.svg)
The Amplitude Account Health Check: 10 Things to Audit Every Quarter
Amplitude implementations don’t break all at once, they drift. A site update quietly kills a tag, a new feature ships without tracking, a definition shifts, costs creep up, and none of it announces itself. The dashboards keep loading, so everyone keeps trusting them, right up until a number doesn’t add up in a meeting. A quarterly health check is how you catch that drift before it corrupts the decisions built on top of it. Here are the ten things to audit, and what each one is really telling you.
An Amplitude account health check is a periodic audit of whether your implementation can still be trusted — because tracking drifts as your product changes. Run it each quarter across ten areas: event taxonomy, event firing and volume, user identity, data accuracy against your source of truth, MTU usage and cost, dashboards and metric definitions, key funnels and reports, data governance and access, integrations, and whatever’s changed since last time. The goal is to catch the quiet problems — a broken tag, a drifting definition, a cost spike — before they break the reports and decisions that depend on them.
Why run a quarterly Amplitude audit
Your product changes constantly, and your tracking has to keep up or it silently falls out of sync. New features ship, pages get redesigned, SDKs update, teams redefine metrics, and event volume shifts — each a chance for something to break without an alert. A health check is the deliberate, scheduled moment where you go looking for those problems instead of waiting for them to surface in a report. Quarterly is the sweet spot for most teams: frequent enough to catch drift early, rare enough not to be a burden.
Work through these ten, top to bottom.
1. Event taxonomy and naming
Check: Scan your event list for inconsistent naming, duplicates (Checkout Complete vs checkout_completed), and orphaned events no one uses. Confirm it still matches your tracking plan. Why it matters: Inconsistent naming means the same behavior gets logged under different names, so every analysis silently under- or over-counts. A clean event taxonomy is the foundation everything else sits on — this is the first thing to fix if it’s messy.
2. Events actually firing
Check: Look at event volume over time. A key event trending toward zero, or missing entirely, usually means a broken tag. Confirm any features shipped since last quarter are actually instrumented. Why it matters: A silently broken event poisons every metric built from it, and because nothing errors, no one notices for weeks. New features that shipped without tracking are invisible in your data — a common and costly gap.
3. User identity resolution
Check: Compare Amplitude’s user count against the number of real users you know you have. A big gap signals fragmenting identity — one person splitting into several as anonymous and known sessions fail to stitch. Why it matters: Fragmented identity distorts every user-level metric downstream, from retention to LTV. It’s especially common with non-email logins, and worth fixing early because so much depends on it.
4. Data accuracy against your source of truth
Check: Spot-check a few key numbers in Amplitude against your backend or finance source — total signups, revenue, conversions. They won’t match perfectly, but large unexplained gaps are a red flag. Why it matters: If Amplitude can’t be reconciled to a defensible range, you can’t fully trust it for decisions. This check tells you whether your implementation reflects reality or has quietly drifted from it.
5. MTU usage and cost
Check: Review your monthly tracked users against your plan, and watch for unexpected spikes. Identify high-volume, low-value events you could stop sending. Why it matters: MTU drives your Amplitude cost, and a stray high-firing event (or a bot problem) can blow through your allocation. Pausing events you don’t analyze protects both your budget and your data cleanliness.
6. Dashboards and metric definitions
Check: Audit your most-used dashboards. Are there duplicate or stale charts? Does each key metric have one agreed definition, or does “active user” mean three different things across teams? Does each dashboard have an owner? Why it matters: Multiple definitions turn every cross-team conversation into a translation problem, and orphaned dashboards decay. One definition per key metric is what keeps the whole org steering by the same numbers.
7. Key funnels and reports still valid
Check: Open your most important funnels and confirm they still work — product changes often break a step or make a required event stop firing. Watch for funnels returning implausible numbers. Why it matters: A funnel that silently broke keeps producing a number, just a wrong one. Validating your critical-path funnels ensures the reports leadership relies on still reflect the actual journey.
8. Data governance and access
Check: Review who can create events (uncontrolled creation is how taxonomies rot), confirm any planned-or-blocked events are enforced, and verify PII masking is in place where it should be — especially on Session Replay. Why it matters: Governance is what stops a clean implementation from degrading over time. And a PII leak is a compliance risk, not just a data one — worth checking every quarter.
9. Integrations and data sources
Check: Confirm your connected sources and destinations are still syncing — ad platforms, your CDP, and tools like Braze or HubSpot. Look for syncs that have silently failed or fallen behind. Why it matters: A broken integration means stale data flowing in (or cohorts not flowing out to your marketing tools), and like everything else here, it usually fails quietly. If Amplitude powers downstream activation, a dead sync has real revenue impact.
10. What’s changed since last audit
Check: List the product releases, redesigns, new pages, and SDK updates since your last check, and trace whether any affected tracking. Why it matters: Change is where tracking breaks. Reviewing what shipped — and confirming it didn’t quietly break instrumentation — is the single highest-yield habit of a healthy Amplitude account. Most of the issues above trace back to a change no one checked.
Make it a habit, not a one-off
The point of a health check isn’t the one-time cleanup — it’s the cadence. Amplitude drifts continuously, so auditing once and walking away guarantees it goes stale again. Put this on a quarterly recurring calendar, keep a short log of what you find and fix, and your account stays trustworthy instead of slowly rotting while the dashboards keep loading. (Many of the issues here also trace back to the same root causes as the most common Amplitude configuration mistakes — worth a read if you keep finding the same problems.)
Want your Amplitude account pressure-tested?
Running this yourself is a strong start. If you’d rather have it done rigorously — or you’re managing Amplitude across multiple products and want ongoing monitoring rather than a once-a-quarter scramble — our Data Stack Audit puts your Amplitude implementation through exactly this, with a prioritized roadmap for what to fix first.
Frequently asked questions
What is an Amplitude account health check?
It’s a periodic audit of whether your Amplitude implementation can still be trusted — covering event taxonomy, whether events are firing, user identity, data accuracy, MTU cost, dashboards and definitions, key funnels, governance, integrations, and what’s changed since last time. Because tracking drifts as your product changes, the check catches quiet problems before they corrupt your reports.
How often should you audit your Amplitude setup?
Quarterly works for most teams — frequent enough to catch tracking drift early, infrequent enough not to become a burden. Fast-moving products that ship constantly may benefit from monthly spot-checks on event firing and new-feature instrumentation, with a fuller audit each quarter.
What are the most common Amplitude issues to look for?
Broken or missing events (a tag that stopped firing or a feature shipped without tracking), fragmented user identity inflating user counts, metric definitions that differ across teams, MTU cost creep from high-firing low-value events, and funnels broken by product changes. Most of these trace back to a change no one verified.
How do you know if your Amplitude data can be trusted?
Reconcile a few key numbers against your source of truth (backend or finance), confirm your critical events are firing at expected volumes, and check that your user count isn’t wildly inflated by identity fragmentation. If those hold up, your data is broadly trustworthy; large unexplained gaps mean it’s drifted.
How does MTU affect my Amplitude cost?
Monthly Tracked Users is a primary driver of Amplitude pricing, so unexpected spikes — from bot traffic, a misfiring high-volume event, or growth — can push you over your plan. Reviewing MTU each quarter and pausing events you don’t actually analyze keeps both cost and data quality in check.
