You cannot learn your brandif you only ask it questions four times a year.

On creative testing at Bolkra. Last updated 21 May 2026.

WHERE CREATIVE DECISIONS GET MADE

Briefings and marketsare very different teachers.

Most fashion brands make their biggest creative decisions inside briefings. A creative director picks a direction. A photography team executes it. The imagery ships across the e-commerce site, the campaign, the paid social, the email. And then the brand waits eleven months to find out whether the choice was correct, by which point another briefing has already happened, another creative direction has already been chosen, and the cycle continues.

Inside those eleven months, the brand has run the experiment. The customers have spoken. The data is there, on the e-commerce platform, in the conversion rates, in the bounce rates, in the markdowns. But the brand cannot act on what it has learned, because acting on it would mean reshooting. Reshooting costs hundreds of thousands of pounds and takes weeks. So the brand keeps running the imagery it has, even when it knows the imagery could be better.

Bolkra changes the unit economics of that decision. A reshoot is no longer a quarterly capital event. It is an afternoon's work, when you need it, in the direction you have just decided would be worth testing. The page that follows is what that changes.

8.3 days

The median time a static image takes to fatigue on modern paid social platforms, before its cost-per-acquisition begins to rise. The traditional fashion shoot cadence is three to four times a year.

Source: aggregated performance data, 47 ad accounts spending £5,000 to £180,000 monthly, 2024.

15 to 30%

The win rate for creative hypotheses generated by intuition alone, across published A/B testing platform benchmarks. Seventy to eighty-five percent of intuition-led creative changes fail to produce a meaningful positive change.

Sources: Optimizely, VWO, ConversionXL aggregated benchmarks.

30 to 50%

The annual revenue growth advantage of organisations with mature creative testing programmes, compared to industry peers that rely on intuition-led decision making. The gap shows up at every traffic level.

Source: Harvard Business Review research on experimentation maturity.

THE FIVE VARIABLES

What actually moves commercial performance.

After a decade of running these tests at scale inside a major fashion e-commerce retailer, the same five variables emerged consistently as the ones that move commercial performance. They are not the only things that matter, but they are the ones worth testing systematically. The public research literature broadly supports this framework, with strong evidence on some variables and emerging evidence on others.

01 / MODEL

The most volatile variable.

Of the five variables, model selection has consistently moved performance more violently than any other within a single trading cycle. The published research supports this. Experimental studies on social-media imagery show statistically significant differences in purchase intent based on model demographics and body type, even when consumers rate competing images as equally attractive. The decision sits at the boundary of brand identity and commercial outcome, which makes it both the highest-leverage variable and the most contested one.

02 / STYLING

Fully styled outfits versus isolated garments.

Whether a product is shown styled with accessories and complementary pieces, or in isolation, changes how a customer imagines wearing it. Interactive mix-and-match styling tools have driven site-wide conversion lifts of more than 90 percent in published enterprise case studies. Beyond the conversion number, complete styling shifts purchase decisions from single items to ensembles, raising average order value at the same time.

03 / BACKGROUND

Studio white, lifestyle context, or somewhere in between.

The choice of background carries one of the better-evidenced commercial impacts in the public literature. Lifestyle and contextual backgrounds have driven conversion lifts in the 20 to 70 percent range in published case studies, with the upper end consistent with our own experience across thousands of tests. The Baymard Institute and recent academic work caution against over-complicating backgrounds, however. The variable is real and powerful. It also has limits.

04 / LOCALE

Where the largest commercial gaps tend to sit.

Most brands ship one set of imagery to every market they sell into. The published research suggests this is a serious commercial leak. Studies applying Hofstede's cultural dimensions to e-commerce imagery have documented conversion lifts above 30 percent from locale-specific imagery, with global brands able to translate this into seven-figure incremental annual revenue at retail scale. In our experience, this is the variable most consistently underused by brands that have the data to know better.

05 / CONTEXT

The right image at the right point in the journey.

The fifth variable is structural. Lifestyle imagery (in-context, environmental, aspirational) and utilitarian imagery (clean cutout, detail-focused, transactional) are both essential, but they serve different moments in the customer journey. Lifestyle leads discovery. Utilitarian closes the decision. Mismatching which image type sits in which gallery position is one of the most common and most easily corrected commercial errors in fashion e-commerce.

WHAT BECOMES POSSIBLE

Different cadence,different decisions.

Localising for a new market

Most brands enter a new market with the imagery they already have. They translate the copy. They adjust the prices. They leave the photography alone. The publicly available research suggests this is leaving meaningful conversion on the table. With Bolkra, the same product can be shown to French, German, US, and Korean customers in imagery calibrated to each market's visual register. Different model selection, different styling, different backgrounds. Each market sees imagery that was made for it, even on a product range that was originally shot for one.

Testing creative direction before committing

The most expensive creative decisions are the ones that get made before any data is collected. A creative director commits to a season's direction in March. The work ships in August. By October, the data is in. By then, the decision is sunk. Bolkra makes it possible to test creative direction the other way around. Generate variants in the proposed direction, test them against the current one, look at the data, then commit. The risk of a wrong direction does not go to zero, but it becomes a managed risk instead of a forecast.

Responding to a signal

Most actionable signals in fashion e-commerce are too small to justify a reshoot. A specific colourway is underperforming. A product is selling well in one market and slowly in another. A new collection is launching against an unexpected competitive backdrop. None of these are big enough to call a photographer. All of them are big enough to be worth a different image. Bolkra closes the gap. The signal arrives on a Monday, the new imagery is live by Wednesday, the data comes back the following Monday.

THE CONVERSATION

This framework is the starting point.Where it applies to your brand is a conversation.

The five variables are the structure. Whether all five matter for your brand, at this moment in your trading cycle, given your category and your customer, is a different question. So is the question of which parts of your creative direction should be tested and which parts should be protected from testing because they are part of what makes the brand itself.

We have spent a decade making these judgement calls inside a major fashion e-commerce retailer at scale. We will spend an hour with you and your team making them for your brand specifically. No software demo. A working conversation about where your biggest creative leaks are likely to sit and what a structured testing programme would do about them. If Bolkra is the right answer for that programme, we will say so. If it isn't, we will say that too.

Book a testing consultation →Read the public research →

WHAT THE RESEARCH DOES NOT YET TELL US

The honest data gaps.

We have presented the strongest publicly available research on each of the five variables. The research is good. It is not complete. Here are the gaps we acknowledge openly, both because we want the page to be honest and because some of them are areas where we expect to see real research catch up over the coming years.

Cross-cultural research outside US and UK.
Most public research on imagery and conversion is conducted in US or Northern European markets. The locale variable is strongly evidenced at the broad level, but the specific question of which imagery works in which market, beyond the major Western markets, is meaningfully understudied. The Hofstede-derived work and the multimodal AI behavioural studies have started to fill this gap. They have not finished.
Long-term brand-equity effects of high-velocity testing.
Most published research measures the short-term conversion impact of imagery changes. The longer-term effects, over years, on brand-equity measures, premium pricing power, and customer loyalty, are largely unmeasured outside the Binet and Field tradition. Practitioner consensus exists. Rigorous longitudinal data does not.
Generative AI imagery in commercial production.
The research on testing AI-generated imagery against traditional photography is too new to have produced large-sample longitudinal evidence. The early enterprise data is promising. The five-year picture is genuinely uncertain.
The interaction between mobile-first usage and the five variables.
Most published variable-by-variable research was conducted at a moment when desktop was the dominant context. Mobile is now the dominant context in fashion e-commerce. Whether the same variables move the same way on the same scale in a mobile-first world is partly known and partly assumed.
The dispersion of testing maturity across the industry.
Public surveys cover the headline question: most fashion retailers do not run mature creative testing programmes. The specific shape of that distribution is not well documented. We can describe what mature looks like. We have less rigorous data on how many brands sit at each level.

GET STARTED

Or start small.

If a consultation feels heavier than the conversation you want to have today, start on the free tier and generate a few testing variants for yourself. The five variables are visible in the studio.

Sources

Aggregated performance marketing data (2024). Creative fatigue study across 47 ad accounts, £5,000 to £180,000 monthly spend.
Optimizely, VWO, ConversionXL. Aggregated A/B testing platform benchmarks on creative hypothesis win rates.
Harvard Business Review. Research on experimentation maturity and revenue growth advantage.
Baymard Institute. Product page UX and imagery research for e-commerce.
Wang et al. (2024). Contrast-composition-distraction framework for e-commerce imagery.
Hofstede cultural dimensions applied to e-commerce imagery and conversion.
Binet and Field, IPA. Long-term brand effects versus short-term activation.
Eileen Fisher. Published enterprise case study on styling and conversion.
Label Emmaus. Published case study on contextual imagery.