Skip to content
Concept demo

A fictional brand, designed and built by Get Ranked On Top. Not a real company.

Skip to content
4.2Tail-based sampling is generally available

Find the slow span before your customers do.

Signalyard is the OpenTelemetry-native console for traces, metrics and logs. One waterfall, one query language, one bill priced by the gigabyte.

prod · ca-central-1Traces7f3a9c04e1d2b6a8…e21b
Last 1 h
p99 latency

612ms

+2.1% vs 1h

Error rate

0.12%

−0.03 pts

Throughput

18.4k rpm

+6.8% vs 1h

Saturation

63% pool

pg-orders-primary

Trace waterfall · 12 spans · 9 services612 ms
CPU profile · checkout-svc · linked to this trace

Live mock. Click any span to trace its path, or flip Inject latency to watch a pool stall propagate.

Paging fewer engineers at 1,900 companies, from seed-stage to the public markets.

  • HALCYON FREIGHT
  • moraine/bank
  • Quillstack
  • PARALLAX HEALTH
  • tundra·games
  • Ordinal Pay
  • Lumen&Ash

The console

Built for the fourth minute of an incident.

Not the first, when everyone is still opening laptops. The fourth, when someone needs to say which service, which query and which deploy. Every screen is designed to answer that out loud.

Click

Every span is a door

Select any span and Signalyard walks the parent chain, pins the critical path and puts its attributes, logs and profile one keystroke away.

G then P

Profiles stitched to traces

Continuous profiling is joined on span ID, so the flame graph you see is the CPU that ran during this request — not an hourly average.

B

Baseline, not vibes

Each operation carries a rolling seven-day p50/p99. Deviations are flagged in the waterfall before anyone has written an alert rule.

⌘ K

Query in the language you have

Filter with SQL, PromQL or our trace query syntax. Saved views become dashboards; dashboards become SLOs with one toggle.

Platform

Three signals. One storage engine. Zero tab-switching.

Traces, metrics and logs share one columnar store and one set of resource attributes, so a click on any of them lands you in the other two with the context already applied.

traces

Every hop, on one ruler.

W3C trace context end to end, span links for async work and queue consumers, and a waterfall that renders 40,000-span traces without a spinner.

  • Tail-based sampling
  • Span links across Kafka and SQS
  • Critical-path highlighting

metrics

Cardinality is not a pricing tier.

Exemplars on every histogram bucket, so any point on a p99 chart opens the exact trace that put it there. Tag with tenant ID; we bill bytes, not series.

  • Native OTLP and Prometheus
  • Exemplars to traces
  • SLOs with burn-rate alerts

logs

Logs that know which request they came from.

trace_id and span_id are parsed on ingest, so the log panel beside a span shows only the lines it wrote. Full-text search over 30 days in under two seconds.

  • Structured and plain-text
  • Automatic trace correlation
  • Field-level redaction

OpenTelemetry-native

No agent. No SDK of ours. No lock-in.

Signalyard speaks OTLP natively. Your instrumentation stays vendor neutral, so leaving us is an environment variable — which is exactly why teams stay.

  1. 01

    Export

    Set three environment variables. Any OpenTelemetry SDK or Collector ships to Signalyard over OTLP/gRPC or OTLP/HTTP.

  2. 02

    Sample at the tail

    Keep 100% of errors and anything over your p95. Drop the boring 90% after the trace completes, not before.

  3. 03

    Ship it

    Median time from first export to first full trace across new workspaces last quarter: 11 minutes.

shellany OTel SDK
# 1 · point any OpenTelemetry SDK at Signalyardexport OTEL_EXPORTER_OTLP_ENDPOINT="https://ingest.ca-central.signalyard.example"export OTEL_EXPORTER_OTLP_HEADERS="x-sy-token=sy_live_••••••••••••"export OTEL_SERVICE_NAME="checkout"export OTEL_RESOURCE_ATTRIBUTES="deployment.environment=prod,service.version=4.18.2"# 2 · run your service exactly as beforenode --require @opentelemetry/auto-instrumentations-node/register server.js
otelcol.yamlCollector 0.112+
processors:  tail_sampling:    decision_wait: 10s    policies:      - name: keep-errors    type: status_code      - name: keep-slow      type: latency  threshold_ms: 450      - name: baseline       type: probabilistic  pct: 5  signalyard/redact:    rules: [pan, email, bearer_token]

Go · Java · Python · Node · .NET · Ruby · Rust · PHP · eBPF

Pricing

One meter: gigabytes. Hosts, seats and series are free.

A 40-service platform sending 6.2 TB a month on Scale pays $992. That number does not move when you add engineers or tags.

Free

$0up to 50 GB / month

For side projects and the first service you instrument.

  • Traces, metrics and logs
  • 7-day retention
  • Unlimited seats
  • Community support
Start free

Team

Most teams

$0.24per GB ingested

For product teams running production on OpenTelemetry.

  • 30-day retention
  • Tail-based sampling
  • Profiles linked to traces
  • Email support, 1 business day
Start 30-day trial

Scale

$0.16per GB, 20 TB / month commit

For platform teams consolidating three tools into one.

  • 90-day retention
  • SAML SSO and SCIM
  • Regional residency
  • 1-hour P1 response in a shared channel
Talk to an engineer
Plan comparison
FeatureFreeTeamScaleEnterprise
Traces, logs, metricsPer GBPer GBPer GBPer GB
Hosts, seats, seriesFreeFreeFreeFree
Retention7 days30 days90 daysUp to 13 months
Tail-based sampling—YesYesYes
Continuous profiling—YesYesYes
SAML SSO / SCIM——YesYes
Data regionsca-centralca-central, us-east+ eu-west+ dedicated cell
SupportCommunity1 business day1 h for P1Named engineer

Prices in USD, billed monthly on uncompressed OTLP bytes after sampling. Enterprise starts at 100 TB a month; contact sales for a dedicated cell.

Customers

Measured in minutes not spent on a bridge call.

lower median time-to-resolve after 90 days, across 140 teams that migrated in 2025
63%
average reduction in observability spend versus the customer's previous vendor
41%
median time from first OTLP export to first complete trace
11 min
spans ingested last month, queried at a p99 of 1.8 s
2.4T

P1 MTTR 47 → 14 min

“We used to grep three tools during an incident. Now the on-call opens one waterfall, clicks the red bar and reads the db.statement. Our P1 retros got shorter because there is less to argue about.”
Priya RamaswamyStaff SRE, Halcyon Freight

Bill down 38% year over year

“Per-GB pricing is the only model where adding a high-cardinality tag doesn't trigger a finance meeting. We tagged every span with tenant ID on day two.”
Marc-Antoine DucharmeHead of Platform, Quillstack

100% of error traces retained

“Tail-based sampling was the unlock. We keep every error and every slow checkout, drop the rest, and the trace you need is always there when the pager goes off.”
Hannah OkaforEngineering Manager, Ordinal Pay

Security

Your telemetry is a map of your company. We treat it that way.

servicesOTLPcollectorredact · sampleyour VPCregionyour KMS key
  • Scrub at the edge

    PII rules run in our Collector distribution inside your VPC. Card numbers and emails are redacted before a byte leaves your network.

  • Choose your region

    Data stays in the region you pick at workspace creation: Toronto, Virginia or Frankfurt. No silent cross-region replication.

  • Bring your own key

    Envelope encryption with keys held in your cloud KMS. Revoke the key and the data is unreadable, including to us.

  • SSO, SCIM, roles

    SAML 2.0 and OIDC, SCIM provisioning, and per-service read scopes so contractors see the payments team's spans only if you say so.

  • Audit everything

    Every query, export and permission change is written to an immutable audit stream you can ship to your own SIEM.

  • Documented controls

    Independent assessment reports, our pen-test summary and a pre-filled security questionnaire are available under NDA.

Start free

Your first trace in eleven minutes. Your first bill at 50 GB.

No credit card, no sales call, no agent to install. If you would rather talk to an engineer first, write to hello@signalyard.example.

Demo form — nothing is sent. Data region defaults to ca-central.

FAQ

Questions engineers actually ask.

What exactly counts as a gigabyte?

Uncompressed OTLP payload after your sampling and scrubbing rules run, measured at our ingest edge. Data you drop at the Collector is never billed. You can see the meter per service, per signal, in real time.

Do I have to replace my existing instrumentation?

No. If it speaks OpenTelemetry, Prometheus remote-write or Fluent Bit, it already works. Most teams change an exporter endpoint and nothing else.

How does tail-based sampling work without losing data?

Our Collector buffers spans until a trace completes, then applies your policies — keep errors, keep anything slower than p95, keep 5% of the rest. Decisions are made with the whole trace in view.

Is there a minimum contract?

Free and Team are month to month with no minimum. Scale is an annual commitment in exchange for the lower per-GB rate; unused volume rolls over for one quarter.

Can we self-host?

The Collector distribution and query SDKs are open source. The storage and query engine run as a managed service, or as a dedicated cell in your cloud account on Enterprise.

What happens if we exceed our plan?

Nothing breaks. We keep ingesting, warn at 80% and 100% of your budget, and you can set a hard cap per service if you would rather drop than pay.