ct.sql

CREATE TABLE calls (
 customer_id INTEGER,
 called_at TIMESTAMP,
 SUCCESS_RESULT VARCHAR(2) NULL,
 ERROR_RESULT INTEGER NULL
 );

data.sql

INSERT INTO calls VALUES (1, '2021-01-01', null, 3);
INSERT INTO calls VALUES (1, '2021-01-02', 'E', null);
INSERT INTO calls VALUES (1, '2021-01-03', null, 2);
INSERT INTO calls VALUES (1, '2021-01-04', 'S', null);
INSERT INTO calls VALUES (2, '2021-01-05', null, 8);
INSERT INTO calls VALUES (2, '2021-01-06', 'X', null);
INSERT INTO calls VALUES (2, '2021-01-07', null, 7);
INSERT INTO calls VALUES (2, '2021-01-08', 'Y', null);
INSERT INTO calls VALUES (1, '2021-01-09', null, 10);
INSERT INTO calls VALUES (3, '2021-01-03', 'Z', null);
INSERT INTO calls VALUES (4, '2021-01-01', null, 3);

query.sql

WITH CTE
AS 
    (
        SELECT
            customer_id,
            called_at,
            CASE
                WHEN success_result ISNULL THEN error_result::CHAR
                WHEN error_result ISNULL THEN success_result
            END AS result,
            call_num
        FROM
        (
            SELECT
                *, ROW_NUMBER() OVER (PARTITION BY customer_id ORDER BY called_at DESC) as call_num
            FROM
                calls
        ) callresults
        WHERE call_num <= 2
    )
SELECT
    customer_id, "1", "2", "3"
FROM (
    SELECT customer_id, result,
    ROW_NUMBER() OVER (PARTITION BY customer_id ORDER BY call_num desc) AS rn
    FROM CTE
) eml
PIVOT (
  min(result)
  FOR rn IN ([1],[2],[3])
) P