How long should I cache social media API results to avoid paying twice?

Paying twice for the same data usually happens by accident: a dashboard that calls the API on every page load, a nightly job that refetches everything, a notebook re-run during analysis. A small cache in front of your calls fixes all three. The only real decision is how long each kind of data may sit before you fetch it again.

Lifetimes that fit social data

Data	Endpoint	Changes	Suggested cache
Account profile, followers	Douyin account profile	slowly	1 to 7 days
Post text and images	Xiaohongshu note detail	almost never	keep indefinitely
Post metrics	same call	fast at first, then slowly	hours in the first week, then days
Comments on posts older than a week	TikTok video comments	slowly	days; re-check only the first page
Hot lists	Weibo hot search	within hours	do not cache; store each snapshot
Search results	per query	depends on the keyword	your monitoring interval

The prices tell you where caching pays most. Xiaohongshu note detail is $0.02 per note, so 1,000 repeat lookups waste $20.00. A Douyin profile is $0.002; for cheap lookups like that, a cache also keeps big jobs clear of rate limits.

A small file cache

import hashlib, json, os, pathlib, time, requests

H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}
CACHE = pathlib.Path(".okcache"); CACHE.mkdir(exist_ok=True)

def cached(path, ttl_seconds, **params):
    key = hashlib.sha256(json.dumps([path, params], sort_keys=True).encode()).hexdigest()
    f = CACHE / key
    if f.exists() and time.time() - f.stat().st_mtime < ttl_seconds:
        return json.loads(f.read_text())
    body = requests.get(f"https://api.okdata.online{path}", headers=H, params=params, timeout=60).json()
    if body["code"] == "ok":
        f.write_text(json.dumps(body, ensure_ascii=False))
    return body

profile = cached("/v1/douyin/users/get", 3 * 86400, id="MS4wLjABAAAA...")

Only successful responses are cached, so an error never sticks; errors were free anyway.

Two things that are not your cache

• Idempotency keys. Resending the same request with the same Idempotency-Key within 24 hours replays the stored result for free. That protects retries; it is not a general cache across different requests.
• meta.cached. The service may answer from a short-lived cache of its own and says so in each response. You should not count on it for savings.

For a monitoring setup that only pays for what is new, see competitor monitoring.