The Pricing Decision Playbook for 2026

Companion to Report 004, "Pricing Myths: Charm Pricing, Tiers, and the Numbers Founders Repeat." Evidence standard: peer-reviewed academic pricing research first (Anderson and Simester on 9-endings, Thomas and Morwitz on the left-digit effect, Marn and Rosiello on the 1% price statistic, Tversky and Kahneman on anchoring), transparent large-sample vendor datasets second, uncited "this trick lifts revenue X%" figures rejected outright. This is the Synthesis layer of the stack. The report proves what is true. This turns it into what to do.

The numbers this playbook runs on

11.1% vs 3.3% - Marn and Rosiello, 2,463 companies: a 1% price improvement lifted average operating profit 11.1%, versus 3.3% for a 1% volume gain. So do this: take your pricing power seriously, and separately, test what volume you actually lose before you bank the 11.1%. The figure holds volume constant. Your market does not.
$39 outsold $34 - Anderson and Simester's field experiment: a women's clothing item sold more at the higher, 9-ending price than at the lower, round one. So do this: test a 9-ending on new items specifically, and drop it the moment you're also running a "Sale" cue, because the two signals compete.
5 to 10% - McKinsey's estimate for the return-on-sales improvement from value-based pricing. So do this: quote this range, not the unsourced "31% higher profit" figure that circulates without a study attached. It does not exist in this report's sources.
$2.99 vs $3.00, not $3.59 vs $3.60 - Thomas and Morwitz's boundary condition for the left-digit effect. So do this: the penny-drop trick only fires when it rolls the dollar digit down. Check that condition before you use it.

Start here

Most pricing "hacks" a founder hears started as a real experiment. Somewhere between the journal and the growth-hacking newsletter, every condition that made the finding true got sanded off, and what's left is a portable rule that sounds like physics: always end in 9, always show three tiers, a 1% price increase always beats a 1% volume increase. None of that survives contact with the original paper.

A few effects do replicate, and they replicate exactly where the study said they would. Charm pricing works on new items when there's no competing discount signal. The left-digit effect fires only when the leftmost digit rolls over. Anchoring is one of the sturdiest findings in decision science, but its dollar-for-dollar effect on your specific pricing page has never been controlled-tested by anyone. The move is not to memorize the tactic. It's to price from your own margin math and your own customers' willingness to pay, and use these effects only where your situation matches the condition that made them true.

Part 1. The SOP

Four levers that hold up under the report's evidence bar, in the order they matter.

A. Get the price level right first

This is the lever with the biggest number attached to it, and the one most founders skip because it feels less clever than a pricing-page trick. Marn and Rosiello's 1992 study of 2,463 companies found a 1% price improvement lifted average operating profit 11.1%, more than any other single lever. The catch, which the statistic itself does not carry, is that it assumes volume does not move when price does.

Model your price change against your own elasticity, not the industry-average 11.1%. Run a real price test on a segment before rolling it out everywhere.
Ask what volume you'd lose at the new price, in writing, before you raise it. If nobody can answer, that's the actual open question, not the arithmetic.
Favor price over cost-cutting as the first lever to pull. It has no cost attached and flows straight to margin, which is why it ranks first in the underlying data.
Re-run the test whenever your product, buyer, or competitive set changes. The 1992 number describes an average company in an average year, not yours in this one.

B. Use charm pricing where the study's conditions actually hold

Anderson and Simester ran three field experiments with a real mail-order retailer and found a $9 ending raised demand every time. The lift was strongest on new items and weakened when a "Sale" cue was also present.

Reach for a 9-ending on new SKUs and new plans, where buyers have no prior reference price to compare against.
Drop the tactic the moment you're also signaling a discount. Two price cues compete instead of stacking.
Check the left-digit condition before you commit to a $X.99 price: it only works when it rolls the dollar digit down ($2.99 vs $3.00), not when it doesn't ($3.59 vs $3.60).
On premium or enterprise positioning, treat a round number as the safer default. The original study never tested luxury or B2B pricing, and a 9-ending can read as cheap exactly where you don't want it to.

C. Build an anchor or a decoy deliberately, then test it

Anchoring is a genuinely sturdy mechanism. Tversky and Kahneman showed that an initial number biases a later estimate through insufficient adjustment, and that finding has held up for fifty years. What has not been controlled-tested is its dollar effect on a real pricing page.

If you add a high anchor tier, treat the lift on the tier below it as a hypothesis, not a guaranteed multiplier. A/B it.
If you build a decoy, build it to be worse than exactly one option and only that option. The effect lifts a specific built target. It does not summon itself by putting three prices on a page.
Expect the decoy to work less reliably than the lab suggests. Yang and Lynn's replication work found the effect largely disappears once the choices look like real products instead of two-number abstractions.
Run the anchor and the decoy as separate tests from each other. Stacking untested tactics makes the result impossible to attribute.

D. Test on your own customers before you adopt anyone else's number

Every tactic in this report that holds up does so under a named condition. None of them holds up as a universal law, and the tactics with the least evidence, like "value-based pricing lifts margin by a specific percentage" or "16.7% is the right annual discount," are exactly the ones sold hardest as settled facts.

Run your own price test before you change a live price for every customer.
Treat any round percentage you can't trace to a named study as marketing, not evidence.
Write down which condition (new item, no sale cue, left-digit rollover, differentiated product) applies to your case before you use a tactic built on it.
Re-test annually. Buyer behavior, competitive pricing, and your own product differentiation all shift the ground the original study stood on.

Part 2. Real effect vs overclaimed rule

Tactic	What actually replicates	What gets oversold
Charm pricing (9-endings)	Raised demand in three field experiments on a real retailer, strongest on new items, absent a "Sale" cue.	"Always end in 9." Can backfire on premium or enterprise positioning, which was never tested.
Left-digit effect	$2.99 reads as meaningfully cheaper than $3.00, because the leftmost digit rolls over.	"Just drop a penny." $3.59 vs $3.60 gets almost none of the effect, because the dollar digit doesn't change.
Decoy / anchoring effect	A real, documented violation of rational choice in controlled, abstract-number settings; a deliberately built decoy can lift a specific target option.	"Put three tiers on the page and the middle wins." The effect lifts a built target, not "the middle," and largely fades once the options look like real products.
The 1% price statistic	Marn and Rosiello's arithmetic is correct: for the average of 2,463 companies, a 1% price gain beat a 1% volume gain, 11.1% vs 3.3%.	"A price increase always pays more than chasing volume." The number assumes zero volume loss, which almost never holds in a real market.
Value-based pricing	Directionally sound doctrine; McKinsey estimates a 5 to 10% return-on-sales improvement.	"Value-based pricing raises profit 31%." No auditable source for that figure exists in this report. The causal lift is confounded with product differentiation.
Annual discount depth	Most SaaS annual discounts land between 15% and 20%, per industry benchmarks.	"16.7% is the optimal discount." No experiment identifies it as profit-maximizing. It's a Schelling point that's easy to communicate ("12 months for the price of 10").
Freemium vs free trial	Each converts differently by funnel: freemium roughly 2.6-5.1%, opt-in trials roughly 8.9%, opt-out (card-required) trials roughly 31.4%, per 2026 benchmark data.	"Free trials beat freemium" or the reverse, stated as universal. The comparison depends on product, acquisition cost, and time-to-value, not a portable winner.
Discounting trains customers to wait	Confirmed over 8.25 years in one packaged-goods category: sustained promotion raised long-run price sensitivity.	"Discounting always erodes the brand" and "94% of promotions fail," stated with no channel or magnitude attached. Unsourced as stated.

To check whether an effect's condition matches your situation, ask three questions before you use it: is this a new offering with no prior reference price, is there a competing discount signal running at the same time, and does the underlying study's population (mail-order retail, packaged goods, SaaS benchmark) resemble your buyer at all. If you can't answer one of the three, run your own test instead of trusting the study's number.

Part 3. The pricing-decision cadence

Step	Move	Why it works (verified mechanic)
1	Model the price change against your own margin and elasticity before touching a live price.	Marn and Rosiello: price flows straight to profit with no cost attached, but only if volume holds, which the statistic assumes and your market won't automatically deliver.
2	If pricing a new item or new plan, test a 9-ending against a round number.	Anderson and Simester: three field experiments, real retailer, real customers, a $9 ending raised demand every time on new items.
3	Check the left-digit condition before setting any $X.99 price.	Thomas and Morwitz: the effect only fires when the dollar digit rolls down. Skip it where it doesn't.
4	If you use an anchor tier or a decoy, build it to lift one specific option, and A/B it rather than assume the lift.	Tversky and Kahneman confirm the anchoring mechanism; Yang and Lynn confirm the pricing-page magnitude is unverified and the effect is fragile with real products.
5	Price to value where your product is genuinely differentiated, not as a blanket policy.	Hinterhuber and McKinsey: directionally sound, 5-10% return-on-sales estimate, but the margin gain is confounded with the differentiation itself.
6	Set your annual discount as a round, communicable number, and test it against your own churn and cash preference, not the 16.7% convention.	OpenView benchmarks show the 15-20% band is convention, not an experimentally identified optimum.
7	Choose freemium or free trial based on your product's time-to-value and acquisition cost, not a universal claim about which converts better.	ProfitWell/Paddle and OpenView/ChartMogul benchmarks show conversion rates differ by funnel type and population, not by a portable ranking.
8	Before you scale any discount program, check whether it's training your buyers to wait.	Mela, Gupta, and Lehmann: 8.25 years of panel data show sustained promotion raises long-run price sensitivity, at least in one packaged-goods category.

Part 4. Measurement and the operating principle

Track three things, and track them from your own data, not from the study that inspired the tactic: realized margin after the price change, willingness-to-pay signals from your own buyers (what they say, what they actually pay, where they drop off), and elasticity measured from an actual price test, not assumed at zero.

None of that is glamorous. A pricing tactic that "always works" is the version of the finding with every condition removed, which is exactly the version that stops working the first time your context doesn't match the paper's.

The one-line operating principle: price from your own margin math and your own customers' willingness to pay, and only borrow a tactic when your situation matches the condition the study actually tested.

Method and sources

Grounded in peer-reviewed pricing and decision research: Anderson and Simester (9-endings, three field experiments, 2003), Thomas and Morwitz (left-digit effect and its boundary condition, 2005), Huber, Payne, and Puto (the original decoy effect, 1982) alongside Yang and Lynn's replication challenge (2014), Marn and Rosiello (the 1% price statistic and its volume-held-constant assumption, 1992), Tversky and Kahneman (anchoring, 1974), and Mela, Gupta, and Lehmann (the long-run promotion-sensitivity study, one packaged-goods category, 1997). Vendor and benchmark data (McKinsey, OpenView, ProfitWell/Paddle) is used for direction, named with its own sample, never blended into a single portable number.

Lab and field effect sizes do not transfer unconditionally to a specific market. Every tactic in this playbook is tied to the condition that made it true in the original study. Test it against your own prices before you trust it against your own margin.

Petrichor Projects | Category Gravity | petrichorgrowth.com/research