The short answer
- Zapier measures successful actions as tasks in Classic Zaps; trigger checks, filters and paths do not use tasks. Make bills credits and describes an operation as a module run. Standard non-AI app operations generally use one credit each, including a polling check.
- For a deliberately simple 1,000-submission example, a successful search, one write and one confirmation per submission model 3,000 Zapier Classic tasks. A Make webhook run with those three modules models about 4,000 credits if the trigger and each action use one credit. Connector behavior, retries and additional modules can change either total. These are units, not prices.
Start with the same handoff
Assume a month with 1,000 form submissions. Each is read once, searched in the CRM, written once through either the create or update branch, and followed by one confirmation in another app. Assume all three actions succeed, each search produces a usable result even when there is no match, and each processed submission produces one record. There are no retries, replay, extra transformations or AI steps.
This is a model of a workflow, not a verified configuration. A connector may return no bundle on a missed search, combine operations into an upsert, send multiple records or require extra steps. Check the exact form and CRM, fields, required permissions, plan, connector versions and behavior before using the arithmetic.
Count tasks and credits
Zapier’s Classic task guide says successful actions count as one task each; triggers, polling, filters and paths do not. Under the stated assumptions, 1,000 searches plus 1,000 writes plus 1,000 confirmations are 3,000 tasks. The guide says a search configured to proceed when nothing is found counts as a task; a search configured not to proceed after no match does not. Confirm that the chosen CRM search can proceed to the create branch.
Make describes each polling trigger check as one operation, regardless of the number of bundles returned, and downstream modules as running per bundle. Its credit guide says non-AI app operations use one credit. For the assumed search, write and confirmation modules, the model is 3,000 credits plus the number of polling checks, including empty checks. With 100 checks returning the 1,000 submissions in total, that is 3,100 credits; with 1,000 checks, it is 4,000. These are illustrations, not a minimum or maximum.
For a different trigger, assume one webhook event starts one Make scenario run and supplies one bundle, with one credit for that trigger and one for each of the three standard non-AI modules. The illustration is four credits per submission, or about 4,000 for 1,000 submissions. This assumes default rates and no extra modules. Confirm the actual trigger and module charges in Make’s usage history; exceptions, batching and retries change the total.
The exact search must pass a no-match case to the create path; do not assume every Make search module emits a bundle for no results. Confirm its output, maximum batch size and any additional tools needed. A different trigger or batching pattern changes the trigger count. Router modules do not use credits according to Make’s pricing FAQ, but additional transformation and other modules may. Make’s If-else and Merge guide says those modules use operations without consuming credits. Check the billing of the actual modules used.
Neither total is a price quote. Choose plans that support the connector and branches, then price the selected task or credit allowance, billing term, add-ons and overage. Extra successful steps, replays and processing behavior change usage. Zapier says its Classic rules are the basis shown here and announces different Next Gen Zap billing from December 1, 2026; recheck the terms that apply when buying.
Test the actual connector
Zapier’s HubSpot guide lists a Create or Update Contact action based on email. Make’s credential-request guide names HubSpot Search for Contacts and Create or Update a Contact modules. These are examples of documented connector operations, not proof that every form, CRM object, custom field or plan supports this exact model.
An upsert may replace the separate search and branch, changing the usage estimate and the matching rule. Zapier documents email matching for its named HubSpot action; the Make material read does not establish that same key. Test a new contact, an existing contact, a missing email, two possible matches and the custom fields that sales needs. Confirm what a form platform sends, the destination record ID, and whether a confirmation means the CRM accepted the write.
Plan for a partial success
A CRM write can succeed even if the confirmation fails or the automation reports a timeout. Zapier warns that replaying a timed-out action can create duplicates. Zapier's full-run replay guide describes rerunning successful steps; Make says its replay runs the current scenario with previous trigger data through every module, including those that succeeded.
Make says storage of incomplete executions is disabled by default. Its error guidance describes using a Retry handler with storage enabled, then reviewing an unresolved execution. Confirm settings, retention, plan eligibility and the scope of the replay; do not assume all external writes roll back.
Use a stable submission identifier and a CRM match key, and check whether the receiving service supports idempotency or uniqueness enforcement. A search followed by create can race with another run, so the search alone does not guarantee a single record. Reconcile source submissions, destination IDs and confirmations; test a timeout after a successful write using synthetic data.
The example assumes exactly one create or update path. Make says a Router evaluates routes in order, while If-else runs the first matching route. Configure mutually exclusive conditions and test them; a branching diagram alone does not guarantee a single write.
Assign ownership and access
Name an owner for each connection and the form, CRM and notification destination. Grant only the permissions needed for the selected actions, define how access is transferred when someone leaves, and record who can edit, replay or disable the workflow. Do not share personal credentials to hand off a client integration.
A connected app, CRM edition or API quota may impose its own access and rate limits. Ask each vendor to demonstrate the exact trigger, object, fields and replay permissions under the intended plan. Verify the notification destination and a way to detect silent failures.
Choose a common trial
Shortlist Zapier and Make using the same form, CRM fields, volume and recovery cases. Consider n8n when its deployment and execution model fit the team; self-hosting also assigns operations and security work to the operator.
n8n’s pricing page describes an execution as a whole workflow run regardless of the number of steps. On that basis, 1,000 completed workflow runs would be 1,000 executions before any separately counted retries or sub-workflows. Ask how those cases are metered and add self-hosted infrastructure or third-party costs where applicable.
Use the form builder guide, CRM guide and Unbounce profile to map the systems on either side. The software trial guide helps assign an owner and pass/fail criteria.
Common questions
Are a Zapier task and a Make credit the same?
No. Zapier counts successful Classic Zap actions as tasks, while Make bills credits based on operations and, for some features, other consumption. Compare the exact workflow.
How many units do 1,000 submissions use?
Under the specified three-action example, the model is 3,000 Classic Zap tasks. At one webhook trigger plus three modules, each charged one credit, it is about 4,000 Make credits. Extra steps, retries, batching, connector output and billing rules can change the result.
Does replay prevent duplicate CRM records?
No. A timeout can leave a successful write behind, and replay may rerun it. Confirm destination-side uniqueness or idempotency and reconcile the records.
Sources and scope
We read the linked vendor documentation and pricing pages on November 5, 2026. This is documentation-only research; we have not built or tested these workflows, connectors or products.
Inclusion and corrections are free; sponsorship does not determine our conclusions. Quillwell also builds Adswell, a paid-media product.
Usage and plan terms change. Verify the specific connector, access and quote before relying on this illustration.
Sponsorship does not buy an editorial ranking or a positive review. How we research.
Source references
- Zapier: How task usage is measured
Checked 2026-11-05. Classic Zap task accounting, searches, polling and announced Next Gen billing change.
- Make: Credits
Checked 2026-11-05. Credits, operations and fixed versus dynamic usage.
- Make: Operations
Checked 2026-11-05. Trigger checks, bundles and downstream operations.
- Make: Pricing and credit FAQ
Checked 2026-11-05. Credits, router exceptions and depleted-credit behavior.
- Zapier: Replay an entire Zap run
Checked 2026-11-05. Full-run replay and previously successful actions.
- Zapier: Zap is creating duplicate data
Checked 2026-11-05. Replay following timeouts can create duplicates.
- Make: Scenario run replay
Checked 2026-11-05. Replay uses current scenario and reprocesses earlier successful modules.
- Make: Incomplete executions
Checked 2026-11-05. Incomplete execution storage is disabled by default.
- Make: Fix errors and warnings
Checked 2026-11-05. Retry handler and incomplete execution storage.
- Zapier: Getting started with HubSpot
Checked 2026-11-05. HubSpot connection and create-or-update action.
- Make: Credential requests
Checked 2026-11-05. HubSpot module examples and eligible credential-request users.
- Zapier: Pricing
Checked 2026-11-05. Plan features and task billing.
- n8n: Pricing
Checked 2026-11-05. Whole-workflow execution billing on listed plans.
- Make: If-else and Merge
Checked 2026-11-05. If-else and Merge use operations without credits; route-selection behavior.