The Smith chart is less intimidating once you treat it as a map with a few fixed "roads". Here is the workflow I use for a lumped L-network match at a single frequency.

1. Normalize. Divide the load impedance by Z0 (usually 50 ohm). A load of 25 - j15 ohm becomes 0.5 - j0.3.

2. Know the moves.
- Series inductor: move clockwise along a constant-resistance circle.
- Series capacitor: counter-clockwise along a constant-resistance circle.
- Shunt capacitor: clockwise along a constant-conductance circle.
- Shunt inductor: counter-clockwise along a constant-conductance circle.
Adding any reactive element moves you along a circle; it never changes the real part you are sitting on (R for series, G for shunt).

3. Find the intersection. For the match you need to land on the r = 1 circle (if your last element is series) or the g = 1 circle (if it is shunt), then use the last element to cancel the remaining reactance and reach the centre.

4. Pick the topology with the least Q. There are usually two solutions. Prefer the one with a smaller path length, because lower loaded Q means wider bandwidth and less sensitivity to component tolerance.

5. Convert to component values. Reactance X = 2*pi*f*L for inductors and X = -1/(2*pi*f*C) for capacitors; remember to de-normalize by Z0.

Practical notes:
- Model real parts. A 0402 capacitor has self-resonance and ESL; at a few GHz a "10 pF" part can look quite different from 10 pF. Use the vendor S-parameter files.
- Include pad and trace parasitics, or at least simulate with your actual layout.
- For broadband matches, use multiple sections and keep each step's Q low (constant-Q contours on the chart help visualize this).

Once the moves are in your fingers, you can sanity-check a simulator's optimizer result in seconds.