William Swelbar started as a flight attendant at North Central in 1979, months after deregulation, and by 1982 was leading a coalition of five unions attempting a $400 million employee buyout of Republic Airlines. He spent nine months on Wall Street raising the money. When Northwest acquired the company two years later, every Republic employee recovered every dollar they had given up in concessions, protected by common stock warrants and a liquidating preferred stock the coalition had negotiated. He now runs Swelbar-Zhong Consultancy and writes at Swelbar on Airlines.
Swelbar’s explanation of airline economics starts with network design. An airline serving ten cities point-to-point has ten city pairs to sell. Connect those cities through a hub and the same ten airports produce 55 sellable pairs. Add two more hubs and the figure reaches 230. The airports and aircraft may be largely the same, but the number of possible itineraries expands sharply. That is why Swelbar describes hub-and-spoke networks as a way to do more with less.
Southwest shows where the point-to-point model starts to break down. At 72 airports, a two-hour drive radius around its network covered about 95% of domestic demand. By roughly 2017, the airline had expanded to around 90 airports, but many of those routes still connected only one city to another. As the network grew, filling aircraft required connecting traffic. Southwest has since moved toward a hybrid model that combines point-to-point flying with more network connectivity.
The next question is how airlines segment the same aircraft. Delta and United now sell as many as five products within a single cabin, shifting competition beyond ticket price toward different levels of service and willingness to pay. Swelbar argues that fares alone have never covered the full cost of carrying passengers, so airlines need other revenue to support the model. Credit cards and loyalty programs now contribute about $14 billion a year across the industry, with Delta expecting its American Express partnership alone to generate $10 billion annually by 2030. Without card revenue, Swelbar estimates United and American would see margins fall from roughly 8-10% to below 3%.
For analytics leaders, the lesson is that structural choices determine how much optimization can achieve. Revenue management cannot fix a network with too few sellable routes, and segmentation only creates value if the pricing system can move customers across the tiers.
Our Recommended Read: Why Try AI
Why Try AI is a weekly newsletter that shows professionals how to use AI without jargon or hype. Get practical tools and workflows you can try today. Join 18,000+ readers who'd rather get things done with AI than just read about it.
Market Pulse
Carpe’s Minerva reasoning engine turns underwriting appetite into configurable rules that produce recommendations to quote, refer, or decline. It evaluates more than 200 business characteristics across 50 million U.S. business profiles and returns the evidence behind each decision. Carpe says it can reduce underwriting touches by up to 25% and save 30-45 minutes of research per submission. Uncertain cases are routed to human review, keeping the decision path auditable.
Project44’s Intelligent TMS debuted as a Leader in seven G2 Fall 2026 reports, including the Enterprise TMS Grid. The platform combines live rates, carrier performance, and predictive ETAs with 95%+ accuracy across a network of 282,000 carriers. Project44 says customers can cut freight costs 4%, improve on-time delivery 17%, and reduce manual work by up to 70%. Intelligent TMS revenue grew 57% year over year in FY27 Q2.
NielsenIQ is working with The OpenAI Deployment Company to embed its consumer and commerce intelligence into enterprise AI applications and workflows. NIQ’s Optiq Bridge will let companies use its data, models, and generative AI capabilities inside their own systems, while Optiq Chat turns business questions into recommendations grounded in NIQ data. The underlying data engine spans 160 petabytes, roughly 260 million product items and 10.5 billion product attributes. Expanded versions of both products are scheduled for early September.
QueryStory is building an agentic data platform that lets business users query messy enterprise data without relying on predefined schemas or data-team tickets. The system returns decision-ready briefs, decks, documents, or dashboards with the underlying sources, definitions, and assumptions attached. Outputs update as the source data changes, while preserving what was decided and why. The approach applies cybersecurity investigation techniques to broader enterprise decision workflows.
Resources and Events
📅 Gartner IT Symposium/Xpo 2026 (Barcelona, Spain - November 9-12, 2026)
Gartner’s four-day conference brings together more than 6,500 CIOs and IT executives, 130+ Gartner analysts, and 125+ technology providers. The 2026 program covers agentic AI, AI infrastructure, cybersecurity, emerging technologies, and operating models, with sessions on decision intelligence, multi-agent systems, responsible AI, and enterprise AI strategy. Details →
📅 Game-Changing AI Technology for Governments (Virtual - September 10, 2026)
SAS is hosting a webinar on how governments can use decision intelligence to combine data, analytics, AI, and behavioral science. The session covers how decision intelligence differs from traditional analytics, how it can improve planning and resource allocation, and what agencies need to consider when implementing a decision intelligence framework. Speakers include Jennifer Robinson, Global Strategic Advisor for Public Sector at SAS, and Lucas Ermino, a SAS systems engineer focused on fraud prevention and AI. Details →
📊 Report Spotlight: The State of AI in 2026 (McKinsey)
McKinsey surveyed 1,719 respondents across 97 countries on how organizations are scaling AI and translating adoption into business value. Enterprise-wide AI scaling rose from 38% to 44%, while 50% of respondents say AI is improving decision-making. Despite broader deployment, only 37% report any EBIT contribution from AI, and just 6% qualify as high performers. Among companies with more than $1 billion in revenue, 40% are now scaling AI agents, up from 27% a year earlier. Read →
The DecideWise Edge
Benjamin Baer, Founding Member of the DecideWise community, argues that AI has moved from a source of small efficiency gains to the main engine behind business strategy, and that this shift raises the cost of getting decision logic wrong. When a company automates a complex choice such as who is approved for a loan, a flaw in the logic does not affect one transaction. It repeats across every case the system handles, and discovering it in a live market can cost millions. The piece makes the case for decision simulation as the safeguard. Running encoded logic in a sandbox lets teams stress-test the underlying strategy, observe how the system behaves under pressure, and identify failure modes before anything reaches production. Read More →
For the Commute
Solving Urban Mobility Problems with Data (Decision Intelligence Lab)
Jon Petersen, Vice President of Data at Archer, explains how data science supports decisions across aircraft design, manufacturing, and future air-taxi operations. The discussion covers how Archer combines engineering data with demand and trip-density models, how data teams work directly with engineers, and why getting useful models into operational workflows is often harder than building them. Petersen also explains how he builds cross-functional data teams and makes the case for data science investment by tying projects to specific engineering and business decisions.
