WITH
users AS (
SELECT
* EXCEPT (timestamp), timestamp as registrationTimestamp
FROM
`logs.registeredUsers` WHERE schedule != "original" AND schedule != "aggressiveStart" AND timestamp >= TIMESTAMP("2021-04-19")),
initialReviews AS (
SELECT
DISTINCT userID,
registrationTimestamp,
schedule,
cardID
FROM
(SELECT * EXCEPT (reviewCount), RANK() OVER (PARTITION BY userID, cardID ORDER BY timestamp) AS reviewCount FROM `logs.reviews`)
JOIN
`users`
USING
(userID)
WHERE
sessionID IS NULL AND reviewCount = 1 AND reviewMarking = "forgotten" AND
cardID IN(
SELECT
cardID
FROM
`logs.latestEssaysCards`
WHERE
essayName="qcvc")
),
subsequentReviews AS (
SELECT
*,
RANK() OVER (PARTITION BY userID, cardID ORDER BY timestamp) AS realReviewCount
FROM
`logs.reviews`
JOIN
initialReviews
USING
(userID,
cardID)
WHERE
sessionID IS NOT NULL AND isRetry IS NOT TRUE),
firstSession AS (
SELECT userID, cardID FROM subsequentReviews WHERE realReviewCount = 1 AND beforeInterval=1000*60*60*24
),
validUsers AS (SELECT userID FROM `logs.reviews` WHERE beforeInterval > 1000*60*60*24*5 AND cardID IN (SELECT cardID FROM `logs.latestEssaysCards` WHERE essayName="qcvc") AND sessionID IS NOT NULL GROUP BY userID HAVING COUNT(DISTINCT cardID) >= 103),
firstFollowup AS (SELECT userID, cardID, MIN(realReviewCount) AS realReviewCount FROM subsequentReviews JOIN firstSession USING (userID, cardID) WHERE beforeInterval > 1000*60*60*24 GROUP BY userID, cardID),
followupSession AS (SELECT userID, cardID, beforeInterval, realReviewCount, reviewMarking FROM subsequentReviews AS s JOIN firstFollowup USING (userID, cardID, realReviewCount) JOIN validUsers USING (userID))
SELECT
followupSession.beforeInterval / (1000*60*60*24) AS days,
COUNT(DISTINCT userID) AS N,
COUNT(*) AS promptN,
COUNTIF(reviewMarking="remembered") / COUNT(*) AS accuracy
FROM
followupSession
GROUP BY
days
ORDER BY
days