How Fast Do Your Videos Peak? Measuring View Velocity With Data

Total view count is a vanity number. A video with 100,000 views that took six months to get there behaves nothing like a video that hit 100,000 in 48 hours, but if you only look at the final total, they're identical. What actually matters for spotting winners, timing paid promotion, and understanding the algorithm's early verdict is velocity: how fast views accumulate, and when they stop.

You can't get this from a single API call, no endpoint tells you "views per hour." You build it by sampling the same videos repeatedly over time and computing the deltas yourself. Here's how.

Why velocity beats totals

Two reasons velocity is the more useful metric:

• It's an early signal. A video's first-day velocity is a strong tell for whether it'll be a winner, long before the total is impressive. Catch that early and you can pour promotion on a video the algorithm is already rewarding.
• It reveals the shape. Some videos spike and die; some slow-burn for weeks. The curve shape tells you whether your content has a long tail worth re-promoting or a flash worth riding immediately.

The catch is that public APIs give you a snapshot, the view count right now. Velocity is the derivative, so you have to build the time series yourself by snapshotting repeatedly.

The mechanic: snapshot, store, diff

The pattern is the same as any monitoring build: pull the current view count on a schedule, store each reading with a timestamp, and compute the change between readings. Base URL https://api.sociavault.com/v1, x-api-key header, 1 credit per call, data under data.

For TikTok, a creator's videos come back under data.aweme_list[] with statistics.play_count and a create_time (Unix seconds):

import os, time, requests

API_KEY = os.environ["SOCIAVAULT_API_KEY"]
BASE = "https://api.sociavault.com/v1"

def snapshot_tiktok(handle, amount=12):
    r = requests.get(f"{BASE}/scrape/tiktok/videos",
                     headers={"x-api-key": API_KEY},
                     params={"handle": handle, "amount": amount}, timeout=60)
    r.raise_for_status()
    data = r.json().get("data", {})
    items = data.get("aweme_list", []) if isinstance(data, dict) else []
    now = int(time.time())
    return [{
        "video_id": v.get("aweme_id"),
        "plays": (v.get("statistics") or {}).get("play_count"),
        "created": v.get("create_time"),
        "captured_at": now,
    } for v in items]

Run this on a schedule (say every 6 hours for the first few days after a post, then daily). Append each snapshot to a table, don't overwrite. The stored history is the whole point.

Computing velocity from snapshots

Once you have two or more snapshots per video, velocity is just the change in plays over the change in time:

def velocity(snap_a, snap_b):
    """views per hour between two snapshots of the same video"""
    dv = (snap_b["plays"] or 0) - (snap_a["plays"] or 0)
    dt_hours = (snap_b["captured_at"] - snap_a["captured_at"]) / 3600
    return dv / dt_hours if dt_hours > 0 else 0

Now you can do the useful things: rank fresh videos by first-24-hour velocity to spot early winners, detect when velocity drops toward zero (the video has peaked, stop promoting), and compare a new post's early curve against your historical winners to predict where it'll land.

Turning it into a practice

A workable routine: snapshot new videos frequently for their first 72 hours (that's where the signal is), taper to daily, and stop once velocity flattens. Store everything, then chart plays-over-time per video. For a set of recent posts that's a modest daily credit spend, and it uses the same infrastructure as a general monitoring dashboard. Keep the credit-saving habits in mind, velocity tracking is easy to over-poll.

The honest limits

• You're reconstructing, not reading, velocity. No endpoint returns views-per-hour; it only exists if you snapshot over time. Your resolution is only as fine as your sampling interval.
• Sampling interval caps accuracy. Snapshot daily and you can't see an intraday spike. Match your cadence to the precision you actually need, and remember finer cadence costs more credits.
• Public totals only. You're tracking public play counts, not retention, watch time, or traffic sources, the things that explain why a video moves. Those are owner-only.
• Counts can update unevenly. Platforms don't refresh public view counts in perfect real time, so very short intervals can show noise. Smooth across a couple of readings.
• Backfill is impossible. You can only measure velocity going forward from when you start snapshotting. Start now if you care about it.

Frequently Asked Questions

What is view velocity?

It's how fast a video accumulates views over time, views per hour or per day, rather than its total. Velocity reveals whether a video is spiking, slow-burning, or already peaked, which total view count alone completely hides.

Can I get views-per-hour from the API directly?

No. Public endpoints return the current view count as a snapshot. To measure velocity you snapshot the same video repeatedly over time and compute the change between readings, the API gives you the points, you build the curve.

How often should I snapshot?

Frequently in a video's first 72 hours (say every few hours), where most of the signal lives, then taper to daily and stop once velocity flattens. Match the interval to the precision you need, since finer sampling costs more credits.

Why does velocity matter more than total views?

Because it's an early signal and it reveals shape. First-day velocity predicts winners before the total looks impressive, and the curve tells you whether to ride a video now or re-promote a slow-burner later.

Can I measure velocity for past videos?

Only going forward. Velocity requires time-series snapshots, so you can't reconstruct it for a period before you started sampling. Begin snapshotting now if early-signal detection matters to you.

What can't this tell me?

Why a video moved. You get public play counts, not retention, watch time, or traffic source, the owner-only metrics that explain the velocity. Use velocity to spot what's working, then investigate the why separately.

---

Want to catch your winners on day one instead of week six? Start free with 50 credits, no card required and start snapshotting your videos today.