ~$5/mo (Hetzner box) | An hour of setup |
| SigNoz Cloud | $30-50/mo depending on log volume | Backend-heavy product |
| Datadog | No standalone option | — |
At ~1M events/month — Series A SaaS, growing product
| Tool | Cost | Notes |
|---|---|---|
| Sentry | $300-500/mo | Team tier + overages, or negotiated Business |
| PostHog | ~$370/mo after free tier ($0.00037/event) | Bundles replay and analytics in the same bill |
| Bugsink Cloud | $70-120/mo | Best price-to-feature on hosted |
| Bugsink self-hosted | ~$10/mo of compute | Operational overhead is hours/year |
| SigNoz Cloud | $200-400/mo with traces | OTel ingest is volume-dense |
| Datadog | Custom; expensive | Only worth it bundled |
At 10M+ events/month — mid-size org or chatty mobile app
This is where the spread blows up. Bugsink’s own blog has a worked comparison putting roughly $6,400/month on Sentry against ~$1,300/month on Bugsink at 50M events. I’m not going to vouch for every line item — vendor-authored math always reads favorably for the vendor — but the order of magnitude matches what I’ve seen at three companies that switched. Sentry negotiates hard at this volume and the list price gap shrinks, but it doesn’t close.
The summary that matters: under a million events, ergonomics differences swamp price differences. Above that, the pricing model — per-event vs per-GB vs flat-self-host — starts dominating the bill.
The Highlight → LaunchDarkly migration is messier than the docs suggest
LaunchDarkly’s official line is that the migration is an SDK swap and you’re done. The Highlight.io blog still hosts a clean step-by-step guide. In practice I’ve watched two teams attempt this and hit the same things:
Data history doesn’t move automatically. You email observability@launchdarkly.com to discuss migrating existing sessions and errors. For a lot of teams the answer is “we’ll just start fresh,” which is fine until someone needs to look at last month’s incident.
The pricing reality is different from the Highlight pricing reality. LaunchDarkly is a feature-flag company first; observability is bundled but priced for accounts already paying LD. If you weren’t on LaunchDarkly before, the renewal conversation is going to include a lot more than observability.
The product is genuinely converging on something interesting — feature flags, observability, experiments tied together. Whether you want to be locked into LaunchDarkly’s roadmap to get that combo is the actual question. I’d think hard before saying yes for an indie or bootstrapped team.
If you were on Highlight specifically because of the open-source story, the migration is also a values shift. LaunchDarkly Observability is not open-source. You can read that as “fine, the OSS story didn’t matter to me” or as “this is exactly why I picked Highlight in the first place.” Both reads are valid. Don’t let LaunchDarkly’s email cadence make the decision for you.
When I’d still pick Sentry
Sentry does one thing nobody else on this list matches: error grouping. When you have ten thousand instances of TypeError: cannot read property 'foo' of undefined coming from sixteen different release artifacts, Sentry’s fingerprinting collapses them sensibly. Everyone else still has lumpy grouping. PostHog’s is improving but inconsistent. Bugsink’s is fine. SigNoz’s is fine for traces, less polished for errors specifically.
Source map handling is the other place Sentry coasts on years of investment. If you ship minified frontends — and you do — the experience of clicking a stack frame and landing in source is meaningfully better on Sentry. Release health, regression detection across deploys, the integrations with Linear/Jira/Slack — all of it is just done on Sentry in a way it isn’t on the alternatives.
What you pay for that polish is the bill. Specifically, the bill at the boundary where you’re not big enough to negotiate a custom contract but big enough that the Team plan’s overages are eating you. A million events a month is roughly where I see teams start asking “can we do this for less.”
PostHog’s case is consolidation, not features
If you’re reading this and you already pay PostHog for product analytics, the math is short. Errors are bundled into the same free tier, the same dashboard, the same session replays. You get error events linked to specific users and the replay of what they were doing when things broke. That’s a story Sentry can’t tell as cleanly.
The catch: PostHog’s error-tracking module is two years younger than its analytics module, and it shows. Stack-frame UI, source map handling, grouping — all of it feels like it’s catching up rather than leading. If errors are your primary observability surface, you’ll miss Sentry. If errors are one of five things you’re tracking and you’d rather have one bill, PostHog is the consolidation play of 2026.
The 100K free-events tier is also genuinely the best free starting point on this list. For a side project or a bootstrapped early-stage thing, you can run on the free tier for a long time before payment becomes a question at all.
Bugsink is the self-host story that finally doesn’t suck
I’ve run self-hosted Sentry. I do not recommend it. The 16GB RAM minimum, the Snuba/ClickHouse/Kafka mess, the upgrade pain — none of it is worth the savings unless you’re huge, and even then you have to want it.
Bugsink replaces all of that. Single binary deploy, runs on a $5 Hetzner box, speaks the Sentry SDK protocol so your existing client libraries just work. Change the DSN and you’re done. The self-hosted version is free for non-competing production use under the Polyform Shield License — read the license before you commit, but for a normal product company it’s friendly.
Where Bugsink isn’t a fit: if you need session replay, performance tracing, or profiling, it can’t help you. It’s deliberately scoped to error tracking. That’s a feature in my view; some teams will want more.
The hosted EU-residency option is the other selling point. If you’re a European team trying to keep error data inside EU borders without operating your own infra, Bugsink Cloud is the path of least resistance, starting at $16/month for 75K events. Sentry has EU data residency too, but the bill is on a different planet.
SigNoz fits a different shape
SigNoz is the right answer if you’re already an OpenTelemetry shop, or if you’re consolidating away from Datadog and want errors, traces, logs, and metrics in one app. The ingest model — $0.30/GB for logs and traces, $0.10 per million samples — is friendly to high-volume backend services where event counts misrepresent the actual data shape.
The OTel-native angle is genuine. Most error trackers reluctantly accept OTel; SigNoz was built around it. If you have a polyglot backend — Go, Python, Rust, Elixir, whatever — instrumenting once with OTel and letting SigNoz pick up errors as a side effect is materially less work than wiring SDK-specific clients to Sentry.
What you give up is frontend polish. If your problem is a React Native app crashing for 0.4% of users on a specific device, Sentry or PostHog will give you a better experience. SigNoz’s strengths point backend, and that’s where I’d use it.
Datadog Error Tracking only makes sense if you’re already there
I’ll keep this short because the answer is short. Datadog Error Tracking is competent. It is not better than Sentry, it is not cheaper than anything on this list, and there is no free tier worth the name. The case for it is that you’re already paying Datadog for APM and ops monitoring, and you’d rather have one pane of glass than a separate vendor.
That’s a real case for ops-led organizations. It is not a case for almost any product engineering team I’ve worked with. If you’re choosing fresh in 2026 and Datadog is a question, the answer is no.
PII, compliance, and the EU question
A few quick notes that come up in every adoption conversation:
- Sentry, PostHog, and Datadog all have HIPAA BAAs available on enterprise plans. Bugsink Cloud and SigNoz Cloud are more limited — check current docs before assuming.
- EU data residency: Sentry (paid add-on), Bugsink Cloud (default for EU), PostHog (EU cloud option), SigNoz (depends on tier), Datadog (yes, paid). If you self-host Bugsink or SigNoz, residency is whatever your VPS provider offers.
- Data scrubbing for PII is roughly equivalent across all five — they all let you redact server-side and ship custom rules.
If compliance is your hard constraint, narrow the list to Sentry Business, Bugsink Cloud (EU-only), or self-hosted Bugsink/SigNoz on a box you control. PostHog Cloud EU is fine for most situations but carries fewer certifications than Sentry today.
The picks, by who you are
A B2C startup under a million events a month with product analytics already in PostHog: stay on PostHog and use its error tracking. One bill, one dashboard, one fewer vendor.
A B2B SaaS company at any scale where engineering velocity matters more than the bill: Sentry. The grouping, source maps, and ecosystem save more in engineering time than the price difference costs.
A bootstrapped indie team, or anyone told to cut infra spend this quarter: Bugsink, self-hosted on a cheap VPS. The setup is one evening; operating cost is a coffee a month.
An EU-based company that takes data residency seriously: Bugsink Cloud (EU) or self-hosted Bugsink. Sentry’s EU offering works but is overkill for most teams that fit this profile.
An OpenTelemetry-native backend with no strong frontend story: SigNoz. The traces-and-errors story is genuinely better than anyone else’s at this stack shape.
A team already paying Datadog north of $10K/month: Datadog Error Tracking, because the marginal cost of adding it is small and the consolidation is real. Otherwise, not Datadog.
A team migrating from Highlight: think hard before defaulting to LaunchDarkly Observability. The SDK swap is easy; the long-term lock-in is not. PostHog and Sentry are both reasonable destinations. Bugsink is the answer if open-source values were why you picked Highlight in the first place.
If you’ve got a few hours this week, the cheapest experiment is to spin up Bugsink in Docker, point one of your services’ DSNs at it, and see how it feels. You’ll know within an afternoon whether you can stop paying someone else for stack traces — and if not, you’ve at least narrowed the question to “which paid one.”