Branded AR filters and masks as an engagement tool
How to design mechanics, not imagery
The difference between a filter that accumulates millions of views and a filter that quietly dies in the effects library is not in the budget for 3D-modeling. It is in whether the user has a personal motive to click 'record' and 'share'. Pepsi achieved 400 million reach and 11 thousand user-generated videos in a week not because the filter was technologically complex, but because it provided a simple, instantly understandable result of action — people wanted to show what they had achieved.
KENZO Sport Challenge (300 million views) and BOSS on TikTok (4,1 billion views) are built on the same principle: pose-tracking or hand-tracking — the most 'cost-effective' from a technical standpoint, but it is precisely these that give the user's body the role of the main character in the video. The more complex the filter is physically, the fewer people are able to replicate the effect without five attempts — and without a first successful attempt, the trend will not happen.
- The result should be visible within 1-3 seconds of interaction, not after a long setup
- The mechanics should work in a duet or as a response to someone else's video — that's what turns viewing into content
- There should be an element that you want to show specifically to a friend — a competitive score, an unexpected transformation, 'guess the result'
- The challenge works better as a single mask: LOL Surprise in Likee collected 70 million views against a plan of 20 million, precisely due to the combination of 3D-mask + challenge + giveaway + blogger support
Platform limitations: what is actually possible and what will be killed during moderation
The landscape has changed sharply in January 2025: Instagram has closed Spark AR, and creating custom filters for Meta is now physically impossible. This is not a technical detail — it means that the entire category of cases involving 'filter in Instagram stories' is no longer reproducible, and brands that built strategy around Meta are now forced to redistribute their budgets.
| Platform | Status for brands | Key limitations |
|---|---|---|
| Instagram / Meta | Spark AR closed as of January 2025 | Creating new branded filters is impossible; old restrictions — 4 MB, prohibition on real photos of people, strict rules for logos, text only as part of mask element |
| TikTok | Priority platform for brands | Effect House — a free tool for organic effects; Branded Effects — a paid format with premium placement and priority display |
| Snapchat | Actively developing, strong AR ecosystem | Performance requirements for lens; emphasis on engagement metrics and share rate as criteria for approval of top placements |
Practical conclusion: whereas before a typical funnel for launching a filter included simultaneous release on Instagram and TikTok, now it makes sense to design the mechanism directly for Effect House with a focus on organic growth, and connect paid Branded Effects as an amplifier, not as the main channel.
Which metrics actually speak to engagement, and which are just pretty numbers
Views are a metric of awareness, not engagement. They answer the question 'how many people saw it,' but not 'how many people felt or did something about it.' Real depth of engagement is measured by other indicators:
- Average time of interaction with effect — with Bollé, the first campaign delivered 13,88 seconds on average, which is notably deeper than standard interaction with banner advertising
- Share rate — the portion of users who didn't just try, but shared the result; the Pringles lens in Snapchat delivered 5 times more shares relative to the benchmark, and the Faces campaign reached 1,54% share rate — that is 5 times higher than the average for the AR market in the region
- Retention/save rate — Pringles also showed 3-fold growth in lens retention, meaning people returned to it repeatedly
- The amount of user-generated content (UGC) — for Pepsi it is 11 thousand videos per week, for LOL Surprise — 85 thousand participants in the challenge against the plan of 20 thousand
If only 'reach' and 'views' are mentioned in the AR campaign report, this is a reason to ask the agency — most likely, the depth of engagement was simply not calculated there.
AR filters and sales: truth and inflated expectations
Here it is worth being honest: an AR filter by itself rarely closes a sale directly, but there are measurable cases of direct influence on the bottom of the funnel. The Bollé campaign is one of the few documented examples: the second wave of the campaign led to a 85% increase in new website users and a 456% increase in sales. This happened not because the filter 'sold' glasses by itself, but because the mechanism was built into the chain — from trying on virtual glasses to navigating to the site where the cart was already waiting.
Mechanics without such an embedded next step (just a mask, just an effect) in most cases remain metrics of the top of the funnel — reach and awareness, as with Pepsi or KENZO. This is not a minus if the campaign's goal is brand awareness, but presenting 400 million views as proof of sales growth — is a methodological error that is often made in reports.
Embedding the filter into the visual identity without violating the platform's rules
The main practical contradiction is that a brand guide requires a specific logo and corporate colors, while the platform prohibits direct placement of the logo as a separate graphic element (which was the rule in Instagram until recently) or limits the file size to 4 MB. The solution that actually works is to transfer identity not into static elements, but into the behavior of the effect: corporate palette as the glow color during transformation, recognizable shape as the silhouette of the result rather than an overlaid logo, characteristic sound or animation as part of the brand code.
The text with the brand name should be embedded as part of the visual mask element — the inscription on the 'medal' that the user receives, rather than as a separate text layer. This approach simultaneously passes moderation and works on brand recognizability even without an explicit logo — people recognize the Pepsi or KENZO filter not by the text, but by the character of the interaction.
What ultimately determines the success of a filter
Technology in AR filters is long outdated. Success is determined by three factors simultaneously: simple mechanics that one wants to repeat and show to others; the right choice of platform for current technical capabilities (currently TikTok Effect House, not Meta); and a fair metrics system that distinguishes viral reach from real impact on the funnel. Campaigns without at least one of these elements deliver a pretty number in the report and almost nothing in business results.
How to design mechanics, not imagery
The difference between a filter that accumulates millions of views and a filter that quietly dies in the effects library is not in the budget for 3D-modeling. It is in whether the user has a personal motive to click 'record' and 'share'. Pepsi achieved 400 million reach and 11 thousand user-generated videos in a week not because the filter was technologically complex, but because it provided a simple, instantly understandable result of action — people wanted to show what they had achieved.
KENZO Sport Challenge (300 million views) and BOSS on TikTok (4,1 billion views) are built on the same principle: pose-tracking or hand-tracking — the most 'cost-effective' from a technical standpoint, but it is precisely these that give the user's body the role of the main character in the video. The more complex the filter is physically, the fewer people are able to replicate the effect without five attempts — and without a first successful attempt, the trend will not happen.
- The result should be visible within 1-3 seconds of interaction, not after a long setup
- The mechanics should work in a duet or as a response to someone else's video — that's what turns viewing into content
- There should be an element that you want to show specifically to a friend — a competitive score, an unexpected transformation, 'guess the result'
- The challenge works better as a single mask: LOL Surprise in Likee collected 70 million views against a plan of 20 million, precisely due to the combination of 3D-mask + challenge + giveaway + blogger support
Platform limitations: what is actually possible and what will be killed during moderation
The landscape has changed sharply in January 2025: Instagram has closed Spark AR, and creating custom filters for Meta is now physically impossible. This is not a technical detail — it means that the entire category of cases involving 'filter in Instagram stories' is no longer reproducible, and brands that built strategy around Meta are now forced to redistribute their budgets.
| Platform | Status for brands | Key limitations |
|---|---|---|
| Instagram / Meta | Spark AR closed as of January 2025 | Creating new branded filters is impossible; old restrictions — 4 MB, prohibition on real photos of people, strict rules for logos, text only as part of mask element |
| TikTok | Priority platform for brands | Effect House — a free tool for organic effects; Branded Effects — a paid format with premium placement and priority display |
| Snapchat | Actively developing, strong AR ecosystem | Performance requirements for lens; emphasis on engagement metrics and share rate as criteria for approval of top placements |
Practical conclusion: whereas before a typical funnel for launching a filter included simultaneous release on Instagram and TikTok, now it makes sense to design the mechanism directly for Effect House with a focus on organic growth, and connect paid Branded Effects as an amplifier, not as the main channel.
Which metrics actually speak to engagement, and which are just pretty numbers
Views are a metric of awareness, not engagement. They answer the question 'how many people saw it,' but not 'how many people felt or did something about it.' Real depth of engagement is measured by other indicators:
- Average time of interaction with effect — with Bollé, the first campaign delivered 13,88 seconds on average, which is notably deeper than standard interaction with banner advertising
- Share rate — the portion of users who didn't just try, but shared the result; the Pringles lens in Snapchat delivered 5 times more shares relative to the benchmark, and the Faces campaign reached 1,54% share rate — that is 5 times higher than the average for the AR market in the region
- Retention/save rate — Pringles also showed 3-fold growth in lens retention, meaning people returned to it repeatedly
- The amount of user-generated content (UGC) — for Pepsi it is 11 thousand videos per week, for LOL Surprise — 85 thousand participants in the challenge against the plan of 20 thousand
If only 'reach' and 'views' are mentioned in the AR campaign report, this is a reason to ask the agency — most likely, the depth of engagement was simply not calculated there.
AR filters and sales: truth and inflated expectations
Here it is worth being honest: an AR filter by itself rarely closes a sale directly, but there are measurable cases of direct influence on the bottom of the funnel. The Bollé campaign is one of the few documented examples: the second wave of the campaign led to a 85% increase in new website users and a 456% increase in sales. This happened not because the filter 'sold' glasses by itself, but because the mechanism was built into the chain — from trying on virtual glasses to navigating to the site where the cart was already waiting.
Mechanics without such an embedded next step (just a mask, just an effect) in most cases remain metrics of the top of the funnel — reach and awareness, as with Pepsi or KENZO. This is not a minus if the campaign's goal is brand awareness, but presenting 400 million views as proof of sales growth — is a methodological error that is often made in reports.
Embedding the filter into the visual identity without violating the platform's rules
The main practical contradiction is that a brand guide requires a specific logo and corporate colors, while the platform prohibits direct placement of the logo as a separate graphic element (which was the rule in Instagram until recently) or limits the file size to 4 MB. The solution that actually works is to transfer identity not into static elements, but into the behavior of the effect: corporate palette as the glow color during transformation, recognizable shape as the silhouette of the result rather than an overlaid logo, characteristic sound or animation as part of the brand code.
The text with the brand name should be embedded as part of the visual mask element — the inscription on the 'medal' that the user receives, rather than as a separate text layer. This approach simultaneously passes moderation and works on brand recognizability even without an explicit logo — people recognize the Pepsi or KENZO filter not by the text, but by the character of the interaction.
What ultimately determines the success of a filter
Technology in AR filters is long outdated. Success is determined by three factors simultaneously: simple mechanics that one wants to repeat and show to others; the right choice of platform for current technical capabilities (currently TikTok Effect House, not Meta); and a fair metrics system that distinguishes viral reach from real impact on the funnel. Campaigns without at least one of these elements deliver a pretty number in the report and almost nothing in business results.
Recent articles
MegaFon, BMW, VTB and more 240+ companies. See how it looks in the showcase reel:
Our branding agency is approached to order:
Branding agency services
