How to Choose Stevia Extract Grades for Food and Beverage Applications

01 Quick selection

Quick Selection: Where Should a Formulator Start?

The table below is intended as a starting screen for pilot work, not a fixed formulation recipe.

There is deliberately no row saying “all beverages = Reb M.”

That recommendation is too broad to be useful.

The correct grade depends on what the stevia has to accomplish inside the formula.

02 Grade definition

What Does “Stevia Grade” Actually Mean?

The word grade is used loosely in the stevia market.

A request such as:

does not yet define one ingredient.

A commercially useful specification needs to distinguish the predominant glycoside, individual assay, total steviol glycoside profile, analytical method, production route and relevant physical characteristics.

Consider four specifications that may all contain the number “95%”:

This distinction is easy to overlook during sourcing.

Two Products Can Both Say “95%” and Still Be Fundamentally Different Grades

In a TSG specification, the assay refers to a defined group of steviol glycosides in combination. The relative amounts of stevioside, Reb A and minor glycosides can influence the character of the material.

In a high-purity Reb A or Reb M grade, the specification is centered on the individual named glycoside.

Research comparing differently processed commercial stevia extracts has shown substantial differences in glycoside composition between manufacturing approaches. In one study, a membrane-purified extract contained considerably more minor natural glycosides than several crystallized extracts, while trained sensory evaluation also examined differences in sweetness, bitterness, licorice and metallic character.

This is why steviol glycoside purity is an analytical specification, not a complete sensory specification.

HPLC tells you whether the material meets a chemical identity and assay requirement.

It does not tell you whether your lemon beverage will taste better.

03 Market

Gate One: Confirm the Destination Market Before You Formulate

A technically excellent sweetener is not useful if the ingredient specification or production route does not fit the regulatory framework of the market where the finished product will be sold.

This becomes especially relevant with Reb D and Reb M because modern commercial materials may be obtained through different manufacturing routes rather than only by conventional extraction and refining from stevia leaf.

For formulation teams selling into multiple regions, regulatory review should therefore happen before final sensory approval, not after months of bench development.

The key procurement question is not only:

It is also:

The detailed classifications do not need to be repeated here. For E 960a–E 960d, ADI, production-route distinctions and EU, U.S. and China requirements, use our Steviol Glycoside Regulation Guide.

The practical rule is simple:

That is the expensive order.

04 Sugar reduction

Gate Two: Start With the Amount of Sugar You Want to Remove

This is one of the most useful filters in how to choose stevia extract grades, yet it is often missing from supplier selection guides.

“Sugar reduction” is not one formulation target.

Replacing 20% of sucrose, replacing 60%, and removing sucrose completely place very different demands on the stevia system.

Partial Sugar Reduction

When substantial sucrose remains, the sugar continues to supply sweetness quality, mouthfeel and temporal characteristics. Stevia only needs to replace part of that sweetness.

That can make Reb A a very rational starting point.

A published iced-tea study gradually replaced sucrose using high-purity Reb A. Formulas at 20% and 40% sucrose replacement remained comparable with the sucrose reference in consumer preference and acceptance, despite measurable artificial sweetness, licorice-like, metallic, sweet-aftertaste and bitter-aftertaste characteristics.

The important commercial lesson is not that 40% is a universal threshold.

It is that premium glycosides should solve a problem rather than simply replace a less expensive grade because they sound more advanced.

If a 30% sugar-reduction formulation already meets its sensory target with Reb A, there may be little economic reason to move automatically to Reb M.

Deep Sugar Reduction

As more sucrose disappears, the high-intensity sweetener has to carry a larger share of the sweetness profile.

At that point, defects that were previously hidden become more visible. Bitterness, delayed sweetness, sweet linger and licorice-like notes can become limiting, while reduced sucrose also changes body and texture.

The question then begins to shift from:

toward:

This is where Reb D, Reb M and glycoside blends become more commercially relevant.

Zero Sugar

Zero-sugar development creates the highest sensory demand because sucrose is no longer present to contribute its characteristic onset, peak and decay.

Consumer research directly comparing Reb A, Reb D and Reb M found Reb D and Reb M closer to sucrose than Reb A for in-mouth sweetness and bitterness under the experimental conditions. Reb A showed significantly greater bitterness.

But that result needs a second sentence.

The same work showed that temporal characteristics remained important, including lingering sweetness. That distinction matters for premium beverages where the finish is often as important as the first sip.

05 Matrix

Gate Three: Judge the Difficulty of the Actual Food Matrix

Steviol glycosides do not have one fixed sensory performance independent of the food around them.

Acidity, aroma intensity, fat, protein, viscosity, temperature and solids all influence the way sweetness and bitterness are perceived.

This is why a useful guide to the best stevia for food and beverage applications must describe formulation challenges rather than simply attach one glycoside to each application.

This is a much more useful way to read application recommendations.

The application identifies the problem.

The problem identifies the candidate grade.

Testing decides whether the candidate is actually correct.

06 Clear beverages

Clear and Lightly Flavored Beverages: Where Premium Glycosides Have the Strongest Case

Flavored waters, light citrus drinks and delicate botanical beverages give the sweetener relatively little sensory cover.

A bitter note that disappears in cocoa can remain obvious in a clear lemon beverage.

For these applications, Reb D and Reb M are logical candidates because published sensory research indicates lower bitterness than Reb A.

But there is an important purchasing nuance.

A zero-sugar flavored water and a beverage retaining 60–70% of its original sucrose should not automatically use the same sweetener system.

The zero-sugar product places much greater weight on the quality of the steviol glycoside itself.

The partially sweetened product may obtain sufficient sensory quality from Reb A while sucrose continues to provide much of the expected profile.

When Is It Worth Upgrading From Reb A to Reb M?

The upgrade becomes commercially interesting when Reb A has already met the target sweetness but bitterness or finish is now the factor preventing the formula from passing sensory approval.

It may also make sense when the product is zero sugar, very lightly flavored, positioned as premium, or requires so much flavor masking with Reb A that the cheaper raw-material price no longer reflects the true formulation cost.

By contrast, if the product contains strong flavors, retains substantial sugar, and consumers cannot meaningfully distinguish the Reb M version from the Reb A version, the premium grade may simply be over-specification.

That decision should come from the formula, not from the supplier’s product hierarchy.

If the specific failure is bitterness, licorice character or a long sweet finish, diagnose the defect before changing the whole sweetening system. Our stevia aftertaste guide separates the main off-notes and the formulation variables that can change them.

07 Fruit beverages

Fruit Beverages: Higher Purity Can Help, but the Matrix Still Wins

Fruit drinks illustrate why stevia selection cannot be performed in water alone.

A study of peach juice compared stevia preparations with different Reb A concentrations. Higher Reb A content reduced the intensity and duration of bitterness relative to a lower-Reb-A extract, yet stevia-sweetened samples still differed sensorially from the sucrose reference and consumer acceptance remained challenging in that matrix.

This gives buyers two useful pieces of information.

First, composition matters.

Second, improving the purity of one glycoside does not guarantee that the entire sensory problem disappears.

Fruit acidity, pulp, natural sugars and aroma all change how a sweetener is perceived. The same research notes that sweeteners can behave differently between juice matrices because fruits differ in chemical composition, total solids, acidity and pH.

So a sample approved in peach, mango or lemon should not be treated as proof of equivalent performance in every fruit beverage.

08 Dairy & frozen

Dairy and Frozen Desserts: Taste the Sweetener at the Temperature It Will Be Consumed

Dairy formulations introduce another layer of complexity.

Fat, protein and viscosity change perception, while frozen desserts add serving temperature as another variable.

A consumer study evaluated ice creams sweetened with Reb A, Reb D and Reb M against a sucrose control. Reb D- and Reb M-sweetened products received significantly higher liking scores than the Reb A version, and among the three steviol glycosides, Reb M produced sweetness intensity comparable with the sucrose control under the study conditions.

That makes Reb D and Reb M useful starting candidates for premium frozen applications.

It does not mean every dairy customer should buy Reb M.

A strongly flavored protein shake and a vanilla frozen dessert have completely different sensory exposure. Their cost limits can also differ dramatically.

A useful dairy pilot therefore evaluates the grade at the actual serving temperature and actual solids level, rather than tasting the ingredient in room-temperature water and extrapolating the result.

09 Grade comparison

Reb A, Reb D, Reb M — or a Blend?

Reb A, Reb D and Reb M are often marketed as if they form a straight quality ladder.

That is not the most useful way to think about them.

Reb A Is Not an Obsolete Grade

Reb A remains commercially relevant because formulation is not a competition to buy the newest molecule.

A good purchasing decision selects the lowest-complexity grade that passes the project requirements.

For partial sugar reduction, strong-flavor matrices and price-sensitive applications, that can still be Reb A.

Reb M Is Not an Automatic Upgrade

Reb M has genuine sensory advantages in published comparisons, but its value should be measured against the problem it solves.

Paying more for a cleaner glycoside makes sense when the cleaner profile changes consumer acceptance, reduces masking requirements or allows a higher level of sugar reduction.

It makes less sense when the final product cannot demonstrate a commercially meaningful difference.

The Fourth Option Is a Blend

The choice does not have to be binary.

Recent work examined Reb A, Reb D and Reb M mixtures at iso-sweet conditions and found significant differences in bitterness between blends. Consumer testing in ice cream also found that combining Reb A with Reb D and M could decrease bitterness and improve liking.

This creates an important middle ground:

The economic optimum may sit between them.

For a deeper physical-property comparison, see Reb A vs Reb D vs Reb M Solubility.

10 Sweetness potency

Do Not Use “200–300 Times Sweeter Than Sugar” as a Dosing Formula

High-intensity sweeteners are often described with one sweetness multiple.

Reb A may be described as roughly 200–300 times sweeter than sucrose, while Reb M is also frequently presented using a broad relative-sweetness range.

Those numbers are useful as introductory descriptions.

They are poor substitutes for formulation work.

Perceived relative sweetness changes with concentration, reference sweetness level and matrix. Reb M research has specifically reported concentration-dependent sucrose equivalence rather than one fixed multiplication factor.

A buyer therefore should not calculate:

That arithmetic assumes a linear relationship that sensory perception does not necessarily follow.

The better sequence is to define the target sweetness, establish an approximate starting concentration, and then determine iso-sweetness in the actual matrix.

A Useful Formulation Rule

That distinction prevents a surprisingly common formulation error.

11 Processing

Gate Four: Processing Can Eliminate an Otherwise Excellent Grade

A sample can win the sensory trial and still lose in manufacturing.

That happens because the concentration and conditions experienced during production may be very different from those in the finished beverage.

pH, Temperature and Time Need to Be Considered Together

Reb A has been evaluated in mock beverages across pH 2.8–4.2 and storage temperatures from 5°C to 40°C over 26 weeks. The study found degradation to be dependent on pH, temperature and time, with greater degradation-product formation under lower pH and higher-temperature conditions.

This does not mean Reb A cannot be used in acidic beverages.

It means that “stable in beverages” is too vague to function as a technical specification.

The buyer needs to know which beverage conditions are being discussed.

Reb M research similarly reports solution stability as dependent on pH and temperature, with greatest reported stability in roughly the pH 4–8 range and lower stability at more extreme acidity.

Again, the commercial lesson is validation, not avoidance.

12 Process concentration

Finished-Product Concentration Is Not the Same as Process Concentration

This is one of the most important practical differences between laboratory formulation and factory operation.

Suppose the finished beverage contains:

That concentration may appear trivial from a dissolution standpoint.

But suppose the plant prepares a 20× concentrated dosing solution.

The intermediate concentration becomes:

The ingredient therefore does not only need to behave at 250 mg/L.

It needs to behave at 5 g/L during the manufacturing step where it is actually dissolved.

This becomes especially relevant for some high-purity minor glycosides.

Published Reb M work reported a substantial difference between crystalline and amorphous forms: crystalline Reb M was only slightly soluble in water under the reported test conditions, while amorphous material showed considerably higher solubility; equilibrium solubility was reported separately at 25°C.

The important purchasing message is not that every commercial Reb M behaves identically to those research samples.

It is that physical form and test conditions matter.

A specification saying only:

is often not enough for an industrial beverage project.

Better Procurement Language

Instead of asking only “Is it water soluble?”, define a test condition such as:

Now solubility becomes a measurable process specification rather than a sales adjective.

13 Production method

Premix, Direct Addition or Dry Blending Can Change the Grade Decision

The method used to introduce stevia into production also deserves attention.

A powdered beverage may disperse the sweetener through a dry carrier system before the consumer adds water.

A bottling line may prepare a concentrated liquid sweetener premix.

A gummy plant may add the material into a hot syrup with acids and flavors.

Those are not equivalent processes.

The same grade can be completely acceptable in one and inconvenient in another.

For this reason, a useful commercial specification should record not just the finished dosage but the highest concentration, lowest dissolution temperature, shortest available mixing time and relevant holding period encountered anywhere in production.

That information often predicts operational problems better than the finished-product ppm alone.

14 Cost in use

Gate Five: Compare Cost-in-Use, Not Price per Kilogram

Stevia purchasing decisions are frequently distorted because suppliers quote dollars per kilogram while formulators work in milligrams per liter.

Those are different economic frames.

The first calculation should therefore be:

Consider a hypothetical comparison:

At first glance the premium system appears 150% more expensive per kilogram.

At finished-beverage level, however, the hypothetical difference shown here is only $9 per 1,000 L before other formulation costs are considered.

That is a very different purchasing conversation.

But the opposite conclusion is also possible.

If the Reb A formula passes consumer testing without extra masking, the additional $9 may deliver no meaningful commercial value.

This is why price per kilogram should never be the final comparison metric.

15 Total system cost

Calculate the Total Sweetening-System Cost

The real cost does not stop with the stevia powder.

A lower-priced grade may require additional flavor correction, sweetness modulation or repeated formulation work. Removing sucrose may also require separate ingredients to restore body or solids, regardless of which high-intensity sweetener is selected.

A better economic model is:

Development time can be treated separately because its financial value differs from one manufacturer to another.

This gives procurement a better decision rule:

This also prevents the opposite mistake: using Reb M simply because it is positioned as a premium ingredient.

Premium only has value when it changes the result.

If erythritol is part of the system, the economics and formulation role change because erythritol contributes bulk as well as sweetness; our Stevia vs Erythritol comparison treats those as separate jobs. If the alternative high-intensity sweetener is monk fruit, see Monk Fruit vs Stevia before assuming one category has the cleaner taste profile.

16 Pilot testing

A Three-Stage Pilot Before Approving a Stevia Grade

Supplier samples should begin qualification, not finish it.

The most defensible grade selection process moves through three distinct stages.

Stage 1 — Equal-Sweetness Screening

Start with the realistic candidate set rather than comparing everything in the catalog.

For example, a project may screen Reb A, Reb D, Reb M and one A/D/M blend.

The samples should be compared at approximately equal perceived sweetness, not equal ppm.

At this stage, focus on sweetness onset, peak sweetness, bitterness, licorice-like notes and lingering sweetness.

The objective is simply to eliminate weak candidates.

Do not issue commercial approval from this test.

Stage 2 — Real-Matrix Comparison

Move the surviving candidates into the actual product.

Keep sugar-reduction target, pH, acid system, flavor loading, protein or fat content, solids and serving temperature consistent while changing the stevia system.

This is where the meaningful grade decision begins.

Research in peach juice and ice cream illustrates exactly why. The sensory behavior observed in a real food matrix contains interactions that cannot be predicted reliably from water tasting alone.

A grade that looks excellent in water may lose most of its advantage once strong flavor is added.

Another grade that seems imperfect in water may become completely acceptable in the commercial formulation.

Stage 3 — Process and Storage Validation

Once the sensory candidate has been identified, reproduce the manufacturing conditions.

Evaluate the highest intermediate concentration, actual water temperature, mixing time, heat exposure, pH, process hold and relevant storage conditions.

A simple internal evaluation can compare Day 0 with later checkpoints and record clarity, precipitation, sweetness intensity, bitterness and lingering character.

At this stage the question changes from:

to:

That is the difference between an ingredient sample and a commercially approvable raw material.

17 Matrix validation

Do Not Approve a Stevia Grade From Water Tasting Alone

Water screening is useful precisely because it exposes differences.

It removes much of the noise created by acids, flavors, proteins and other ingredients.

But that strength is also its limitation.

The consumer will not drink your laboratory water standard.

The finished matrix can suppress, reveal or reshape sensory characteristics.

Therefore:

That simple procedure is much more defensible than choosing a grade because a supplier's neat solution tasted clean during a sales meeting.

18 Consistency

Batch-to-Batch Consistency Matters More Than a Perfect Sample

This is often overlooked in online stevia buying guides.

A supplier can send one excellent 50 g sample.

Commercial procurement needs the fifth and fiftieth production batch to behave predictably as well.

A buyer therefore should think beyond:

and ask:

For stevia ingredients, chemical assay is only one part of that question.

Not every buyer needs every parameter as a formal release specification.

The point is to identify which parameters actually affect your process and finished product.

A Better Way to Approve Samples

Instead of writing:

record what the sample has established.

For example:

Now the sample has created commercial acceptance criteria.

That is far more useful than remembering that “Sample A tasted good.”

19 Supplier qualification

Supplier Qualification: What Buyers Should Actually Ask

A capable supplier should be able to answer more than “COA available.”

The purchasing file should establish what the ingredient is, how it was made, whether its regulatory documentation matches the destination market, and whether its physical behavior is relevant to the manufacturing process.

This is where supplier selection becomes more efficient.

The more precise the buyer's technical question, the easier it becomes to distinguish a specification-based supplier from a seller who only knows the commercial product name.

20 Supplier brief

How to Brief a Stevia Supplier So You Get a Useful Recommendation

A supplier cannot make a meaningful grade recommendation from:

At minimum, the recommendation should be based on the intended application, target degree of sugar reduction, destination market and important processing conditions.

For beverage work, the supplier should also know whether you make a concentrate and roughly what the maximum in-process stevia concentration will be.

For sensory-sensitive applications, explaining whether the product is clear/lightly flavored or strongly flavored can be more useful than simply saying “drink.”

The difference between these two enquiries is substantial:

versus:

The second enquiry gives a competent supplier enough information to make an actual recommendation.

21 Applications

What Should You Start With for Common Applications?

By this point, the application table becomes more meaningful because we can interpret it as a shortlist rather than an absolute answer.

For a partially sugar-reduced beverage, Reb A or a suitable high-purity stevia extract remains a logical benchmark. If that formulation fails because of bitterness rather than sweetness level, Reb D, Reb M or a blend becomes the next comparison.

For a zero-sugar clear beverage, beginning with Reb M or a D/M-containing system can save unnecessary sensory iterations, particularly where the product has little flavor masking.

For dairy and frozen products, Reb D and Reb M deserve early evaluation based on published matrix studies, while the cost advantage of Reb A should still be tested rather than dismissed.

For gummies and confectionery, the acid and flavor system should be developed alongside the sweetener because those variables materially affect the final perception.

For coffee, cocoa and other strong-flavor foods, an expensive minor-glycoside specification should earn its place through testing. Strong product flavor may make an economical Reb A or TSG system perfectly adequate.

For powdered beverages, cold-water reconstitution and dosage consistency can be at least as important as the neat sensory profile of the powder.

The pattern is consistent:

22 Decision framework

Final Decision Framework

A practical buying decision can now be reduced to seven questions.

This sequence is more useful than asking whether Reb M is “better” than Reb A.

Sometimes it is the better choice.

Sometimes it is unnecessary.

And sometimes neither individual glycoside is the best answer because a blend provides the better compromise.

23 Bottom line

The Core Buying Principle

There is no single best stevia extract grade for food and beverage applications.

Reb A remains valuable where cost efficiency, established supply and partial sugar reduction are the priorities. Reb D and Reb M become increasingly useful when the sensory limitations of Reb A prevent a formula from reaching its target. Blended glycoside systems provide another route when the cleanest single-glycoside option is more than the application actually needs.

The assay matters.

But 95% is not a taste specification.

The application matters.

But “beverage” is not enough information to choose a grade.

Sweetness potency matters.

But “300 times sweeter than sugar” is not a dosing equation.

Solubility matters.

But finished-product concentration may not be the concentration your factory actually has to dissolve.

Price matters.

But $/kg is not the same as cost per finished product.

And finally, a good sample matters.

But batch-to-batch repeatability matters more than one perfect sample.

For procurement teams, the best definition of the right grade is therefore straightforward:

If you are selecting between Total Steviol Glycosides, Reb A, Reb D or Reb M, send us your application, sugar-reduction target, destination market and highest in-process concentration. We can recommend a practical starting grade for pilot testing and provide the relevant specification, COA and sample.

24 Questions

Frequently Asked Questions

25 Evidence base

References

[1] Tao R, Cho S. Consumer-Based Sensory Characterization of Steviol Glycosides (Rebaudioside A, D, and M). Foods. 2020;9(8):1026. DOI: 10.3390/foods9081026. The study used 126 consumers to compare Reb A, Reb D, Reb M and sucrose, including immediate and lingering sweetness and bitterness. PMID 32751835

[2] Majchrzak D, Ipsen A, Koenig J. Sucrose-replacement by rebaudioside A in a model beverage. Journal of Food Science and Technology. 2015;52:6031–6036. DOI: 10.1007/s13197-014-1624-z. It evaluated progressive sucrose replacement in iced tea and supports the distinction between partial and deeper sugar reduction. PMC4554609

[3] Espinoza MI, Vincken JP, Sanders M, et al. Identification, quantification, and sensory characterization of steviol glycosides from differently processed Stevia rebaudiana commercial extracts. Journal of Agricultural and Food Chemistry. 2014;62(49):11797–11804. DOI: 10.1021/jf502878k. PMID 25393842

[4] Medeiros A, Tavares E, Bolini HMA. Descriptive Sensory Profile and Consumer Study Impact of Different Nutritive and Non-Nutritive Sweeteners in a Peach Juice Matrix. Foods. 2022. The work demonstrates the interaction between stevia composition and a real fruit matrix. PMC8775159

[5] The Effect of Steviol Glycosides on Sensory Properties and Acceptability of Ice Cream. The consumer study compared Reb A, Reb D and Reb M in a real frozen dairy matrix and found higher liking for Reb D and Reb M than Reb A under the test conditions. PMID 35741943

[6] Prakash I, Markosyan A, Bunders C. Development of Next Generation Stevia Sweetener: Rebaudioside M. Foods. 2014;3:162–175. It reports Reb M sweetness, physical form, aqueous solubility and stability characteristics. Because the work includes industry-affiliated authors, its physicochemical data are useful here but should not be treated as the sole basis for sensory claims. PMC5302307

[7] Prakash I, Clos JF, Chaturvedula VSP. Stability of Rebaudioside A under acidic conditions and its degradation products. Food Research International. 2012;48(1). DOI: 10.1016/j.foodres.2012.02.015. Reb A was studied in mock beverages across pH 2.8–4.2 and several storage temperatures. DOI source

[8] Moazzem MS, et al. Taste profiles of steviol glycoside blends and their application in ice cream. International Journal of Gastronomy and Food Science. 2025;40:101143. The study examined Reb A/D/M blends and provides useful evidence for considering a blended system rather than treating A, D and M as mutually exclusive choices. Article