Optimising Accuracy and Reducing Response Time

Use this guide to improve the accuracy and speed of receipt and invoice processing with Taggun's API.

Most performance improvements come from three areas:

submitting a clear, correctly sized document;
keeping files within the recommended size and format constraints;
disabling product line item extraction when you do not need product-level data.

Submit clean receipt images

Image quality has the biggest effect on extraction accuracy.

Ask users to submit images that:

show one receipt or invoice per request;
include the full receipt, including merchant name, date, totals, and line items if you need product data;
are in focus, well lit, and not blurred;
avoid glare, shadows, heavy perspective angle, and background text;
use the original digital receipt or invoice file where available instead of a screenshot or re-photographed copy.

For user-facing capture guidance, see End-User Guidance.

Keep files practical to process

Upload files must be under 20 MB. Smaller files usually upload faster and reduce end-to-end response time, provided the receipt text remains clear enough to read.

Supported formats include JPEG, PDF, PNG, GIF, and HEIC/HEIF. See Supported File Formats.

Disable optional extraction work you do not need

Product line item extraction can increase processing time. Set extractLineItems deliberately instead of relying on account defaults.

Set extractLineItems to false unless your workflow needs product-level data such as item names, quantities, unit prices, or total prices.

Example request field without product line items:

{
  "extractLineItems": false
}

Provide useful hints when you have them

Hints can improve accuracy and reduce avoidable ambiguity.

Field	When to use it
`language`	Set a two-letter language hint when you know the receipt language. Leave it empty for automatic detection.
`near`	Provide a city, state, and country when merchant location context is useful.
`ignoreMerchantName`	Use when a known customer or account name is often mistaken for the merchant name.

Do not add hints that you are unsure about. Incorrect hints can reduce accuracy.

Handle errors and retries safely

Use retries for transient failures, not for validation errors.

If you receive 429 Too Many Requests, retry with exponential backoff.
If a file is rejected, check the file type, file size, and image dimensions before resubmitting.

Quick checklist

Before sending receipts to production, confirm that your integration:

uploads one receipt or invoice per request;
keeps files under 20 MB;
sets extractLineItems: false when product-level data is not needed;
passes language, near, or ignoreMerchantName only when those hints are reliable.