A 25-line check for botted crypto KOL accounts before you pay them

Matt Cepulis — Tue, 20 Oct 2026 12:08:27 +0000

Paying a crypto KOL is mostly a bet on whether their audience is real. Follower counts are easy to fake. Engagement patterns are harder.

We vet KOLs for a living, so here is the first-pass check we run before anyone looks at a wallet or a rate card. It will not prove an account is clean. It tells you which accounts deserve a closer look.

function kolRedFlags(k) {
  const flags = [];
  const engagement = k.avgLikes + k.avgReplies + k.avgReposts;
  const er = engagement / k.followers;
  if (er < 0.002) flags.push('engagement under 0.2% of followers');
  if (k.avgReplies / Math.max(k.avgLikes, 1) < 0.01)
    flags.push('almost no replies for the likes');
  const growth = k.followers30dAgo
    ? (k.followers - k.followers30dAgo) / k.followers30dAgo : 0;
  if (growth > 0.3) flags.push('followers up more than 30% in 30 days');
  if (k.topRepliersShare > 0.4)
    flags.push('a handful of accounts write most of the replies');
  if (k.promoPostsShare > 0.5)
    flags.push('more than half of recent posts are paid promos');
  return { engagementRate: er, flags };
}

The thresholds are starting points, not rules. Tune them to the niche: a memecoin caller and a DeFi researcher have very different normal engagement.

What each flag catches

Bought followers inflate the denominator, so the engagement rate collapses. Bought likes rarely come with replies. Sudden follower jumps usually mean a purchase or a giveaway. Reply rings show up as the same few handles under every post. And an account that is mostly ads has an audience that has learned to scroll past them.

What it misses

Wallet behaviour. The most expensive KOL problem is not fake followers. It is a real audience being sold into by the KOL's own wallet. That needs on-chain checks, which we cover in our crypto KOL due diligence checklist and the open crypto KOL campaign guide.

If you are pricing streamers rather than X accounts, the same idea applies to live viewers. We put that maths in a free streamer sponsorship cost calculator.

Feedback on the thresholds is welcome in the comments.

The funnel maths behind pricing a streamer sponsorship (and the formula we put in a free calculator)

Matt Cepulis — Mon, 19 Oct 2026 16:49:23 +0000

Most streamer sponsorship quotes are priced on followers. Followers tell you how many people clicked a button once. They tell you nothing about how many people are watching right now.

We run creator campaigns for a living, so we needed a way to price a live deal on what actually happens during a stream. We ended up with a small set of formulas and turned them into a free streamer sponsorship cost calculator.

Here is the maths, so you can check it or build your own.

Inputs

Cost side:

number of streams
creator fee per stream
production and clip editing
usage rights and exclusivity
management, tracking and reporting
optional payment per first purchase (for hybrid deals)

Delivery side:

average concurrent viewers per stream
hours live per stream

Funnel side:

share of average viewers who click the link per stream
share of clicks that register
share of registrations that make a first purchase
your target cost per first purchase

The formulas

function priceDeal(d) {
  const fixed = d.production + d.rights + d.management;

  const clicks = d.avgViewers * d.clickRate * d.streams;
  const registrations = clicks * d.regRate;
  const firstPurchases = registrations * d.purchaseRate;

  const totalCost =
    d.streams * d.feePerStream + fixed + d.perPurchase * firstPurchases;

  const viewerHours = d.avgViewers * d.hoursLive * d.streams;

  return {
    totalCost,
    viewerHours,
    costPer1kViewerHours: totalCost / (viewerHours / 1000),
    costPerClick: totalCost / clicks,
    costPerRegistration: totalCost / registrations,
    costPerFirstPurchase: totalCost / firstPurchases,
    // the highest fee per stream that still hits your target
    maxFeePerStream:
      (d.target * firstPurchases - fixed - d.perPurchase * firstPurchases) /
      d.streams,
  };
}

The last line is the useful one. It turns the question around: instead of "is this quote fair?", you ask "what is the most I can pay per stream and still hit my number?"

A worked example (made-up numbers)

These inputs are invented to show the mechanics. They are not real rates.

4 streams at $1,500 each
$800 production, $400 rights, $600 management
1,200 average concurrent viewers, 4 hours live
5% of viewers click, 30% of clicks register, 25% of registrations make a first purchase
target: $150 per first purchase

Results:

total cost: $7,800
viewer-hours: 19,200, so $406.25 per 1,000 viewer-hours
240 clicks at $32.50 each
72 registrations at $108.33 each
18 first purchases at $433.33 each
max fee per stream for a $150 target: $225

So the quote is roughly 6.7 times what the funnel can carry. You either negotiate the fee down, improve the funnel (a better offer, a pinned link, a clip programme that keeps working after the stream), or walk away.

Why viewer-hours and not followers

Two creators with the same follower count can differ by ten times on concurrent viewers. Concurrent viewers times hours live is the closest thing to a real impression count on a live platform, and it is the number you can verify from public stream stats.

What the model leaves out

Clip reach after the stream. One two-hour stream can hold 20 to 30 usable short clips. If they are cut and posted, the cost per result keeps falling for weeks.
Brand lift and retention. The model only counts first purchases.
Variance. Real campaigns are lumpy. Treat the output as a negotiating range, not a forecast.

If you want the acquisition side in more depth, we wrote up what a casino creator costs per depositing player and how to size a first creator pilot.

Feedback on the funnel assumptions is welcome in the comments.

DEV Community: Matt Cepulis