We've covered the bowl, the seat, and the blend between them. Now we go one step downstream: the transition from the bowl into the port floor. This is where the air, having just negotiated the valve seat and short-side turn, meets the long, mostly-straight section of the runner. Get this transition wrong and you undo every gain from your seat work.

The bowl is a roughly spherical cavity directly under the valve. The port floor is the flatter, lower surface of the runner extending back toward the intake face. Between them sits a geometric discontinuity: the bowl curves in three dimensions, while the floor is essentially a 2D surface with mild curvature. Where they meet, you get a step, a hump, or a hollow depending on how the casting cores fit together at the factory.

The three defects you'll find on a stock head:

• Core shift step: The sand core forming the bowl was misaligned with the core forming the runner. You'll feel a ridge with your fingernail — often 0.030"–0.060" tall — right at the transition. Air separates off this step and forms a low-pressure wake that occupies port cross-section without flowing.
• Casting flash hump: A thin fin of iron or aluminum sticking into the port where the cores parted. Trips the boundary layer and creates turbulence at exactly the wrong spot.
• Sump hollow: A low spot in the floor just aft of the bowl where casting shrinkage pulled the surface down. Fuel puddles here on port-injected engines during transient throttle.

The blend technique: Using a carbide burr followed by cartridge rolls, you're not trying to enlarge the port — you're smoothing the transition so the streamline sees a continuous curve from bowl to floor. Aim for a radius roughly equal to 1/4 of the port height at that station. On a 1.5" tall port, that's a 0.375" blend radius.

Real-world example: The Ford 4.6L 2V PI heads have a notorious core-shift step about 3/4" downstream of the intake valve. Stock, they flow around 205 CFM at 0.500" lift. Just blending that transition — no seat work, no port enlargement, maybe 30 minutes per port with a rotary tool — picks up 8–12 CFM. That's pure "found money" because the metal you're removing was actively hurting flow.

Rule of thumb: If your fingernail catches on a ridge in the port, air is separating there. Boundary layer separation costs roughly 2x the flow area of the step itself because the wake behind the step blocks additional cross-section. A 0.050" step in a 1.5" port can effectively narrow it by 0.100"+.

Blend the floor before you touch anything else. It's the highest return-per-hour work in the entire port.