with t as (
	select '1000' UserID, convert(datetime, '2019-01-01') dt
	union all
	select '1000' UserID, convert(datetime, '2019-01-02') dt
	union all
	select '1000' UserID, convert(datetime, '2019-01-03') dt
	union all
	select '1000' UserID, convert(datetime, '2019-01-04') dt
	union all
	select '1002' UserID, convert(datetime, '2018-12-03') dt
	union all
	select '1002' UserID, convert(datetime, '2018-12-20') dt
	union all
	select '1003' UserID, convert(datetime, '2018-12-01') dt
	union all
	select '1003' UserID, convert(datetime, '2018-12-11') dt
)
select UserID, Cnt, datediff(day, PreDt, dt) df
from (
	select t.*
		,lag(dt, 1, null) over(partition by UserID order by dt) PreDt
		,ROW_NUMBER() over(partition by UserID order by dt) Cnt
	from t
) lst
where Cnt > 1
;
 

 lag 함수를 이용하면 될 것으로 보입니다.

위 쿼리 결과로 각 고객별 몇번째 구매인지(Cnt)와 재구매 까지 날 수(df)를 알 수 있습니다.

 박인호님 소중한 답변 정말 감사합니다 ! :) 

정말 최고입니다!!!

3번째 구매의 재구매 일수 구하는 기준일자가
첫번째 일자인지? 두번째 일자인지? 에 따라서
MIN 을 사용할지 LAG 를 사용할지 정하시면 됩니다.

WITH t AS
(
SELECT 1000 UserID, '2019-01-01' dt
UNION ALL SELECT 1000, '2019-01-02'
UNION ALL SELECT 1000, '2019-01-03'
UNION ALL SELECT 1000, '2019-01-04'
UNION ALL SELECT 1002, '2018-12-03'
UNION ALL SELECT 1002, '2018-12-20'
UNION ALL SELECT 1003, '2018-12-01'
UNION ALL SELECT 1003, '2018-12-11'
)
SELECT *
  FROM (SELECT UserID
             , ROW_NUMBER() OVER(PARTITION BY UserID ORDER BY dt) rn
             , DATEDIFF(d, MIN(dt) OVER(PARTITION BY UserID), dt) df
          FROM t
        ) a
 PIVOT (COUNT(UserID)  FOR rn IN ("2", "3", "4")) a
 WHERE df != 0
;

 마농님도 답변 감사합니다. 복 많이 받으세요~