Hidden RUM Costs 2026
The published per-session rate is not the bill. Sampling, retention, replay storage, overage rates, and ingest meter sharing all produce real cost variance. Here's what to model before signing.
Cost 1: Sampling vs full capture
Most RUM vendors default to 100% capture. Reducing sampling to 25% reduces the bill by 75%. For non-critical user segments (anonymous visitors, low-engagement pages), 10-25% sampling is operationally sufficient, the statistical signal at the 75th-percentile measurement Google uses doesn't require 100% capture.
Failure mode: enterprise teams default to 100% sampling because “we don't know what we'll need”, then sign a contract that quotes 5M sessions/month. Six months later, they realise 25% sampling would have captured the same signal at 25% of the bill.
Cost 2: Retention surcharges
Standard retention varies by vendor:
- 30 days: Datadog, AWS CloudWatch RUM, LogRocket Core (cheapest tier), Sematext Basic.
- 35 days: Dynatrace RUM.
- 60 days: Raygun.
- 90 days: Sentry paid tiers, LogRocket Team.
- 12 months: Fullstory free tier.
Extending retention costs roughly: 90 days adds 20-30%; 13 months adds 50%+; multi-year compliance retention can double the bill. Most RUM vendors don't publish extended-retention rates, quote-only conversation.
Cost 3: Session replay overages
Vendors that bill replay separately (Sentry, Dynatrace) charge per-replay overage above included allowance. At 50% replay coverage on 1M sessions, replay overage alone can be $1,000-$2,000/month, often more than the base RUM subscription.
Cost 4: Shared ingest meter overflow
New Relic and Elastic price per GB of telemetry ingested across all products. Adding RUM is cheap in isolation but pushes the shared meter higher, accelerating tier-ceiling overages on the other products (APM, logs, infrastructure). Model total ingest, not RUM ingest.
Cost 5: Mobile RUM as separate meter
Datadog Mobile RUM, Dynatrace Mobile RUM, and Fullstory Mobile are separately-priced from their web RUM products. Teams instrumenting both web and mobile pay double meters. Plan for the dual-meter scenario before signing for web only.
Cost 6: Enterprise feature gating
SSO/SAML, audit logging, custom dashboards, advanced alerting (Core Web Vitals regression alerts on Raygun, conditional recording on LogRocket) are typically gated to higher tiers. Mid-market teams find they need the enterprise tier for features the cheapest tier doesn't include.
Cost 7: Annual commit overage exposure
Annual-commit pricing is cheaper per session, but overage above the committed volume is typically billed at on-demand rates (33% premium). A bursty workload that exceeds annual commit can produce surprise year-end bills.
Cost 8: SDK upgrade cost
SDK upgrades happen 2-4 times per year for active vendors. Each upgrade requires QA against your application, typically 1-3 hours of engineering time. Across 4 upgrades per year per project, that's 4-12 hours per project per year of soft engineering cost not on the invoice.
Cost 9: Alert noise and on-call burden
New RUM alerts produce noise during the first 30-60 days while you tune thresholds. The on-call burden during this period is real but typically not modelled in procurement. Allow 0.1 FTE of engineering attention for the first 60 days post-launch.
Cost 10: Sub-account migration friction
Switching RUM vendors or consolidating across acquired-company sub-accounts requires SDK swap-out, alert reconfiguration, historical data export, dashboard rebuild. Budget 2-4 weeks of engineering time for a clean migration; budget 6-8 weeks for a multi-property consolidation.
For the cost-reduction playbook, see how to reduce RUM costs.