blog

By Thom King, CFS, Founder, Chairman, Chief Innovation Officer Icon Foods

SteviaSweet™ RebD 95 deserves a serious place on the formulation bench. It is a high-purity rebaudioside D ingredient that can help build appealing sweetness in reduced-sugar foods and beverages. Its strongest argument is the finished product, the right sweetness, an acceptable finish, a workable process, and a cost that lets the product live beyond the sample room.

But we need to retire the idea that buying a more sophisticated stevia automatically produces a more sophisticated formula. You can put an all-star quarterback behind a terrible offensive line and still spend Sunday watching sacks. Sugar reduction works the same way. Sweetener selection matters. So does everything around it.

My recommendation is to screen RebD 95 alongside RebM and RM95D at matched sweetness, then select against the actual product brief. RebD 95 is particularly useful when you want to control the D contribution independently. RebM provides an M-focused reference. RM95D offers a supplied M–D combination worth evaluating when blend convenience and the final sensory result justify it.

Know which ingredient is on the bench

Icon Foods markets RebD 95 as SteviaSweet RD95, with a minimum of 95% rebaudioside D. For a concrete comparison, this paper uses SteviaSweet RM95, minimum 95% rebaudioside M, as the RebM reference. The published RM95D page describes an M-dominant blend with ≥80% RebM and ≤20% RebD. Those bounds do not establish an exact 80:20 formula. Confirm the current specification, assay basis, and lot COA before calculating glycoside contributions. 

How to use this paper

Product identity comes from published Icon Foods information. Sensory principles draw on the cited research. Application ranges, trial designs, and processing suggestions are proposed development starting points, not validated Icon Foods performance data, regulatory limits, or guaranteed sugar-replacement ratios. Qualification of the actual commercial grade and intended use still controls the final formula.

What RebD brings to a sweetening system

RebD and RebM are distinct steviol glycosides. Their structural differences matter, but a molecule diagram will not tell you whether your citrus drink tastes balanced or your protein shake finishes cleanly. Published consumer research comparing RebA, RebD, and RebM found greater bitter and chemical character for RebA under the tested conditions. That supports screening D and M for improved sensory quality; it does not prove that D or M wins every application. 

That study evaluated 0.1% weight per volume glycoside solutions, including comparisons with 14% sucrose. This is useful evidence at a particular concentration and in a particular matrix. It is not a validated usage recommendation for a commercial beverage, nor a head-to-head test of these three Icon Foods products. 

Compare at equal sweetness before judging quality

At equal ppm, one sample may simply be sweeter. That can alter the apparent acidity, bitterness, and flavor intensity. First map each ingredient’s dose response. Then compare the doses that produce the same perceived sweetness. Record onset, peak, residual sweetness, bitterness, astringency, and flavor identity separately. A single “tastes good” score throws away the clues you need to fix the formula.

Published RebM work describes a nonlinear concentration–response relationship and shows that physical form affects solubility. The practical lesson is straightforward: doubling dose does not guarantee doubling sweetness, and a solubility number belongs to a defined material under defined conditions. 

There is no defensible universal ranking of sweetness onset, linger, or cost across these three products without comparable data. Treat supplier positioning as a reason to run a trial. Let the trial decide who gets the jersey.

Practical usage levels and dosing math

All ppm below mean milligrams of commercial ingredient as supplied per kilogram of finished food or beverage, on a weight basis. These are suggested screening windows for RebD 95. For RebM and RM95D, begin with the same window to map response, then adjust to equal sweetness. The windows are not evidence that the three ingredients have equal potency.

These windows do not guarantee full replacement of a high-sugar benchmark. Start near the low end and move in 25–50 ppm steps. If the upper end still tastes weak, investigate the sweetness target, matrix suppression, and bulk sweetener contribution before chasing the target with more stevia. Higher doses require fresh sensory and solubility qualification.

Convert units before the batch goes sideways

100 ppm = 0.010% = 0.10 g/kg. A 1 kg trial at 200 ppm requires 0.200 g ingredient. A 1,000 kg batch at 200 ppm requires 200 g. Use batch mass; mg/L is only approximately equivalent to ppm in a dilute beverage with density near 1 kg/L.

Ingredient mass in grams = target ppm × batch mass in kilograms ÷ 1,000.

At exactly 95% assay, 200 ppm RD95 contributes 190 ppm RebD. If the target is 200 ppm active RebD, the required ingredient dose is 200 ÷ 0.95 = 210.5 ppm. Use the actual assay and moisture basis when precision matters. Do not purity-correct a sensory recommendation already stated on an as-supplied basis.

Dissolution and process best practices

Make the manufacturing route part of the experiment

Specify water temperature, addition order, mixing time, shear, solids, pH, and the smallest water volume the sweetener will encounter. A full-strength beverage and its syrup concentrate are different solubility problems. At 300 ppm in the drink, a fivefold concentrate requires 1,500 ppm ingredient. Passing the drink does not qualify the concentrate.

For a liquid process, reserve enough formulation water to dissolve the sweetener without creating an unqualified concentrated stock. Add the powder gradually with good circulation. If the supplier’s instructions allow it, compare ambient dissolution with a warm-water screen, for example 50–60°C. Record how long dissolution takes, then cool to the actual storage temperature and inspect again. This temperature range is a proposed trial condition, not a product-specific supplier instruction.

A clear warm solution can be supersaturated after cooling. Extra mixing may help dissolve particles; it cannot make an unfavorable equilibrium disappear. Do not release a formula because the beaker looked handsome at the end of a hot mix. That is the movie trailer, not the ending.

Keep concentrated stocks honest

Do not assume a convenient 1% aqueous stock is stable for any of these grades. Qualify the chosen stock concentration at its preparation and holding conditions, or weigh directly with a suitable calibrated balance. A 100 g bench trial at 200 ppm contains only 20 mg ingredient; increasing batch size can be more reliable than pretending a coarse balance is a microbalance.

If a qualified stock is used, calculate its water contribution and deduct that water from the formula. Log its concentration, preparation time, temperature, and appearance. A suspension is not a solution: sampling a settling stock can turn a tidy dose-response trial into random-number generation.

Design a process that fits the product

For powders, use geometric dilution into a compatible carrier already permitted in the formula, then validate blend uniformity and resistance to segregation. Evaluate reconstitution in cold water at the labeled serving size. A powder that performs only after prolonged hot stirring is a poor candidate for an instant hydration mix.

For chocolate and other low-water fat systems, avoid adding an aqueous sweetener stock. Evaluate particle size and distribution through the relevant refining or mixing process, and taste for local sweetness bursts. For bakery, account for moisture loss: ingredient ppm based on initial dough weight will rise as finished-product yield falls.

Check after the actual process

Taste and inspect after pasteurization, hot fill, UHT, baking, or other intended processing. Include cold holds, expected distribution conditions, and the intended package. Chemical stability and physical solubility are separate questions. A glycoside can remain chemically intact while precipitating. Use analytical work where needed to distinguish degradation, crystallization, and simple blending error.

Where each ingredient makes sense

Clear and sparkling beverages

RebD 95 belongs in the initial screen when the objective is appealing sweetness in an exposed flavor system. Lemon, lime, tea, and lightly flavored water leave little room to hide a lingering finish. Screen D against M and RM95D at equal sweetness and at serving temperature. Select D if it delivers the desired flavor and finish at an acceptable dose and passes clarity and cold-hold checks. Do not assume M or the blend is automatically more soluble.

Carbonate the finalists before selecting the winner. A still base does not reproduce the sensory experience of the finished sparkling drink. Retune acid and flavor only after you have established a useful sweetener comparison; changing everything at once destroys your ability to learn what helped.

Protein drinks dairy and plant based products

Include all protein, calcium, electrolytes, stabilizers, fat, and processing conditions before finalizing sweetness. An attractive result in flavored water tells you very little about a heat-treated protein drink. RebD 95 is useful here as an independent variable in a blend screen. RebM establishes the M-only reference, while RM95D tests whether the supplied combination produces a preferred balance with fewer ingredients to weigh.

Sweetness may reduce the perception of some off-notes, but none of these ingredients is a universal bitterness eraser. If a plant protein or mineral system remains harsh at the correct sweetness, investigate ingredient quality, mineral form, flavor design, and a separately qualified modulator. Turning up sweetness to bury bitterness can leave you with a drink that is both bitter and cloying. An impressive achievement, for all the wrong reasons.

Coffee cocoa and indulgent flavors

Evaluate RM95D and RebM early, with RebD 95 as a meaningful challenger, especially when creaminess and a pleasant post-swallow profile are priorities. This is a proposed screening order, not proof that the blend is intrinsically creamier. Sweetness can change the impression of richness; it does not supply the viscosity, lubrication, or solids missing from the formula. Taste coffee hot and cold if consumers will use it both ways.

Frozen desserts bars bakery and confectionery

Use any of the three to close a sweetness gap after designing the bulk system. In frozen desserts, stevia at these levels does not replace sugar’s substantial contribution to freezing-point depression. In bakery, it does not replace sugar’s contribution to bulk, water management, spread, or browning. In gummies, it does not restore the soluble-solids and gel conditions lost when sugar is removed.

Screen RD95 where independent D adjustment helps finish the sweetness profile. Select M or RM95D when the matched-sweetness trial shows a better result or easier production. Evaluate frozen products at realistic eating temperatures, bars through storage, and confections across repeated bites. The first bite is an audition. The whole serving decides whether the consumer buys it again.

Build blends with a defined job

Start by deciding what the other ingredients must contribute. Residual sucrose, allulose, tagatose, polyols, and fibers are not interchangeable bulk. Their sweetness, water interactions, processing behavior, digestive tolerance, and labeling implications differ. Choose them for the functions the product needs, then use RebD, RebM, or RM95D to close the remaining sensory gap.

Use a sweetness estimate without mistaking it for a result

Suppose the target is the perceived sweetness of 6% sucrose and the rest of the formula is provisionally estimated to deliver 3% sucrose-equivalent sweetness. At an assumed potency of 200 times sucrose, closing the remaining 3% would suggest 0.015% sweetener, or 150 ppm. That is a useful center point for a trial. It is not proof that the blend will taste like 6% sugar.

The 200-fold assumption is broadly consistent with Icon’s published RD95 and RM95 descriptions, but product descriptions also give differing potency estimates across the portfolio. Sweetness contributions can be non-additive. Confirm the complete blend by tasting it against a reference instead of adding marketing multipliers in a spreadsheet. [1–3]

A trial plan that produces usable answers

Round one maps each commercial ingredient at 100, 200, and 300 ppm in the same finished matrix. Include the target-sweetness reference and an unsweetened matrix control. This creates nine ingredient treatments plus two controls. For a partial-reduction product, move the levels lower; for a challenging matrix, bracket around the application window on page 3.

Round two refines the dose of each ingredient to match the reference sweetness. Compare the matched samples using random three-digit codes and balanced serving order. Use more than one taster, repeat on another day, and split sessions to control adaptation and fatigue. Provide consistent rinse procedures and recovery time.

Round three asks whether custom D addition improves the M reference. As an exploratory design, screen RM95:RD95 at 100:0, 90:10, 75:25, and 50:50 by commercial ingredient mass. First hold total ingredient dose constant to map behavior, then rematch sweetness for preference testing. Include commercial RM95D as its own sample. A laboratory 90:10 mixture is not established as compositionally or functionally identical to RM95D.

These blend ratios are experimental factors, not recommended commercial recipes. Keep the total solids, acid, flavor, and all other sweeteners constant while you learn the M–D effect. A preferred mixture does not itself prove sweetness synergy; demonstrating synergy requires comparison against a defined additive expectation from the individual dose-response curves.

Capture the finish as carefully as the first impression

Score sweetness and bitterness during consumption, shortly after swallowing, and again around 30 and 60 seconds later. Record astringency, cooling, flavor suppression, and overall preference separately. Retest the best candidates after processing and storage. Do not invent a universal rule that D supplies the front and M supplies the tail; measure what happens in your product.

Cost and troubleshooting

Compare cost at the dose that actually wins

Price per kilogram is only half the equation. Compare qualified doses that achieve the same sweetness and acceptable sensory quality. Then include changes to bulk ingredients, flavor, masking, process time, yield, and rejected product. A cheap ingredient that needs extra masking and still leaves a weak finish is not a bargain.

Sweetener cost per metric ton of finished product = dose in ppm × ingredient price per kg ÷ 1,000.

All prices and performance doses in this example are hypothetical, not quotations or test results. They show why a more expensive kilogram can produce a cheaper finished product. Obtain current delivered prices and use your own matched-sweetness data before choosing.

Diagnose the failure before adding another ingredient

When a sample fails, preserve its formula, lot identity, process record, and storage history. Those details make a failed trial useful. “It tasted weird” is not a technical record.

Qualification and final selection

Verify the commercial material

Before a production decision, obtain current specifications and representative COAs for RD95, RM95, and RM95D. Confirm individual and total glycoside assay, moisture basis, physical form, particle size where relevant, dissolution guidance, microbiological limits, contaminants, and applicable certifications. Verify the production route rather than inferring leaf extraction, bioconversion, or fermentation from the glycoside name.

Lock the qualified grade and supplier change-control expectations. A similar assay does not establish identical powder behavior or taste. Requalify meaningful changes in source, process, physical form, or glycoside composition, especially for clear drinks and concentrated premixes.

Keep labeling attached to the intended use

FDA describes certain highly purified steviol glycosides as subjects of GRAS notices to which it has not objected. That does not mean every stevia material or manufacturing route is automatically covered, and it is not blanket “FDA approval” of a brand. Confirm that the documentation for the specific material covers the proposed food use and market. 

Have regulatory review confirm ingredient naming, claims, category permissions, and any applicable limits before commercialization. A low use level by itself does not establish eligibility for declaration as natural flavor. Organic, non-GMO, and other certifications also require the relevant current documentation; a glycoside name is not a certificate.

Choose the ingredient for the actual job

Choose RebD 95 when it wins the matched-sweetness comparison or gives you useful control over the D contribution to a custom system. It is a strong candidate for formulators who want to tune the final blend themselves and can qualify its behavior in their process.

Choose RebM when the M-focused reference delivers the preferred flavor and finish at an acceptable cost, and the commercial grade fits the manufacturing route. RebM does not need an additional glycoside simply because a blend sounds more advanced.

Choose RM95D when the supplied M–D combination wins on sensory performance, economics, or manufacturing simplicity. Its practical benefit may be one less component to handle and a supplier-controlled blend. Do not promise improved solubility, universal superiority, or proven synergy without data for the material and application.

My position is simple: RebD 95 should compete on the bench, not sit on the sidelines because RebM gets more attention. Build the product around its required functions, compare the candidates fairly, and follow the winner through processing and shelf life. Consumers will never ask which glycoside won the trial. They will tell you whether the product is worth buying again.

For RD95, RM95, and RM95D samples and current technical documentation, contact Icon Foods at sales@iconfoods.com or visit iconfoods.com.

Sources and technical notes

Sources reviewed September 2026. Product pages establish published supplier descriptions; current controlled specifications and lot COAs remain the basis for commercial qualification. Research on defined glycoside preparations is not direct validation of a branded grade. No unpublished Icon Foods comparative trials were supplied for this paper.

1. Icon Foods. SteviaSweet RD95. Product identity and published potency description. https://www.iconfoods.com/product/steviasweet-rd95/
2. Icon Foods. SteviaSweet RM95. Product identity and published potency description. https://www.iconfoods.com/product/steviasweet-rm95/
3. Icon Foods. SteviaSweet RM95D. Published M and D composition bounds and product positioning. https://www.iconfoods.com/product/steviasweet-rm95d/
4. Consumer-Based Sensory Characterization of Steviol Glycosides (Rebaudioside A, D, and M). Foods, 2020, 9, 1026. DOI 10.3390/foods9081026. Consumer study; results apply to the tested preparations and conditions. https://pmc.ncbi.nlm.nih.gov/articles/PMC7466183/
5. Prakash I, Markosyan A, Bunders C. Development of Next Generation Stevia Sweetener: Rebaudioside M. Foods, 2014, 3, 162–175. DOI 10.3390/foods3010162. Concentration response and material-specific physical behavior. https://pmc.ncbi.nlm.nih.gov/articles/PMC5302307/
6. US Food and Drug Administration. Aspartame and Other Sweeteners in Food. Regulatory background on highly purified steviol glycosides. https://www.fda.gov/food/food-additives-petitions/aspartame-and-other-sweeteners-food

Interpretation of the development recommendations

The application windows and experiments are author-proposed formulation guidance. They are deliberately presented as screening tools because a useful commercial dose depends on target sweetness, residual sugars, bulk sweeteners, acidity, protein, minerals, flavor, processing, temperature, and storage. They should not be presented externally as supplier-validated performance claims without supporting trials.

The paper does not assign a single solubility limit to RD95, RM95, or RM95D because grade-specific equilibrium and handling data were not available. It also does not assign a fixed price advantage, a universal temporal profile, or a guaranteed sugar-equivalent value to any of the three products.