A leading formalwear expert has shared his advice on how retailers can better utilise AI to improve suit sizing.
Mike Dobell, founder of online formalwear retailer Dobell, says fit remains one of the biggest problems in online fashion.
That view is backed by research with fit cited as a reason for returning an online purchase by 57.6 per cent of consumers in a survey of 1,000 UK shoppers by IMRG and nShift.
Separate research from Ingrid found that size and fit issues accounted for 52.9 per cent of UK returns across fashion, sportswear and accessories.
The scale of the problem is also reflected in the wider returns market. ZigZag’s UK Annual Returns Benchmark 2025 forecasts £25.1 billion in UK non-food returns in 2025, with 23.6 per cent of clothing purchases expected to be sent back. Its research names product fit as the dominant reason for returning clothing and footwear, with two-fifths of all returns coming down to sizing issues.
Mike believes AI could help reduce that uncertainty, but tailoring presents a more complicated sizing problem.
“Suits make that worse than most categories, because a jacket, trousers and a shirt all fit differently on the same body. Someone can be spot-on for a 40 chest and still need the trousers taken in, and a size chart doesn’t tell you that.”
Here, Mike outlines three areas retailers should get right if they want AI sizing tools to make useful suit recommendations.
One: Start with consistent product data
“A 40 Regular jacket in one style can fit differently from a 40 Regular in another, and that’s a problem retailers create for themselves long before AI enters the picture,” Mike said.
“Feed that inconsistency into a sizing tool and it comes out the other end looking authoritative. Getting your own measurements consistent across the range is the boring bit nobody wants to do before the software goes live.”
For suit retailers, the value of a recommendation depends heavily on the quality of the garment measurements behind it. A retailer may have detailed information about a customer’s chest, waist and height, but the recommendation will still be unreliable if the measurements of the products themselves are inconsistent.
AI cannot remove inconsistencies in underlying product data simply by presenting a confident answer. Retailers should ensure their systems record more than a standard label such as “40 Regular”, including the garment’s actual measurements, cut, fabric characteristics and intended fit.
Two: Treat a suit as more than one size
Mike also argues that retailers should not treat a single suggested size as the finished job.
“An algorithm will give you a number and sound very sure about it,” he said.
A customer is effectively fitting several garments at once. The jacket may fit correctly across the chest and shoulders while the trousers need a different waist or leg measurement. The shirt may introduce another set of considerations around the neck, sleeve length and preferred room through the body.
The useful role for AI is therefore to help narrow the best combination, rather than reduce the entire fit decision to one label. A stronger system could explain why it has recommended a particular size, identify where the fit is most likely to be close and flag when an alternative cut or professional alteration may be more appropriate.
That distinction is important because fit is both measurable and personal. One customer may prefer a close contemporary silhouette, while another may want extra room for movement or to wear a waistcoat underneath.
For help choosing your size, consult Dobell’s size guide or contact the team for advice before ordering.
Three: think beyond standard sizes
“Most of what I’ve seen performs reasonably in the middle of the size range and gets shakier at either end, which happens to be exactly where a lot of formalwear customers sit, including plenty of ours,” Mike said.
“Testing a tool against standard sizes and assuming it holds up for big-and-tall customers as well is a guess dressed up as due diligence.”
Any AI sizing tool should be tested across the full-size range, different body proportions and different garment cuts. Retailers should also assess whether recommendations work for customers with preferences that do not match a standard fit profile.
That testing should continue after launch. Retailers can analyse whether recommended items are kept, exchanged or returned, and record the reasons customers give for those decisions. Over time, the data may reveal that a particular cut is regularly returned as tight through the thigh, or that customers between two jacket sizes tend to prefer the larger option.
This type of feedback can improve both the technology and the products being recommended.
Mike believes the opportunity is real, but says retailers risk undermining the technology if they treat it as a shortcut around the fundamentals of fit.
“Fashion has already spent years improving size charts and product information, and fit is still one of the biggest reasons for returns,” he said. “A rushed version of this just adds another layer of false confidence on top of a problem that’s already expensive. It’s worth taking the time to get right.”
About Dobell
Dobell is a British-designed menswear brand specialising in accessible formalwear for weddings, black-tie events, work and other special occasions. Founded in 2003, the company offers suits, tuxedos, shirts, footwear and accessories across an extensive range of sizes.
Shop Dobell for men’s formalwear in a wide choice of sizes, styles and fits.
