SaaS engineering · Updated May 2026
SaaS observability without the per-host pricing trap.
Per-host APM bills punish redundancy; per-tenant slicing demands one query layer; security needs the same span data your APM already pays for. Here's a SaaS-native take on observability — and what one tool that does it all looks like.
The pricing model is the architecture
Tell me how an observability vendor charges, and I'll tell you what they assumed about your business. Per-host charging assumes you scale infrastructure with revenue. SaaS doesn't. SaaS scales tenants with revenue. The host count goes up because of multi-AZ redundancy, blue/green deploys, and burst pods — none of which produce proportionally more useful data.
Per-TB-scanned pricing inverts this. Every byte you query is a byte an engineer found valuable. The bill rewards instrumentation that actually gets used.
Five things SaaS observability needs that single-tenant observability doesn't
- tenant.id everywhere.Spans, logs, errors, security events. If you can't filter by tenant in one query, you don't have observability — you have a search engine.
- Per-tenant SLOs.Aggregate p99 hides the one customer who's about to churn over a 4-second checkout.
- Per-tenant cost attribution.Which customer is chewing 80% of your CPU? Without tenant-aware tracing you can't price-up your enterprise tier honestly.
- Security that travels with the span. An attack that targets one tenant is invisible in fleet-wide latency dashboards. ASM data needs the same tenant tags.
- Per-tenant kill switches. When a single tenant has a bad day — abusive integration, accidental DDoS — you need a dial, not a deploy.
The reference architecture
- OpenTelemetry SDKs in every service.
tenant.idas a baggage attribute, propagated through every downstream call. - One collector (OTel collector or vendor) that ingests everything to one query layer.
- One query layer (ClickHouse for self-hosters; a SaaS for everyone else) — APM dashboards, logs, and security events on the same indices.
- One enforcement plane — IP firewall + per-tenant blocklist + automation rules, controlled from the same UI as the dashboards.
- One billing axis — data scanned, not host count or seat count.
What SecureNow does
SecureNow ships exactly that stack as one product. npm install securenow instruments every Node framework with a tenant-aware OTel SDK. The dashboard splits everything by service.name (per app), tenant.id (per customer), and time. The firewall has a per-app toggle (UI, CLI, or API) so you can suspend protection on a single environment without a redeploy. Pricing is $5 per TB scanned with a free tier.
See it in detail on the Datadog comparison, the Sentry comparison, or the application security monitoring guide.
Frequently asked
What is SaaS observability?
Observability for software-as-a-service applications: distributed traces, logs, metrics, and security events from a multi-tenant or single-tenant SaaS, correlated so you can answer 'why is customer X slow' or 'is anyone abusing our API'. Different from generic observability mainly because of multi-tenancy: every signal needs a tenant ID, and per-tenant slicing is table stakes.
Why does per-host pricing fail SaaS?
SaaS hosts scale with redundancy and burst capacity, not with revenue. A 30-pod Kubernetes deployment serving 50 customers and one serving 5,000 cost the same on per-host pricing — but the second one is producing 100x more useful telemetry. Per-TB pricing matches the bill to the value.
Should I use one tool or several?
Several only if you genuinely have separate teams with separate budgets. The friction tax of three dashboards usually outweighs the marginal feature wins. One OpenTelemetry-native tool that does APM + logs + security + firewall is the 2026 default for most SaaS shops under ~200 engineers.
What about per-tenant cost analysis?
Tag every span with `tenant.id` and the dashboard splits costs and SLOs by customer. SecureNow does this natively; in fragmented stacks you'll write three queries.
How does SecureNow handle multi-tenant security?
Per-application API keys (each tenant or environment gets its own), per-app firewall toggle (the new Firewall menu we shipped in v7.3.0), per-tenant block/allow lists, and tenant-scoped AI investigations. One backend, isolated views.
SaaS observability priced like SaaS.
1 GB/month free, $5/TB after. Add the firewall. Stop budgeting around redundancy.