Running Amplitude?

Amplitude funnels disagreeing with the database? We fix it.

Amplitude starts as a tidy list of events and ends with duplicate properties, orphaned user IDs, and a billing warning about the monthly tracked user count. We rebuild the taxonomy, fix the identity resolution, line up your funnels with the database numbers, and bring the MTU count back inside the plan.

  • Since 2015Fixing technology
  • FreeEstimates
  • US teamReal engineers

amplitude-ingest.log

still broken

What it is

Amplitude at a glance

Amplitude is a product analytics platform founded in 2012, aimed at helping product and growth teams understand how users move through a digital product. It tracks events sent from web, mobile and server SDKs, resolves them to user identities, and exposes funnels, retention charts, pathfinder analyses and dashboards built on top.

The data model is event-and-property. Each event has a name and a set of typed properties, each user has a resolved identity across devices, and the platform charges based on monthly tracked users (MTUs) rather than raw event count. The tracking plan, Amplitude's schema for what events and properties are allowed, is the single most important artifact in any deployment that has to survive more than a quarter.

Type
Product analytics platform for event-based user behavior
First released
2012
License
Proprietary SaaS, MTU-based pricing
Maintained by
Amplitude, Inc. (NASDAQ: AMPL)
Typically used for
Funnels, retention, cohort analysis, feature usage, growth experiments

The honest version

Amplitude pros and cons

No technology is all upside. This is where Amplitude earns its place and where it tends to cause trouble.

Where Amplitude is strong

  • Funnels and retention are the core strengthThe interaction model for defining a funnel, a retention cohort or a path is purpose-built and quick to work with once the taxonomy is sane.
  • Cross-platform identity resolution worksA user on web, mobile and backend can be stitched into one timeline through the Amplitude Identity API without rolling your own user graph.
  • Tracking plans enforce schemaGovernance mode rejects events that are not in the plan, which stops the taxonomy decaying every time a new engineer ships a feature.
  • Experiment integration is usefulAmplitude Experiment plus the analytics side means feature flags, exposures and outcomes land in the same place, so win/loss analysis is one dashboard away.
  • SDKs for most mainstream platformsBrowser, iOS, Android, React Native, Flutter, Node, Python, Java, Go and more are all covered, with batching and retries built in.
  • Clean data export to a warehouseHourly exports to S3, BigQuery, Snowflake and Redshift mean the raw events can live in your warehouse for deep analysis without re-ingesting.

Where Amplitude bites

  • MTU pricing surprises growing productsEvery anonymous visitor is an MTU for the billing period. A sudden traffic spike from a marketing campaign can push the count past plan limits before anyone notices.
  • Taxonomy drift is the defaultWithout a locked tracking plan, every team adds their own event names and property types. Within a year, funnels built in different quarters are measuring different things.
  • Funnel numbers disagree with the warehouseSession windowing, re-attribution, and the way Amplitude counts users at each step mean the funnel can show a number the data warehouse cannot reproduce, which erodes trust.
  • Identity resolution has sharp edgesSetting user_id after some anonymous activity has already been tracked can merge or split identities in ways engineers do not expect, and the behavior is hard to reason about from the docs.
  • Historical data has retention limitsQuery time ranges are capped on export jobs and some chart types, which means very long-term analysis has to happen in the warehouse, not in Amplitude.
  • The admin UI gets heavyA mature instance with thousands of events, hundreds of properties and dozens of saved charts is slow to navigate, and renaming anything safely is a careful process.

Sound familiar?

Common Amplitude problems we fix

The Amplitude problems that bring people to us, and what you have once each one is fixed.

MTU count 142k of 100k

Billing plan overshot in the middle of the month

An anonymous traffic spike from a campaign, a bot net that was not filtered out, or a new client-side SDK on a high-traffic page. Each unique device counts whether or not it ever converted.

MTUs back inside the plan

unknown event type 'checkout-complete'

Engineers send events the plan does not allow

A new feature shipped with a fresh event name that was not added to the tracking plan, or a property type that does not match the schema. Governance mode drops the event silently.

Tracking plan matches the shipped code

funnel step drop-off > 97%

Funnels look impossibly bad

An event firing on only some clients, a property filter that excludes most users, or a session window shorter than the real user journey. The number is wrong, not the product.

Funnels match the ground truth

identify failed: user_id and device_id missing

Users split across devices and anonymous sessions

SDK initialization that races with identify, or an identify call fired before the user_id is actually known. The result is two users where there is one, or a merged user that eats the wrong history.

Identity resolved to one real user

property 'user_type' has 14k distinct values

Properties explode in cardinality

A property that was supposed to be an enum now carries user IDs, URLs, or raw strings. Charts slow down, exports balloon, and the property is useless as a filter.

Properties return to usable cardinality

schema violation: revenue expected number

Revenue analytics returns wrong totals

Revenue sent as a string from one client and a number from another, or in cents from the backend and dollars from the frontend. Charts quietly ignore the string version.

Revenue typed and totalled correctly

insert_id collision within window

Events duplicate or disappear in deduplication

Retried HTTP calls without stable insert_ids, or clients generating the same insert_id across sessions. Amplitude dedupes within a window and the user count looks wrong.

Dedup logic reflects reality

export job failed: 365-day limit

Historical exports to the warehouse fail

An export over more than a year of data, or an export that includes too many events in the window. The job dies partway and the warehouse has gaps.

Historical data landed in the warehouse

What we do

Amplitude help, from one broken funnel to the whole deployment

Point us at one unreliable funnel, or hand us the whole analytics stack.

Rescue
Funnels disagreeing with the database, MTU count over the plan, or a dashboard that nobody trusts. We fix the immediate pain and schedule the deeper work.
Taxonomy rebuilds
A lean tracking plan, governance mode on, deprecations handled properly, and property types locked so a careless shipment does not poison a funnel.
Identity resolution
Anonymous-to-identified transitions handled correctly across web, mobile and server, with existing mismatched identities reconciled where feasible.
Warehouse integration
Hourly or daily export to Snowflake, BigQuery or Redshift, with modeling in dbt or similar so Amplitude's numbers and the warehouse agree.
Migrations
Moving in from Mixpanel, Heap, PostHog, or Google Analytics, with event mapping and history preserved.
Managed Tech
Taxonomy policed, MTU trend watched, deprecations handled before they bite, and an engineer to call when a release weekend produces dashboards that make no sense.

How it works

From first message to fixed

  1. Look Free estimate

    Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.

  2. Fix

    We fix the cause, not the symptom. If something is down, getting it back up comes first.

  3. Harden

    Backups, monitoring, updates and access, so the same problem can't quietly come back.

  4. Hand back or stay on

    You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.

Often found together

Running Amplitude alongside something else?

We work on the whole stack, so the problem can be wherever it actually is.

Questions

Amplitude questions

Something else? Call+1 737 727 7327

Why do my Amplitude funnels disagree with my database?

Three common reasons. The funnel defines a time window that is shorter than the real user journey. The events at different steps fire on different client types and only some of them reach Amplitude. Or an identity merge has collapsed two users into one between the steps. We walk through each step with a cohort of known users and show where the gap comes from.

How do I control MTU spend on Amplitude?

The fastest wins are usually filtering bots and internal traffic out at the SDK or ingestion rule level, suppressing anonymous MTUs on low-value marketing pages where the conversion is tracked elsewhere, and reviewing which event streams are tracked identified vs anonymous. We audit the last three months of MTU trend and show which traffic is driving the bill.

Should I use Amplitude or Mixpanel for product analytics?

Both cover the same core job. Amplitude tends to be stronger on growth and retention features and has deeper experimentation integration. Mixpanel tends to be a little nimbler on ad hoc analysis and has aggressive pricing for smaller teams. The decision usually comes down to which team already uses which, and whether the experimentation integration matters for your roadmap.

How do I migrate from Google Analytics to Amplitude?

The event model is different enough that a direct mapping is rarely clean. We start by designing a tracking plan around the product actions that actually drive the business, instrument those in parallel with the existing GA setup, and run both for a quarter before retiring GA. Historical GA data stays queryable through its own export rather than being copied into Amplitude.

Can Amplitude replace our warehouse analytics?

Usually no, and we would not recommend trying. Amplitude is excellent for product questions in a product UI and is not meant to replace SQL over the full data model. The common pattern is Amplitude for product and growth teams, the warehouse for finance, operations and anyone asking a question Amplitude's chart types do not answer.

How much does Amplitude help cost?

Scoping is free of charge. Describe the deployment, the dashboard that is misbehaving, and the MTU plan you are on, and a scoped price lands with you before any taxonomy is touched.

Free estimate

Amplitude giving you trouble?

Tell us what you're running and what's going wrong. A person reads it and comes back with an estimate.

  • No charge to find out what it will take
  • If it's urgent, say so

Replies come from a person. By sending, you authorize us to email you back. No marketing.