Amplitude vs Heap

Autocapture versus a planned tracking contract

The central choice is how your team wants to learn. Heap records a wide interaction surface first. An analyst can later label a button click, navigation path or form interaction as a virtual event and inspect history that was already captured. That is useful when a product team has unanswered questions and cannot wait for a release cycle. It also creates a responsibility: an interaction that looks meaningful in one interface can become noisy after a redesign, and two analysts can label the same behavior differently unless the team documents a shared definition.

Amplitude starts with the opposite operating habit. Product and engineering agree that an event such as `workspace_created` or `report_exported` has a name, properties, owner and expected meaning. The team sends it deliberately and uses that catalog for funnels, retention and cohorts. Planned instrumentation takes longer at the start, but a metric can be traced to a release and reviewed when its meaning changes. For a B2B SaaS with several squads, that governance often matters more than the speed of the first dashboard.

A practical approach is to use Heap during discovery if the organization is still learning which interactions deserve investment, then decide whether stable revenue and activation metrics need a stricter contract. If Amplitude is the long-term system, set a migration rule early: exploratory clicks may inform a question, but a metric used in a board report or customer health workflow must be promoted to a named event with an owner. If Heap remains the system of record, apply the same rule to virtual events and keep a changelog for selectors, page areas and identity joins.

• Write the business question before choosing an event definition: activation, adoption, retention or expansion.
• Record the actor, account, workspace, plan and release properties needed to answer that question.
• Assign an owner who can retire a definition when the UI or product behavior changes.
• Keep exploratory analysis separate from the small set of metrics used in operating reviews.

Retroactive analysis, taxonomy and governance

Retroactive analysis is valuable because B2B SaaS teams often discover the important question after an incident or customer interview. Heap’s autocapture can show whether a past cohort used a feature or followed a path that was not in the original plan. That evidence can shorten discovery, but it does not remove the need for a taxonomy. A retroactive event is still a product definition: the team must state which elements count, which accounts are excluded, how identity is resolved and what happens when the page changes.

Amplitude’s planned approach makes those decisions visible before the data is used. An event catalog can distinguish a stable product action from an implementation detail. It also helps data teams reconcile product events with warehouse facts, CRM accounts and billing periods. The trade-off is that a missing event remains missing. Before selecting Amplitude, ask engineering whether the release process can include instrumentation review and whether product managers will maintain ownership after launch.

Governance should be tested in a live workflow rather than a slide deck. Create one definition, change its properties, deprecate it and ask a second team to find the current version. Check who can create events, who can edit a cohort, how permissions work, and whether an analyst can explain an old dashboard six months later. Heap’s capture breadth and Amplitude’s planned catalog solve different parts of this problem; either can become unreliable without a named steward.

Session replay and the Contentsquare relationship

Session replay answers a different question from a funnel. A funnel shows that users dropped between steps; a replay can reveal whether a form was confusing, a control failed, or a page was used in an unexpected order. Amplitude lists session replay as part of its product analytics plans and includes 10K monthly session replays on Free. Heap lists Session Replay as an add-on on Pro. Compare the practical package: monthly replay allowance, retention, masking of personal data, sampling, search, export and the teams that can access recordings.

Heap is part of the Contentsquare group. That ownership can be useful for a team looking for a broader digital experience stack, but it also makes product boundaries and commercial packaging worth checking. Ask whether the replay experience in your proposal is Heap’s current feature, a Contentsquare capability, or an add-on with separate administration. Confirm the support path, data residency options and how an account can request deletion. Do not infer any of those terms from the acquisition alone.

For a B2B SaaS, replay sampling should follow a privacy and research plan. Mask workspace names, customer content, tokens and support conversations. Use replay to investigate a defined product question, then turn the finding into an event, usability fix or support workflow. Replays are evidence for a decision; they are not a substitute for a governed metric or a retention definition.

Warehouse sync and identity design

A warehouse connection determines whether product analytics can be reconciled with contracts, support and revenue. Heap’s pricing record describes Heap Connect as syncing data to a warehouse automatically and retroactively. That is particularly relevant when the team wants to join newly defined events to an older captured history. Confirm the destination engines, refresh cadence, schema stability, event property types, user deletion handling and whether the connector is included in the plan or quoted separately.

With Amplitude, start by deciding which system owns the canonical event model. If the warehouse is the source of truth, keep event names, account IDs, workspace IDs and timestamps consistent across the SDK and warehouse pipelines. If Amplitude is the analysis workspace, document which cohorts or exports are sent downstream and how their membership is reproduced. In either case, use a stable account key alongside a user key: a person can change workspaces, and a company can have many users with different product roles.

Run a small reconciliation before buying a larger tier. Pick a known workspace, compare signup, activation and expansion events in the product analytics tool with warehouse rows, and deliberately exercise a user merge and deletion request. The result should be a short data contract that names the owner of identity mapping, retention and schema changes. This work often exposes a more consequential cost than the headline analytics price.

A worked scenario for N monthly tracked users

Assume a B2B SaaS team tracks N monthly users across customer workspaces. Each user generates an average of E meaningful product events and a wider set of browser interactions. The team wants to answer three questions: which onboarding steps predict a first successful workflow, which accounts return after a trial, and why users abandon a report builder. N and E are variables because the correct commercial comparison depends on actual traffic and event design, not on a generic seat count.

The team first estimates its Amplitude volume as N × E events per month, then adds a documented allowance for system events and QA traffic. If the result stays within the listed 2M events/month Free allowance, the Free plan is a $0 starting point. If it grows beyond that allowance, the team requests an event based quote and asks how the first 2M events, retention, session replay and experiments are represented. It also budgets the engineering work to instrument `onboarding_step_completed`, `first_report_published` and `workspace_invited` with account and plan properties.

For Heap, the team estimates monthly sessions rather than events. A session ends after 30 minutes of pageview inactivity on web, so N monthly tracked users can produce more or fewer than N sessions depending on repeat visits and usage patterns. The Free plan is $0 up to 10K monthly sessions with 6 months of history. If the team exceeds that allowance or needs Pro replay and warehouse capabilities, it requests a custom session pricing quote and asks what happens above the plan limit. It should not convert Amplitude’s event estimate into a Heap price without a vendor quote.

During a six-week evaluation, Heap answers the discovery question quickly: define a report-builder abandonment event from captured interactions and inspect prior sessions. Amplitude answers the operating question: instrument the stable workflow, create a governed activation cohort and reconcile it to account expansion in the warehouse. The team chooses Heap if recovering past behavior is the urgent constraint and it can staff curation. It chooses Amplitude if the product organization needs a durable event contract across squads. If both are viable, compare the written quote plus implementation and governance hours for the same N, session volume, retention and replay requirements.

1. Measure N monthly tracked users, monthly sessions, meaningful events and replay needs for the last complete month.
2. Write the three decisions the analytics program must support and identify the account and workspace keys behind each one.
3. Run one discovery question through Heap autocapture and one governed workflow through Amplitude instrumentation.
4. Reconcile a sample to the warehouse, including identity changes, deleted users and a product release.
5. Request like-for-like quotes and record the internal engineering, analytics and governance time beside the vendor price.

Migration and rollout checklist

Moving from Heap to Amplitude is a shift from captured evidence to an explicit event contract. Inventory the virtual events and keep only definitions that answer a current business question. Map each one to an Amplitude event and properties, instrument it in a release, then run both tools in parallel long enough to compare counts and identity. Expect gaps where Heap’s broad capture included interactions that were never promoted to a stable product event.

Moving from Amplitude to Heap can shorten the time to investigate an unknown behavior, but do not delete the existing catalog on day one. Keep the Amplitude names as reporting requirements, label their Heap equivalents, and test whether selectors, URL changes and single-page application navigation affect capture. Rebuild cohorts with account context and confirm that historical definitions are still valid before changing an operating dashboard.

1. Export a definition inventory with owner, purpose, properties, account key and last review date.
2. Choose a bounded product area and run dual collection for a release cycle.
3. Compare event counts, session boundaries, identity joins and replay privacy behavior.
4. Document the retirement date for old dashboards and keep a link to the replacement definition.

Questions to ask in a demo and proposal

• Can the team show a new event definition, its owner, approval path, history and retirement workflow?
• For Heap, how are virtual events protected from selector changes and how are noisy autocaptured interactions curated?
• For Amplitude, what engineering and QA workflow is expected for a new event, property or account identity?
• What session replay allowance, retention, masking, sampling and add-on terms apply to our traffic?
• What exactly does warehouse sync include: history, refresh timing, schemas, deletion propagation and destination support?
• How are N monthly tracked users translated into sessions, events, limits and the written commercial proposal?