Beyond traditional flour specifications: overview
- Protein, moisture and Falling Number remain essential, yet flour functionality increasingly depends on starch behaviour, enzyme activity and dough rheology
- Two flours with similar laboratory results can perform very differently on the production line because of differences in their functional properties
- Manufacturers are increasingly looking beyond standard quality measures to understand how flour will behave during mixing, proofing and baking
A flour delivery arrives at the bakery with a certificate of analysis showing it meets every agreed specification: protein content is within tolerance; moisture is on target; the Falling Number – which measures the activity of starch-degrading enzymes that influence dough behaviour – also sits comfortably inside specification.
Yet once production begins, dough absorbs more water than expected, mixing times drift, proofing becomes less predictable and finished products no longer deliver the consistency customers expect.
That scenario is becoming increasingly familiar across industrial baking. Manufacturers are asking production lines to run faster than ever, often producing multiple products on the same equipment while trying to reduce waste, cut energy use and maintain consistent quality.
At the same time, climate pressures are making wheat crops more variable, meaning two flours with remarkably similar laboratory specifications can behave very differently once they reach the mixer.
Recent research suggests flour performance depends on far more than protein content alone, with starch functionality, dough rheology and the interaction between flour components helping determine how dough behaves during production.
The commercial consequences are already being felt across the wider grain supply chain.
After France’s rain-hit 2024 wheat harvest – the country’s smallest soft wheat crop in around 40 years – starch producers including Tereos, Roquette, ADM and Cargill lowered their usual grain acceptance standards to accommodate unusually small kernels containing less starch and more cellulose. The poorer-quality wheat increased the risk of filter blockages, machine wear, production slowdowns and higher processing costs, illustrating how changes in grain composition can have consequences long before flour reaches the bakery.
However, according to Arnaud Dubat, global business development director at KPM Analytics, the problem isn’t that flour specifications are wrong. It’s that many no longer capture all the variables that determine performance on today’s high-speed production lines.
When the specification doesn’t match the production line
“This is definitely the key mistake that most bakers have been making for decades and are still making in modern bakeries today,” says Dubat.
“They bet on ‘the one’ characteristic that ‘tells’ everything they need to have a successful bake. Still a lot believe that protein content is the most important, some think gluten (which is basically the same parameter). This mindset comes from a long legacy in baking tradition, but it is not working anymore.”
That doesn’t mean protein has become less important. It remains fundamental to gluten development and dough structure. What has changed is scientists’ understanding of how it behaves.
A review published in May found that flour functionality depends on the interaction between protein, starch and dough rheology, rather than on any single measurement in isolation, helping explain why similar specifications can produce very different baking performance.
Dubat argues the answer doesn’t sit inside a universal specification because every production environment behaves differently. “The truth lies in the production line,” he notes. “The answer to the question is not only product-specific, but also plant-specific. What can be a key parameter for a baker at one plant or product type might not be so critical for another one.”
In other words, a specification that works perfectly for one bakery may tell another bakery surprisingly little about how the same flour will behave.
That helps explain why flour that satisfies every contractual specification can still produce inconsistent results. Existing certificates of analysis provide valuable information, but they offer only part of the picture.
“Because of what we explain above, flour specifications today provide only a partial view of flour performance. By repeating the same process over and over but expecting different results, bakers today still receive flour that fully meet existing specifications from their supplier, but the flour is not processing as expected, creating production issues, product loss, and lots of frustration.”
Recent research reinforces that observation. Scientists studying individual flour streams from the same roller milling process found substantial differences in damaged starch, water absorption, Falling Number and dough rheology between streams produced from the same wheat. Those variations translated into measurable differences in processing behaviour and finished product quality, suggesting traditional flour specifications cannot always predict how flour will perform in production.
Why starch deserves a bigger role in the conversation
One reason is that traditional specifications tend to focus on composition, while production performance depends just as heavily on functionality. Two flours can contain similar levels of protein yet behave very differently because their starch, enzyme activity and rheological properties aren’t the same.
Protein may dominate most flour specifications, but starch makes up around 70% of wheat flour and largely determines how dough absorbs water, gelatinises during baking and develops crumb structure. Damage to starch granules during milling isn’t automatically undesirable. Bread flour benefits from a controlled amount because damaged starch absorbs more water and provides fermentable sugars for yeast. Too much, however, can produce sticky doughs, inconsistent handling and reduced product quality.
That growing reliance on enzymes reflects the limitations of traditional specifications. Enzymes help compensate for natural flour variation, improve dough stability and maintain consistent products, but Dubat believes they are often being used to correct problems that could be better understood much earlier through a fuller assessment of flour functionality.
“Enzymes offer great help to correct flour properties,” he says. “I do not see any limitations in relying on enzymes except, maybe, the financial cost involved.”
He believes manufacturers should spend more time understanding why enzymes have become necessary in the first place. “The good question is why do bakers need to add enzymes? This is another evidence that current specifications are missing something.”
Rather than relying on standard flour specifications, he argues that bakeries should define the flour characteristics that genuinely matter on their own production lines before sharing those requirements with their millers. “If the ‘perfect flour’ characteristics for a baker can be established, it can therefore be communicated to the miller (becoming the new COA),” he states.
“As the miller has a much better-defined target, he can do his job (selecting wheat, blending them, grinding, and fine tuning with enzymes) to provide a baker with flour that does not need any enzymes or improvers.”
Dubat acknowledges that today’s reality is often very different.
“This is a best-case scenario as today bakers sometimes blend different flours and add enzymes without evaluating if these are really the ones needed (enzymes work differently based on the flour they are added in),” he says.
The next competitive advantage could be better flour data
The industry is entering a period where experience alone may no longer be enough. Experienced master bakers continue to retire; production lines are becoming increasingly automated; and artificial intelligence is beginning to influence everything from quality control to process optimisation.
“Bakeries are bigger, faster, producing many different products to adapt to increasingly demanding customer requests,” acknowledges Dubat. “And maybe more important, experience is leaving industry bakeries. Master bakers are retiring more and more rapidly at many of the world’s largest baking brands without being replaced. Additionally, AI implementation on the lines is rapidly becoming more common, adding more complexity to the operation.”
AI could accelerate this shift because predictive systems can only work with the information they’re given. If today’s flour specifications don’t fully describe how flour behaves during production, AI will simply make decisions using incomplete data.
“We all see how AI is already changing the way we are working. And we know that milling and baking industry are very active implementing AI-based solutions. AI efficiency is impacted by the type and the quality of data one is using to feed it.”
Researchers are reaching similar conclusions: studies suggest predictive baking models become significantly more reliable when they incorporate rheological measurements, starch functionality and process-specific flour characteristics rather than relying solely on conventional compositional data.
“Either the baker or miller can use same analytical data as always and as everybody, or, they can work together on an adapted dataset and use it to improve processes,” says Dubat. “I believe the future lies in this second dynamic road.”
Meeting specification has long been treated as the finish line. Increasingly, it may be just the starting point. Passing laboratory tests is one thing; delivering predictable performance shift after shift is another.
Studies:
Zhao Y, Karrar E, Peterson J, et al. Contribution of Protein, Starch, and Fiber Composition to the Prediction of Dough Rheology and Baking Quality in U.S. Hard Red Spring Wheat. Foods 2026, 15, 650. https://doi.org/10.3390/foods15040650
Van Rooyen J, Simsek S, Oyeyinka SA, et al. (2023). Wheat starch structure–function relationship in breadmaking: A review. Comprehensive Reviews in Food Science and Food Safety, 22, 2292–2309. https://doi.org/10.1111/1541-4337.13147
Kang S, Lee WS and Ha Y. Addressing Challenges in Wheat Processing and Storage Through Advanced Approaches. Food Bioprocess Technol 19, 339 (2026). https://doi.org/10.1007/s11947-026-04398-8
Hitlamani V and Inamdar AA. Stream-specific functional and rheological variability in roller-milled wheat flours: a sustainable approach to cookie quality optimization and flour utilization. Sustainable Food Technol. 2026; https://doi.org/10.1039/d6fb00108d
