Stop Buying Followers: A Data-Backed Approach to Influencer Selection

Follower count predicts almost nothing

Across campaigns we have measured, the correlation between an influencer's follower count and revenue generated per dollar spent is close to zero. Sometimes mildly negative, because large accounts price on reach and reach is what you are overpaying for.

Four signals that do predict outcomes

1. Audience overlap with your existing buyers

The strongest signal available. If your CDP holds purchase data, you can measure how much an influencer's audience resembles people who already buy from you — not just demographics, but category affinity and price sensitivity.

2. Save and share rate, not likes

Likes are a low-cost signal of mild approval. Saves indicate purchase intent. Shares indicate advocacy. Both correlate far better with conversion than likes do.

3. Comment substance

Automated sentiment scoring on comment text separates "🔥🔥🔥" from "does this come in a larger size?". The second kind is the audience that buys.

4. Historical commercial performance

Not whether they have done brand deals, but whether those deals converted. Many creators will share this. The ones who will not, often have a reason.

A workable selection process

1. Build a longlist by category relevance
2. Score audience overlap against your first-party data
3. Filter on save/share rate against a category benchmark
4. Sample comment sentiment on the last 20 posts
5. Request historical performance on comparable campaigns
6. Test small across 5–8 creators before concentrating budget

The most common mistake is concentrating budget on one large creator before testing. Six small tests will teach you more than one large bet, and cost less.

Measure it properly

Last-click will systematically under-credit influencer work, because discovery and purchase are usually separated by days and channels.

Explore influencer marketing services.