How many fields should a checkout have?

As few as your fulfillment actually needs. Most Shopify checkouts can drop company name, second address line, and optional phone without losing anything, and each removed field removes a chance to mistype on a phone.

Does inline validation really matter?

Yes. Telling the shopper about an error the moment they leave the field, in plain words, prevents the pile-up of errors at submit that causes the most abandonment.

Where does field friction actually cost you checkout conversions?

Most checkout drop-off blamed on price is really friction in the form: too many fields, fields that fight the mobile keyboard, and validation that punishes typos instead of fixing them. Cutting a checkout from twelve fields to seven reliably beats another round of button color tests.

Count the fields, then cut

Every field is a small decision and a chance to mistype on a phone. Audit your checkout field by field and ask what each one earns. Company name, second address line, and phone number (unless you need it for delivery updates) are the usual candidates for removal, or for moving behind a toggle that most shoppers never open.

Match the keyboard to the field

Email fields should open the email keyboard, phone fields the numeric pad, postal codes the layout for the shopper's country. This is a one-attribute change per field and it removes a surprising amount of fumbling. Test it on a real phone, not just a desktop device emulator, because the emulator lies about how keyboards behave.

Validate inline, not at submit

Waiting until the shopper taps pay to reveal four errors is the worst version of validation. Check each field as the shopper leaves it, say exactly what is wrong in plain words, and accept common formats (spaces in card numbers, lowercase emails) instead of rejecting them. Forgiving inputs convert better than strict ones.

Measure the fix

Track field-level abandonment: which field was last touched before sessions died. After a change, compare completion rate for started checkouts week over week, holding traffic mix roughly constant. If the number moves, keep the change. If not, revert it and try the next field on the list.

Should I remove the phone field?: If you do not use it for delivery updates or fraud review, yes. If you do need it, make it clearly optional and say why you are asking.
What about address autocomplete?: Autocomplete shortens the form and reduces typos at once. It is one of the highest-leverage single changes most checkouts can make.