One video pulled 107,968 views and sold nothing. Another pulled 8,280 views and generated Rp 5,133,151.
Same product. Same platform. Same month. The video that reached thirteen times fewer people produced all of the revenue.
We found that pattern while auditing affiliate performance across nine FMCG product lines in Indonesia: pest control, home care, haircare, baby care, automotive, and home fragrance. The dataset covered 5,871 affiliate videos. We pulled the ten highest-GMV and ten lowest-GMV videos from each line, then ran frame-by-frame teardowns on the extremes.
The headline number: 3,718 of those 5,871 videos generated zero rupiah. That is 63.3% of everything the affiliate network produced.
The Zero-Revenue Majority
Most affiliate content does not underperform. It produces nothing at all.
| Product line | Total videos | Videos with zero GMV | Failure rate |
|---|---|---|---|
| Baby wipes | 296 | 256 | 86.5% |
| Reed diffuser | 1,043 | 858 | 82.3% |
| Roach spray | 326 | 247 | 75.8% |
| Car air freshener | 733 | 540 | 73.7% |
| Shampoo | 1,546 | 1,072 | 69.3% |
| Home air device | 643 | 416 | 64.7% |
| Drain cleaner | 184 | 111 | 60.3% |
| Pest adhesive | 276 | 81 | 29.3% |
| Plug-in mosquito repellent | 824 | 137 | 16.6% |
The spread matters more than the average. The mosquito repellent line failed on 16.6% of its videos. Baby wipes failed on 86.5%. Same seller, same affiliate pool, same platform mechanics, five times the failure rate.
Category difficulty explains some of that. Baby care carries a trust burden that pest control does not. A parent buying wipes for a newborn needs more reassurance than someone buying glue traps for a rat problem. But category difficulty does not explain why one shampoo video earns Rp 620 per view while another shampoo video earns nothing across 108,000 views.
Where View Count Stops Working
Across all 90 top-performing videos, views and GMV correlate at r = 0.73. That is a strong relationship. More reach does generally mean more revenue, and a model that ignored views entirely would be wrong.
The problem is the residual. Rupiah per view across those same top performers ranged from Rp 0.14 to Rp 2,684, a spread of roughly 18,800x inside the winners' bracket alone.
| Product | Views | GMV | Rp per view |
|---|---|---|---|
| Shampoo | 8,280 | Rp 5,133,151 | Rp 620 |
| Roach spray | 10,980 | Rp 3,765,342 | Rp 343 |
| Pest adhesive | 208,438 | Rp 11,297,474 | Rp 54 |
| Mosquito repellent | 4,254,968 | Rp 119,104,499 | Rp 28 |
| Reed diffuser | 1,330,998 | Rp 12,813,000 | Rp 10 |
| Shampoo (failed) | 107,968 | Rp 0 | Rp 0 |
An affiliate manager optimising for reach would have promoted the 108,000-view shampoo video and ignored the 8,280-view one. At the extremes, the reach signal points in the wrong direction.
Boosting budget follows view counts because view counts arrive first and feel like progress. Revenue arrives later, and by then most teams have attributed it to the platform algorithm rather than to the creative decision that caused it.
What the Teardowns Found
We transcribed and frame-extracted the extremes, highest earners against highest-traffic zero earners, to isolate what separated them. Four patterns stood out. Not every one held in every category, so we name the teardown each one came from.
Losing videos open with the problem. Winning videos open with the product.
Every failed video in the mosquito repellent teardown opened on a burning mosquito coil: smoke, ash, discomfort. Textbook Problem-Agitate-Solution. Every winner opened on the device or its pack, with the benefit stated inside the first three seconds.
PAS works in long-form copy where the reader has committed attention. In three-second commerce it costs you category clarity at the exact moment the viewer is deciding whether this video is about something they might buy. The problem framing is relatable. It is also a delay.
Winning videos do not prove the product works.
This one surprised us. None of the top-performing mosquito repellent videos showed a dead mosquito, a collection tray, or an overnight result. Not one. The highest earner in the entire dataset, Rp 119 million from a single video, showed hands plugging in a device, inserting a refill, and pressing a mode switch.
Those videos were not proving efficacy. They were removing purchase uncertainty. What does it cost, what is in the box, how do I install it, how long does the refill last, where do I tap to buy.
The losing videos answered "does this work?" quite well. The winning videos answered "should I buy this right now?" Only the second question moves GMV.
Longer did not mean better.
In the pest adhesive teardown, top performers averaged 31 seconds and zero-GMV videos averaged 37. The longer videos were not adding proof. They were repeating claims.
The second-highest earner in the dataset, a mosquito repellent video, ran 14 seconds. It had time for four beats: price surprise, bundle reveal, mode explanation, cart location.
Failing videos still told viewers where to buy.
This is the finding that should worry anyone running affiliate programs. All three zero-GMV videos in the mosquito repellent teardown pointed viewers to the cart, and two of the three had a clear cart CTA. One named the exact spot: "Keranjang kuning kiri bawah." Lower left. Tap here. The instruction was present.
What they lacked was a reason to act now. No exact price. No bundle math. No verified promo window. The top performers all attached something concrete: a specific kit price, how long a refill lasts, a stated promo condition. Cart location is necessary but it is not sufficient, and almost all of the revenue sits in that gap.
The Pattern Underneath
The failed videos were not bad videos. They were informative videos.
They explained the product clearly, demonstrated the setup, named the benefits, and pointed at the cart. A viewer finishing one of those videos understood the product. A viewer finishing a top performer felt ready to buy it.
That gap between understandable and buyable is what the teardowns kept pointing at, more than production quality, creator follower count, or view volume. It also costs nothing to fix, because reordering the beats in a script is free.
The Structure That Worked
From those teardowns we distilled one sequence for a 20 to 28 second runtime. It is a recommended structure, not an average we measured across all nine lines.
| Time | Beat | Job |
|---|---|---|
| 0:00–0:02 | Product visible, one specific claim | Establish category and reason to keep watching |
| 0:02–0:06 | One differentiator | Answer "why this one." Not five benefits. One. |
| 0:06–0:14 | Proof tied to use case | Show it working where the buyer will use it |
| 0:14–0:20 | Objection removal | Price, duration, safety, installation difficulty |
| 0:20–0:28 | Offer plus cart location | Exact price or bundle, then where to tap |
One detail is easy to miss. In the pest adhesive and mosquito repellent teardowns, the CTA landed between 70% and 85% of total runtime in every winner, never earlier.
What Sellers Should Change First
If you are running an affiliate program with a 60%-plus zero-GMV rate, recruitment is not your problem. Adding more creators to a broken brief multiplies the zero-GMV pool.
Rewrite the brief around the first two seconds. Specify that the product must be visible in frame one and named within three seconds. Most affiliate briefs specify hashtags, mention requirements, and disclosure language, then leave the opening shot to the creator. That is the first instruction we would add.
Supply the offer, not just the product. Failing videos said "check my cart." Winning videos said "[price] for the device plus refill, promo running now, yellow cart lower left." Creators cannot invent a bundle price. If the brief does not hand them a concrete offer, the CTA defaults to generic.
Stop scoring affiliates on views. The dataset contains creators whose highest-view video earned nothing and whose lowest-view video earned millions. Rank your affiliate pool by reach and you will promote the wrong people. Rank by GMV per video and the tier list reorders.
For how volume interacts with these conversion mechanics, see the volume math behind TikTok affiliate scale.
The Number That Matters
63.3% of 5,871 videos produced nothing. That is not a creator quality problem and it is not a platform problem.
The failures we tore down failed on sequence. Product too late, proof aimed at the wrong question, offer left unspecified. Each of those is a briefing decision made before a creator opens a camera.
The video that made Rp 119 million ran 24 seconds and showed a pair of hands plugging in a device. The one that made nothing across 108,000 views was a 14-second, creator-led clip that named the product late. Length was not the constraint. Order was.