103A-Real-Time Pricing: What It Takes, What It’s Worth, and Where It Breaks

Industries with a perishable seat, room, or ticket eventually discover real-time pricing. Airlines found it first, in 1978. Everyone else is still catching up, and catching up badly in a lot of cases. This blog lays out what real-time pricing is, what it takes to run, what it’s realistically worth in revenue and margin, and — the part pricing vendors leave out — where it has blown up in companies’ faces. The short version: it works, it’s not free, and the return is a lot smaller than the case studies you’ve been shown, and it risks your relationship with the customer if it’s done without sensitivity.

Posted August 2026

What Real-Time Pricing Is

Real-time pricing means the price of a unit of Capacity (Pattern: The total unit volume a facility, or group of facilities, can produce annually at its highest practical operating mode) — a seat, a room, a ride, a SKU — moves continuously, driven by current demand, remaining inventory, competitor prices, and time until the sale window closes. It is not the same thing as a seasonal price, and it is not the same thing as an Optional Component of the Price (Pattern: An add-on element, fees, bonuses, terms, or performance charges, that changes the net cash-equivalent price paid), which is a bolt-on fee layered over a fixed base price. Real-time pricing changes the base price itself, and it can change it multiple times a day.

Airlines invented the modern version after deregulation in 1978, when they discovered that a seat sold today at a discount and a seat sold tomorrow at full fare are the same seat priced two different ways to two different Segments (Pattern: Groups of customers who share similar buying or usage characteristics). Everything that followed — hotels, ride-share, ticketing, retail, ski resorts — is a variation on that same idea: price the unit of capacity to the customer standing in front of it right now, not to some average customer from a price list.

What It Takes to Implement and Manage It

Real-time pricing is not a switch you flip. It’s an operating capability with five parts, and skipping any of them is how a pilot algorithm project turns into a mess.

• Data infrastructure — live feeds on demand, remaining inventory, competitor prices, and booking pace. Without clean, current data, the algorithm is guessing.
• A demand model — usually machine learning included — that estimates how price moves sales at a given moment, built and validated on real transaction history before it touches a live price.
• Guardrails — hard minimum and maximum bounds set by humans, not the algorithm. Every real-time pricing failure below has a guardrail that was set too loose or wasn’t set at all.
• Systems integration — the price has to actually reach the point of sale, in real time, across every channel. This is the unglamorous part, and it’s usually the expensive part, because it has to run inside whatever legacy booking or POS system the company already has.
• A feedback loop and a human override — someone watching results, adjusting the model, and empowered to pull the price back when it’s about to become a headline.

McKinsey and Company has studied real-time pricing. McKinsey’s client work makes the sequencing point clearly: don’t roll this out cold. Pilot it in a handful of categories, prove the model, get the frontline pricing team who has to live with the tool involved in building it — then scale. Retailers who skipped the pilot and went straight to a full rollout are the ones McKinsey describes as running “half-hearted and poorly planned pilots that, unsurprisingly, had little impact.” The infrastructure is necessary. It is not sufficient. The organization has to actually trust and use the tool, or it sits there unused.

How Much Improvement Should You Actually Expect

This is the number pricing vendors pitch may inflate, so it’s worth being precise about it.

The best-corroborated, non-vendor benchmark — McKinsey’s client work across a large set of retailers — puts mature dynamic pricing benefits at 2 to 5% sales growth and 5 to 10% margin improvement, with individual pilot categories occasionally showing simultaneous revenue and margin gains around 3%. Airlines, the most mature practitioners, see gains in a similar 2 to 7% range from standard revenue management, with McKinsey’s 2024 estimate of up to $45 billion in industry-wide value over five years from modern retailing techniques implying roughly 2 to 3% revenue uplift per airline, or as much as 15% of EBITDA. Hospitality figures in the 2 to 7% revenue and 15 to 30% RevPAR range circulate widely in the trade press attributed to McKinsey, but I did not review a specific McKinsey report backing that exact number.

Two things are worth flagging about that range. First, the margin number is usually bigger than the revenue number, in percentage-point terms, because pricing is a direct pass-through to profit — McKinsey’s own classic finding, on the average income statement of an S&P 500 company, is that a 1% price increase at stable volume yields roughly an 8% increase in operating profit, nearly 50% more than the impact of an equivalent 1% cut in variable costs. That’s the real reason pricing gets this much executive attention relative to its apparent size: the leverage is real.

Second, ignore the 20 to 40% revenue-lift numbers that show up in vendor case studies. They’re real numbers from real (favorable, cherry-picked, unaudited) engagements, not a planning benchmark. Plan against the McKinsey range. Anything promising more than that from a single vendor is describing a best case, not a base case.

Where Real-Time Pricing Has Succeeded

Success clusters in a specific set of conditions: perishable Capacity (an empty seat or room is revenue lost forever, not revenue deferred), genuinely variable demand, and a customer base with some ability to shift when or how it buys.

• Airlines — the original and still the deepest implementation. Nearly five decades of continuous refinement, from simple booking-class buckets to today’s continuous, AI-set fares.
• Hotels — RevPAR management is now table stakes at every major chain; Marriott, Hilton, and Accor have run mature systems since the 2000s.
• Ride-share — Uber’s surge pricing does what it’s designed to do: pull more drivers onto the road exactly when riders need them, functioning as a real-time supply signal as much as a revenue lever.
• E-commerce — Amazon reprices its catalog continuously; this is now standard practice across large online retailers, not a novelty.
• Ski resorts — with a caveat covered in Blog 103‘s most recent update: real-time walk-up pricing works narrowly, on a shrinking single-day-ticket minority of total skier visits, while the season pass — a fixed, pre-sold price — does most of the actual revenue and demand-heavy lifting.

The common thread: in every one of these, the company controls a large pool of comparable, perishable units and sells enough of them that the algorithm has real data to learn from. Small-volume, infrequent-purchase businesses don’t have enough transactions to make the model worth building.

Where Real-Time Pricing Has Failed or Disappointed

The failures cluster too, and they cluster around the same root cause: the price moved further or faster than the Segment being charged was willing to accept as fair, and the company had no story ready for why.

• Coca-Cola (1999) — piloted vending machines that raised prices in hot weather. The idea was framed internally as efficient allocation; the public read it as gouging on a hot day, Pepsi publicly distanced itself, and Coke shelved the plan.
• Wendy’s (2024) — announced menu “surge pricing,” drew an immediate and sharp backlash, and reversed course within days. The press comparison to Coca-Cola’s 1999 episode was immediate and accurate.
• Ticketmaster (2024) — dynamic pricing on the Oasis reunion tour saw prices triple mid-checkout, triggering a UK regulatory investigation and lasting reputational damage to an already-unpopular company.
• Uber — surge pricing during emergencies (storms, attacks) repeatedly generates gouging accusations, even though the underlying mechanism is identical to ordinary rush-hour surge. Context, not mechanics, is what turns the same algorithm into a scandal.
• Ski resorts, current — walk-up lift ticket prices pushed past $300–350 at destination resorts, alienating the low-frequency skiers the industry needs, and an August 2026 antitrust suit alleges several operators coordinated pricing through a shared ticketing platform. Vail is already reversing course with steep advance-purchase discounts — a company admitting its own pricing went too far. (See Blog 103.)
• Retailers running “half-hearted” pilots — McKinsey’s finding: pilots built without frontline buy-in or a clear category strategy produce “little impact” and fail to win the organization’s confidence, which kills the wider rollout before it starts. This isn’t a headline failure, but it’s the most common one — quiet underperformance rather than a scandal.

Every failure above shares a pattern: the company treated real-time pricing as a pure optimization problem and forgot it was also a Reliability (Pattern: The consistency with which a company delivers on its promises: reliability of delivery, of function, and of market presence) and trust problem. A price that moves in a way the customer can’t explain to themselves reads as exploitation, not efficiency — no matter how sound the underlying math is.

Who’s Actually Running It — And Who Isn’t

There’s a question sitting underneath all of the above that matters more than it first appears: is real-time pricing available to every competitor in these industries, or only to the largest ones? The answer, industry by industry, is that it’s mostly restricted to the top tier — and that restriction is itself a competitive weapon. It is a good example of economies of scale.

• Airlines — Delta, United, and American run in-house AI-driven revenue management; Southwest, Frontier, and Spirit lease shallower, off-the-shelf pricing engines from vendors. A former airline captain now teaching at Georgetown put it plainly to PBS: dynamic pricing has taken away one of the last structural advantages low-cost carriers had, because the big network carriers can now use pricing algorithms to match or beat budget fares selectively while protecting margin everywhere else. Spirit’s 2025 shutdown sits downstream of that shift.
• Ski resorts — real-time walk-up dynamic pricing exists almost nowhere outside Vail Resorts and Alterra, who between them dominate the destination-resort segment and share a pricing platform now under antitrust scrutiny. Most independent ski areas still sell a fixed day-rate ticket. (See Blog 103.)
• Ride-share — Uber and Lyft’s surge algorithms are proprietary; there is no vendor-tool equivalent for a traditional taxi fleet to lease.
• Grocery — electronic shelf labels and the pricing systems behind them are a real capital outlay; Walmart, Kroger, Whole Foods, and Lidl are rolling them out, independents mostly are not.
• Hotels and e-commerce are the partial exceptions. Cloud revenue-management tools have genuinely lowered the entry cost for independent hotels and small online sellers. But even here, the large chains keep the deeper edge: their algorithms train on their own enormous transaction history, and a purchased tool can’t manufacture data a small property doesn’t have. Scale buys a better-fed model even when it doesn’t buy exclusive access to the software.

That’s a genuine economies of scale cost advantage compounding for a large Standard Leader (Pattern: The competitor(s) or product that sets the market’s standard performance and price, selling more than half its volume at the industry’s most common price point): the fixed cost of building and running a real-time pricing system is largely flat regardless of volume, so it dilutes to near nothing per unit for a Delta or a Walmart and stays prohibitive for a small competitor. A Price Leader (Pattern: A competitor or product offering below-standard performance at a very low price, selling more than half its volume below the Standard Leader’s price) has an additional implementation problem: its whole identity is built on simple, predictable low pricing, so adopting the same tool the majors use risks undercutting the one thing that made it a Price Leader in the first place. Southwest’s long resistance to fencing and algorithmic fare segmentation, and its recent partial retreat from that position, is a clear illustration of the bind.

Who Pays For It

If the capability is concentrated at the top, the next question is who ends up funding the advantage. Sorted by Customer Sizes (VL, L, M, S) (Pattern: The segmentation of customers by rank-ordered total purchase volume: Very Large (50%), Large (30%), Medium (15%), and Small (5%)), the pattern is consistent: the largest, most powerful customers buy their way out of real-time pricing, and the smallest customers absorb it almost undiluted.

• Very Large customers (roughly 5% of customers, 50% of volume) — largely insulated. This is the segment with enough leverage to negotiate a fixed arrangement instead: the season pass holder, the corporate travel account with a negotiated fare discount, the bulk-contract retail buyer.
• Large customers (roughly 15% of customers, 30% of volume) — mostly insulated too, through loyalty-tier pricing and advance-purchase programs, at a smaller scale than VL.
• Medium customers (roughly 25% of customers, 15% of volume) — mixed exposure. Enough frequency to sometimes access advance-purchase pricing, not enough leverage to negotiate a contract. This is the segment resorts and airlines are now scrambling to build new products for, having realized they got caught by pricing built for the walk-up buyer.
• Small customers (roughly 55% of customers, 5% of volume) — bear the real-time price almost in full. The walk-up ski ticket buyer, the leisure traveler with no corporate contract, the limited-items grocery shopper, the occasional rider. No volume to negotiate with, no advance-purchase habit, and the least ability to notice or shop around a price that just moved against them.

Airlines complicate this picture slightly: the single worst-priced individual transaction is often a late-booking business traveler who works for a Very Large corporate account, and that account’s negotiated discount only partially offsets the fare volatility the traveler experiences day to day. But the broader shape holds across ski, grocery, and hotels — real-time pricing is, in practice, a mechanism that extracts more from the customers with the least power to resist it, while the suppliers running it capture a cost advantage on top of a revenue advantage. Scale wins twice.

The Bottom Line

Real-time pricing is a real, durable tool with a real payoff — call it low-single-digit revenue growth and mid-single-to-low-double-digit margin growth for a well-run program, not the 20–40% a vendor deck will claim. It requires real infrastructure and real organizational discipline to run well, and that infrastructure is expensive enough that it’s mostly a large-competitor’s tool — reinforcing the scale advantage of the large Standard Leaders who can afford it, at the expense of Price Leaders and smaller competitors who mostly can’t. And on the customer side, it fails, reliably and publicly, in exactly one circumstance: when the company lets the algorithm outrun what its least powerful customers — the Small, walk-up, one-off buyers who have no way to negotiate around it — consider a fair deal. The companies that have run it longest — airlines, hotels — didn’t get there by moving fast. They got there by moving the price just far enough, just often enough, to stay ahead of the backlash rather than behind it, while quietly building the case that scale itself has become one more source of competitive advantage.

***