Unable to marshal response: Object of type Decimal is not JSON serializable

What it means

A Lambda handler does not return a value to a caller directly. It returns it to the runtime, which serialises it to JSON and hands the bytes back to the invoker. That serialisation step is invisible until it fails, and when it does the failure lands after every piece of work your function did.

This is what makes the error feel out of order. Your logs show the request arriving, the database being read, the result being assembled, and a final line saying the work is complete. Then the invocation fails. Nothing your code did went wrong — the value it produced simply could not be expressed in JSON, and the result is discarded along with everything it took to compute.

Almost every Python instance traces back to decimal.Decimal. Boto3's DynamoDB resource deserialises every numeric attribute as Decimal rather than int or float, deliberately, because floats cannot represent decimal fractions exactly and silently corrupting a monetary value would be worse than an inconvenient type. The consequence is that any item returned straight from DynamoDB is full of objects json.dumps refuses, often nested inside lists and maps where they are invisible in a quick read of the code.

The reason it survives local testing so reliably is that print() and logging call str(), which every one of these types implements perfectly. A Decimal prints as 4. A datetime prints as a readable timestamp. So a developer logs the object, sees exactly what they expected, returns it, and the runtime rejects it — the same value passing one representation and failing another. Being able to read the successful log lines and the marshal failure as one invocation is what makes that sequence obvious rather than baffling.

What it looks like in CloudWatch

This is the shape the failure arrives in: the lines of one invocation scattered among everything else the log group received at the same moment.

The panel on the right is generated by running the excerpt on the left through the same parser that powers the free web stitcher — it is what the tool actually produces for this input, not an illustration of it.

How to confirm it from the logs

Position is the giveaway. This appears after all of your own logging, including whatever you logged just before returning, because it happens when the runtime marshals the value your handler already produced. If your logs show the work completing and then a failure, this is the shape.

The type named in the message tells you what to fix, and Decimal is by far the most common because that is what boto3 returns for every DynamoDB number. datetime is second. Both are ordinary Python objects with no JSON equivalent, and neither is anything you wrote.

Causes, most likely first

Fixes

pythonimport json
from decimal import Decimal
from datetime import date, datetime
from uuid import UUID

def encode(value):
    if isinstance(value, Decimal):
        # Whole numbers should not become 4.0 in the response.
        return int(value) if value % 1 == 0 else float(value)
    if isinstance(value, (datetime, date)):
        return value.isoformat()
    if isinstance(value, (set, frozenset)):
        return sorted(value)
    if isinstance(value, UUID):
        return str(value)
    if isinstance(value, bytes):
        return value.decode()
    raise TypeError(f"Cannot serialise {type(value).__name__}")

def lambda_handler(event, context):
    item = table.get_item(Key={"id": event["id"]})["Item"]
    return {
        "statusCode": 200,
        "body": json.dumps(item, default=encode),
    }

pythonfrom decimal import Decimal

def normalise(value):
    if isinstance(value, list):
        return [normalise(item) for item in value]
    if isinstance(value, dict):
        return {key: normalise(item) for key, item in value.items()}
    if isinstance(value, Decimal):
        return int(value) if value % 1 == 0 else float(value)
    return value

item = normalise(table.get_item(Key=key)["Item"])

Also seen as

The same underlying failure, worded differently by a different runtime, SDK version, or logging layer. All of these land here — there is no separate page for each phrasing.

Related errors

Errors that show up alongside this one, or that people mistake for it.

References

• AWS Lambda Developer Guide — Define Lambda function handler in Python
• Boto3 documentation — DynamoDB conditions and types