Most brands run TikTok and assume it lifts Amazon — but never test it, because the two platforms don't talk. We measured it directly on our own authorised data: does TikTok activity lift Amazon search demand for the same product, against a stable baseline, once paid ads, seasonality and promotions are screened out. It does — for some products, not all. Our own brand; the numbers are sealed.
What you're seeing. Amazon search demand holds inside its normal week-to-week noise band (the dashed lines) — until a burst of TikTok activity, after which it climbs clear of the band on a 0–5 week lag and decays back over the following month. The delay is the tell: a shared cause would move both on the same day. Axis values sealed.
Two things make this a halo and not a coincidence. The lift arrives on a 0–5 week lag, not the same day — which is what you'd see if some third thing drove both. And it survives screening out paid ads, seasonality and promotions — the boring explanations that usually account for a bump. Strip those away and a real, TikTok-shaped lift in Amazon search is still standing. For collagen it's large; for creatine it's cleanest; for one product it isn't there at all.
The lift lands on a 0–5 week lag and survives screening out ads, seasonality and promotions — that's what separates a halo from a coincidence.
TikTok and Amazon are walled gardens. You can see TikTok views on one screen and Amazon sales on another, but nothing joins them — so "is our TikTok working for Amazon?" gets answered by gut feel and a hopeful glance at the sales chart. We answered it with a bridge that needs no cooperation from either platform: does a burst of TikTok activity move Amazon search demand for the same product — the leading indicator that shows up before the organic sales do?
We score each product as signal versus its own baseline — how many standard deviations search demand jumps — not as a "views went up and so did sales" story, which proves nothing. A lift only counts if it clears the product's normal week-to-week noise and lands on the lag a halo would produce. That's the whole method in one line, and it's why the number survives a sceptic.
A halo isn't a brand-wide multiplier. Collagen halos hardest, creatine cleanest — and one product not at all.
A halo is not a brand-wide multiplier you can assume onto everything you post. It's per-product, and the shape differs by what the content is doing — building awareness for a considered buy, or nudging an existing habit:
| Product | Peak lift | Shape of the halo | Verdict |
|---|---|---|---|
| Marine Collagen | +▓–▓ SD | Large, broad, 0–5 wk — an awareness lift on a considered purchase | Strongest |
| Creatine | +▓ SD | Smaller but tight — least noise, clearest attribution | Cleanest signal |
| Inositol | +▓ SD? | Overlaps a seasonal rise we couldn't fully separate | Contested — hold |
| Everything else | — | No lift clears the baseline on the halo lag | No halo — don't assume one |
It tells you which products to put behind TikTok and which to stop crediting to it. Collagen earns the awareness spend; creatine is the one to trust the measurement on because its signal is cleanest; inositol stays out of the business case until we can separate it from the season. A brand-wide "TikTok lifts Amazon" claim would have quietly funded the products that don't halo at all.
Before crediting a single sale to TikTok, we killed the boring explanations — and the lift outlived every one.
The reason most "TikTok drove our Amazon sales" claims don't survive contact is that something more boring usually explains the bump. So before we credited a single sale to the halo, we killed the obvious co-drivers in the same window — and the lift was still there:
| The boring explanation | What we checked | Verdict |
|---|---|---|
| "It was just paid ads." | Sponsored-Products spend and impressions across the lift window | Flat — no ad surge to explain it |
| "It's the season." | The product's own prior-year seasonal curve | No matching peak (except inositol) |
| "It was a promo." | Deals, coupons and price cuts in the window | None coincided with the lift |
| "Search just wobbles." | The lift against the 8-month noise band | Clears the band by a wide margin |
"Views went up, sales went up" is a story — it can't tell a halo from a coincidence. A lift that lands on the right lag and outlives a co-driver screen is a measurement. It's slower and less flattering, and it's the only version worth putting a budget behind.
In pounds, not standard deviations — and as a range with an honest floor you can actually plan on.
A halo in standard deviations doesn't pay the bills, so we translate it into incremental Amazon profit — and we do it as a range with an honest floor, not a single flattering figure. The floor gates the lift hard (only demand we can defend as TikTok-shaped); the upper is the generous read. The truth sits between, and you should plan on the floor:
Two disciplines keep this honest, and they're the ones most agency "TikTok ROI" decks skip. The time basis is labelled — this is per month at current posting cadence, not a once-ever windfall annualised into a big number. And the floor is dual-gated: a sale only counts if it clears the baseline and the co-driver screen. Under-claiming on purpose is what makes the number safe to build a decision on.
Note the minus sign. On the contested product the honest range includes a loss — TikTok spend that may not be earning its keep on Amazon. Leaving that in is the point; a halo study that only ever finds upside isn't measuring anything.
It turns TikTok from a channel you fund on faith into one you can steer.
The study turns TikTok from a channel you fund on faith into one you can steer:
1. Point TikTok at the products that halo. Weight content and spend toward collagen — the awareness lift is real and largest — and stop crediting TikTok for the products with no measured lift.
2. Time it against the Amazon season. The lift takes 0–5 weeks to land, so a burst timed weeks ahead of a product's Amazon peak arrives exactly when demand is climbing anyway, instead of firing into a flat week.
3. Keep the receipts. Re-run the same signal-vs-baseline read each campaign so the business case updates itself — and so the day the halo fades, you see it, rather than funding a ghost.
Everything so far is search demand — the leading indicator. Here's what tightens it further.
Everything above is measured on search demand — the leading indicator — because it moves first and needs no personal data. The same connection tightens the picture as you feed it more of your own authorised data:
Trace the halo from search demand through to the organic (non-ad) units and profit that followed it, per product — turning the floor estimate from "defensible" to "reconciled to sales."
Resolve the lift down to posting bursts and formats, so the next brief backs the content that actually moved Amazon — not the one with the most views.
TikTok lifts open-web search too. Run the identical method against Google demand and Shopify sessions to see the full off-Amazon halo, not just the marketplace slice.
Keep the baseline live so a product that stops responding to TikTok flags itself — before you've spent another quarter funding a channel it's outgrown.
Same method throughout: your own authorised data, measured against its own baseline, with the boring explanations screened out first. The halo most brands pay for blind, finally on a meter.
The same measurement, on your authorised data — in your own cloud, yours to keep.