Make Credits vs Operations: Trace Your Usage

If your Make scenario uses ordinary one-credit modules, its billable operations and credits usually match. If it uses built-in AI or other specially metered features, count the operations first, then apply each module’s credit rule. Before buying more capacity or switching platforms, check whether empty polling, repeated items or AI usage explains the difference. Make’s credit rules.

What changed from operations to credits?

Credits replaced operations as Make’s billing-unit name. Operations still describe the work performed: a module running to process data or check for new data. A scenario run can contain many operations, and an operation can consume a fixed or variable number of credits. The terminology change does not make every scenario run cost one credit. Credits, operations.

Keep three quantities separate: module operations, credits deducted by Make, and charges billed by connected providers. Non-AI app operations generally use one credit each. Built-in AI can depend on tokens, the model and connection type; some other features use pages, file size or processing time. Use the selected module’s documentation rather than applying a universal operation-to-credit conversion. Make credit accounting.

Trace polling, branches and a loop

Consider an illustrative support workflow. These are assumed volumes and inputs, not a tested template or measured benchmark:

• Over one 24-hour day, Gmail Watch Emails checks every 15 minutes: 96 checks, including empty checks. Across those checks, it retrieves exactly 100 distinct emails.
• A router sends 80 ordinary emails to Google Sheets Add a Row. The other 20 match an exclusive structured-request route.
• Each structured email contains valid JSON with an array of exactly five text requests. Parse JSON returns one bundle containing that array. An Iterator separates its five items.
• Each item goes through Make AI Toolkit Summarize Text using a custom provider connection, then Google Sheets Add a Row. Assume an account already entitled to that connection.
• Every action succeeds once. There are no retries, duplicate deliveries, additional retrieval steps or backfills.

Make Academy documents one credit for each Gmail polling check, even when it returns no email, and no credits for routers or filters. Ordinary downstream actions consume credits when they execute. The two routes below are deliberately mutually exclusive; do not assume every router configuration selects just one route. Academy credit examples, router conditions.

The Iterator’s 100 output bundles do not mean it ran 100 times: it processed 20 input arrays. The next two modules each process 100 individual items. Make Academy’s worked array example establishes the one-credit Iterator charge for an input array and the subsequent per-item action charges. Its Parse JSON example also uses one credit for a single input producing an array-containing bundle. Iterator accounting.

For the AI row, the exact rule is specific to Make AI Toolkit with a custom provider connection: one Make credit per operation, plus charges paid directly to that provider. It is not a universal AI rate. AI Toolkit connection and billing rules.

Repeating these assumptions for a 30-day month gives 12,480 Make credits and 3,000 external AI calls. That is a consumption estimate, not a subscription quote. If each structured email instead contains ten items, daily consumption becomes 616 credits: the Iterator count stays at 20, while AI and item-row actions each rise to 200.

Account for AI on the correct bill

For a separate token estimate, assume each of those 3,000 calls uses 800 input tokens and 120 output tokens. That gives 2.4 million input tokens and 360,000 output tokens per month. These are budgeting assumptions; replace them with the selected provider’s reported usage, including any separately billed token categories.

If that provider quotes input and output rates per million tokens, the corresponding model charge is 2.4 × input rate + 0.36 × output rate, before other applicable fees or taxes. The rates must come from the exact model and billing account. No dollar amount is implied here. Name the owner of the Make subscription and the owner of the provider account; they may be different people or organizations.

With Make’s provider selected in AI Toolkit, the module’s credit consumption varies with tokens and model. Keep the non-AI subtotal of 316 credits per day, then add the AI Toolkit credits reported for the 100 calls. Do not retain the custom-provider assumption of 100 AI credits or add a separate provider invoice for those same Make-provided calls. AI Toolkit billing.

Connection eligibility must be checked in the account before changing routes. The Toolkit documentation describes custom connections on paid plans. Public pricing displays can differ, so this example assumes existing access and makes no recommendation to buy a particular tier. Toolkit requirements, Make pricing and access.

Find the cause before buying more credits

Open a representative run’s details. Inspect each module’s operation bubble and its input and output bundles; enable the credit display so that operation counts and credit deductions can be compared. Find the first point where either the bundle count expands or credits diverge from operations. Make’s operation inspection guide.

Use that evidence to choose the next action:

• If polling dominates, consider a slower schedule only if the response-time requirement permits it. In this example, hourly checks would reduce 96 checks to 24 and save 72 credits per day, assuming all 100 emails are still retrieved and processed exactly once. That changes latency, not the required output.
• If repeated items dominate, check for accidental duplicates and unnecessary work inside the loop. Do not remove required per-item summaries merely to shrink the estimate.
• If AI dominates, compare the same task and acceptable output quality across connection and model choices. Include both Make and provider charges.
• If the counts are justified and the existing allowance covers them, stay with the current setup. Extra capacity does not repair an unintended loop or duplicate processing.

Before converting credits into money, confirm the organization’s selected allowance, checkout currency and region, taxes, monthly billing or annual commitment, and any extra-credit charges. Make supplies the hosted platform, but your team still owns scenario maintenance and failure handling; lower credit consumption alone does not prove lower operating cost. Make’s subscription and hosting information.

Use the Make pricing guide to compare the selected allowance and the Make overview to assess product fit. Your immediate next step is to build this ledger for one normal run and one busy run, then project the actual workload. Buy more capacity only when the required work explains the consumption; investigate the scenario first when it does not.