Routes & Connections
Connection Risk Calculator
You land at a connecting airport and have to catch the next flight. Enter the inbound route and your layover to see how often that connection actually worked — the share of past arrivals that left enough time — from the U.S. DOT's on-time filings over the last 12 months.
JFK → ATL ✈
New York, NY → Atlanta, GA · 45-min layover · all airlines
of past JFK → ATL arrivals left enough time to make a 45-minute connection (slack +0 min after a 45-min MCT) · Jun 2025 – May 2026 · 4,171 filed arrivals
Slack = 45 − 45 = 0 min. The connection is made whenever the inbound arrives no more than 0 minutes late.
Informational only — a historical base rate for the inbound route, not a prediction for any booked itinerary, and a conservative floor (it ignores a late-departing outbound and an early inbound, both of which only help). On-time delay measured per 14 CFR § 234.2.
Methodology
The make-rate is read directly from the arrival-delay distribution of your inbound route in the U.S. Department of Transportation's Marketing Carrier On-Time Performance filing (Bureau of Transportation Statistics, table 236). For the origin → hub direction and window you choose, the tool sums the monthly arrival-delay histograms — exact for any window, because histograms add — and asks what share of those past arrivals were early enough.
- Slack = scheduled layover − minimum connection time (MCT). This is the cushion you have: how late the inbound can be and you still make the outbound.
- Connection made ⇔ inbound arrival delay ≤ slack. A flight's arrival delay uses the federal on-time measure (14 CFR § 234.2).
- Make-rate = the cumulative share of the inbound route's historical arrivals with delay ≤ slack — the CDF of the merged arrival-delay distribution at your slack. Delays are stored to a −60…+240 minute domain; a slack at or beyond +240 reads as 100%.
Why conservative, and its limits. The model counts a connection as made only if the inbound arrived within slack, ignoring that a late-departing outbound or an early inbound would save more connections — so the true make-rate is at least this. It assumes a same-terminal connection with no security or customs re-clear and no gate walk beyond your MCT; it can't see your booked itinerary's real MCT, terminal change, or bag re-tag, which is why the MCT is your input. Flights are attributed to the marketing carrier (the airline on the ticket). U.S. domestic operations of reporting marketing carriers; this tool serves Jun 2025 onward. Historical base rates only — no live tracking, no predictions.
Last reviewed: July 2026
Frequently asked questions
What is minimum connection time (MCT), and why do I set it?
MCT is the shortest gap an airline and airport will accept between a connecting arrival and departure — it is airline- and airport-published, not a federal figure, and runs roughly 30–90 minutes for domestic connections. There is no single correct number, so this tool does not bake one in: it defaults to a labelled 45 minutes and lets you change it to your own itinerary's MCT. Your slack is your layover minus this value.
Why is this called conservative — wouldn't the real make-rate be lower?
The opposite: the true make-rate is at least what this shows. The model only asks whether the inbound arrived within your slack, and ignores two things that only ever help you — an outbound that itself departs late, and an inbound that arrives early. It never counts a delay that a late outbound would have absorbed. So treat the figure as a floor, not a point estimate.
Why can't I enter my actual flight numbers?
Because the tool scores the route, not a booked itinerary. It reads the DOT's historical arrival-delay distribution for every past flight on your inbound route and asks how many would have made your connection — a base rate over months of data, not a prediction for one flight. It also can't see your real booked MCT, terminal change, or bag re-tag, which is why you set the connection time yourself.
What exactly counts as the connection being "made"?
The inbound flight arrives at the gate no more than your slack (layover minus MCT) after its scheduled arrival. Arrival delay is measured by the federal on-time rule (14 CFR § 234.2). The model assumes a same-terminal connection with no security or customs re-clear and no gate-distance walk beyond the MCT you entered — all of which push the real-world result the other way.