Why is my Microsoft Azure Usage Details API propterties.cost sum fluctuating?

I wrote a python script that returns me the Usage Details for all my subscriptions and saves them in a dataframe. It later sums up the properties.cost and returns the total of the ongoing month. My goal ist to know how much i already spent this month. But with every new request i make, the total of my costs goes up or down.

Output example:

{'request1': {'costTotal': 96.99047619273139,
  'time': '17-03-2020T13:39:19:000Z'},
 'request2': {'costTotal': 105.28680062444838,
  'time': '17-03-2020T13:41:19:000Z'},
 'request3': {'costTotal': 103.50071786335579,
  'time': '17-03-2020T13:43:19:000Z'},
 'request4': {'costTotal': 103.04300834107798, 
'time': '17-03-2020T13:45:19:000Z'}}

The API im using:

https://management.azure.com/{scope}/providers/Microsoft.Consumption/usageDetails?api-version=2019-10-01

Part of my download.py script which concats the dataframes and saves it as pkl:

ENV = sys.argv[1]
ORG_FILES = None
RESULTING_FILES = ["billing_file.pkl"]

def main():

    execution = {"file": RESULTING_FILES, "orgfile": ORG_FILES}

    def generate_billinginfo(subscription, cred):

        scope = "subscriptions/" + subscription
        url = (
            "https://management.azure.com/"
            + scope
            + "/providers/Microsoft.Consumption/usageDetails?api-version=2019-10-01"
        )
        headers = {
            "Authorization": "Bearer " + cred.token["access_token"],
            "Content-Type": "application/json",
        }

        connection_attempts = 0
        while True:
            httpresponse = requests.get(url=url, headers=headers, verify=False)
            connection_attempts = connection_attempts + 1
            if httpresponse.status_code != 200:
                logger.error(
                    "Problem accessing server status code: %s",
                    httpresponse.status_code,
                )
                if connection_attempts >= 3:
                    raise ValueError(
                        "Can not access server. Status code: %s",
                        httpresponse.status_code,
                    )
                else:
                    time.sleep(3 * connection_attempts)
            else:
                break

        # Extracting tags & properties
        Tags = []
        Properties = []
        for resource in httpresponse.json()["value"]:
            Tags.append(resource["tags"])
            Properties.append(resource["properties"])

        dfgeneral = pd.DataFrame(httpresponse.json()["value"])

        dfprop = pd.DataFrame(Properties)

        # Rejoining dataframes
        billingdf = dfprop.join(dfgeneral[["tags"]])

        return billingdf

    cred = ServicePrincipalCredentials(
        client_id=config["client_id"],
        secret=config["password"],
        tenant=config["tenant_id"],
    )

    subscriptions = [
        "0f5dd24a-f6e0-473f-8643-75dc3481bdc1",
    ....
    ]

    for subscription in subscriptions:
        try:
            billinginfo
            billinginfoexists = True
        except:
            billinginfoexists = False

        if billinginfoexists:
            billinginfo = pd.concat(
                [billinginfo, generate_billinginfo(subscription, cred)]
            )
        else:
            billinginfo = generate_billinginfo(subscription, cred)

    billinginfo.to_pickle(RESULTING_FILES[0])

    # Resulting file
    with open(Path("./execution.json"), "w") as outfile:
        json.dump(execution, outfile, indent=4)

Any ideas what the problem could be?