What I Learned on My (Multiple) Calls with TikTok

Hi there,

Last week, I spoke with an account manager from TikTok. We were going over all things attribution, halo effect, demand creation, and TikTok Shop.

She said, flat out, that proving the halo effect to brands is hard on their end too. They don't have a great way to help brands measure it for themselves yet.

I've been non-stop talking about the halo effect because it is a real challenge for the brands we work with. And the platform most associated with the halo effect is still working out how to prove it's real. But I think TikTok is where this phenomenon is the most fascinating, and where there is the most opportunity.

What TikTok never sees

For the past year, most of the conversation about TikTok's halo effect has focused on brands not getting credit for what TikTok drives elsewhere. They can't see who bought, or what put that person in the aisle in the first place. Once a purchase moves into a store or a marketplace, the identity and the influence both get stripped out in the same motion.

TikTok's blind spot runs deeper. Unless a shopper checks out on TikTok Shop, it doesn't know a sale happened at all.

Adweek reported earlier this year that TikTok built a pixel-based tool, called Off-site Performance Analysis, to connect organic posts, TikTok Shop, Live videos, and paid ads to purchases made on a brand's own website [1]. Per Adweek, TikTok had treated its commerce ecosystem "like a fortress" for years: what happened in the app stayed in the app. That's a real admission. TikTok's own team knew they were flying blind on nearly everything that happened after someone left the platform.

But look closely at what that pixel actually covers: a brand's own website.

So that really only reflects ~20% of an omnichannel brand's business. The other 80% of sales made on Amazon, Target, Ulta, etc. don't connect back to TikTok at all.

The lift shows up after the algorithm has already moved on

Another piece of the attribution problem is timing. Haus's incrementality research found that TikTok's ad impact is a slow burn. Lift often stays flat for the first half of a campaign, then accelerates sharply, and keeps building for up to a week after the ads stop running [2]. In one geo-based experiment, brands saw an additional 68% lift to their primary KPI in that post-treatment window. In another, almost all of the impact for a higher-priced DTC product landed in the final three days of a month-long test, timed to a promotion [2].

Sit with that for a second. Most campaigns get evaluated, optimized, and reallocated on windows shorter than the time it actually takes TikTok to convert someone. The algorithm is adjusting spend before the real effect has even shown up. And per that same eMarketer coverage, only 33% of CPG marketers measure incrementality beyond a basic level, which means most of us are grading TikTok on a curve that was never built to capture what it's actually doing [2].

These two blind spots compound.

Smart+ is only as smart as what gets fed back to it

Along with under-reporting your results (making CAC look bloated and ROAS look weak), another aspect of the halo effect is changing how I think about 2026 planning: the algorithm running your campaigns is under-learning from your actual customers. Every purchase it never sees is a data point it never gets to train on.

On that same call, someone from our team walked through the mechanism plainly: pass a verified purchase back to TikTok's Events API, and TikTok can match it to someone who saw an ad in the days before. Feed enough of those back, and the algorithm starts learning who your real buyers actually are, not just who clicked.

TikTok's own Smart+ system shows how much this matters. Smart+ is built to "leverage AI and high-quality signals to automate campaigns," and in closed beta testing, advertisers using Smart+ Catalog Ads saw a 29% improvement in CPA over manual campaigns [3].

"High-quality signals" still mostly means pixel fires and in-app checkouts. If your TikTok Shop (and maybe DTC site) are the only sources feeding that system, and those channels represent a fraction of where your customers actually buy, Smart+ is optimizing beautifully against a small, skewed slice of your business. The Target run, the Amazon cart, the Walmart pickup -- the ~80% percent of purchases that happen nowhere near a TikTok checkout -- never touch the algorithm at all.

The platforms building these systems have as much riding on closing this gap as you do, and the credit fight over attribution matters just as much as what you're actually feeding the algorithm to learn from.

If you're learning anything new about TikTok Shop or the halo effect, I'd love to keep the conversation going.

Talk soon,

Kait

A few extras:

Some of my favorite brands are hiring…

1. Lowe's Media Network - Director, Product Management

Sources:

[1] Adweek, "TikTok Turns to Pixels to Show Advertisers What Happens Beyond Its Walled Garden."

[2] eMarketer, "TikTok's True Ad Impact Is a Slow Burn, Incrementality Research Shows."

[3] TikTok for Business, "How to Build Campaigns Your Way With Smart+."