Amplification model
Read this when deciding whether to put paid spend behind an organic post, how much, and when to stop.
The principle: paid does not resuscitate a weak post, it widens one that organic traffic already validated. Publish naturally, let early signal prove it works, then amplify only the winners. Boost the original post so it stays native rather than rebuilding it as an ad, which discards the social proof and the comment thread that made it work.
Layer 1 — Monitor
Track the first six hours: impression velocity, engagement rate, save and share ratio, comment sentiment.
Instruments report signal. They do not make the call. A dashboard that auto-boosts on a volume threshold will reliably spend behind posts that got attention from the wrong audience, because volume cannot distinguish between the two.
Layer 2 — Trigger
Start only when all three hold:
| Condition | Threshold |
|---|---|
| Engagement rate | ~1.5–2x the account's own trailing 30-day average |
| Timing | Within 6–8 hours of publish |
| Sentiment | Clearly positive; negative share below ~5% |
The baseline is that account's own recent performance, never a cross-industry benchmark. A strong post on a 12,000-follower account and a weak post on a 400,000-follower account produce similar absolute numbers and opposite conclusions.
Layer 3 — Deploy
Set spend caps at three levels so no single decision can consume the programme:
| Cap | Purpose |
|---|---|
| Per post | Stops one promising post absorbing the cycle |
| Per creator, per period | Stops a favourite creator crowding out the roster |
| Per category, per period | Stops one content type dominating the mix |
Hold fifteen to twenty percent of the budget back for a manual call on something the trigger missed. Triggers are tuned on past patterns and will not recognise a genuinely new one.
Test two to four posts simultaneously rather than one, over one to three days. Short tests, quick reads. A long single test buys certainty about one asset while the window closes on the others.
Layer 4 — Stop
Cut on any one of these, not on all four:
| Condition | Reading |
|---|---|
| Cost per action >30% over target for 48h | Not going to correct on its own |
| Frequency rising while click-through falls | Saturation; the audience has seen it |
| Organic engagement on the original below its baseline | Paid distribution is now reaching people who do not want it |
| Cluster of brand-safety-negative comments | Spend is buying an argument |
Write the stop conditions before the first unit of spend goes out. Written afterward, they get relaxed.
Worked example
A sponsored post goes live at 09:00 on an account whose trailing 30-day average engagement rate is 1.2 percent.
T+2h. Engagement rate 1.9 percent, roughly 1.6x baseline. Under the trigger this is promising but early, and the six-hour condition has not been met. Do not spend. Keep working the comments.
T+6h. Engagement rate 2.3 percent, now 1.9x baseline. Reshares are running ahead of the account's usual reshare-to-like ratio. Sentiment reads clearly positive with two critical comments, both substantive rather than hostile. Trigger conditions met.
Check the contract before spending. Paid usage rights confirmed, thirty-day window. Deploy at the per-post cap over a 48-hour test, alongside two other posts from the same cycle.
T+30h. Cost per action is 18 percent over target and frequency has reached 2.4 with click-through flat. Over target but not by the stop margin, and frequency is not yet coupled with declining clicks. Continue and re-read at T+48h.
T+54h. Cost per action now 34 percent over target for the second consecutive day. Stop condition met on a single criterion. Cut, regardless of the fact that the post is still the best performer in the cycle. It is the best performer organically; paid distribution has exhausted the audience that wanted it.
The judgment that matters in this example is at T+2h. Most teams spend there, because the number looks good and the excitement is highest. Spending at T+2h means spending before the signal has separated genuine spread from a burst of early enthusiasm from people who already follow the account.
What not to amplify
A post below the account's own median. A post carrying divided or negative sentiment, where spend buys a larger argument. A post with likes but no replies, which the ranker has already scored as low quality and paid distribution will not repair. A post whose creator agreement lacks paid usage rights, which is not a judgment call but a legal one.
Attribution, set up before publish
One distinct tracking parameter or discount code per creator, never shared across a roster, or the post-cycle analysis cannot separate them. Where the surface does not permit links, read branded-search lift over the window instead and accept the coarser signal. Request back-end screenshots from creators as a cross-check on self-reported reach; the gap between reported and actual is itself a data point on that creator.