Query results half right after joining two tables - sql

Question

The following query resulted in correct results only for the inner query (post_engagement, website purchases) while all other numbers were incorrectly increased manyfold. Any ideas? Thanks.
Schema of the two tables:
Favorite_ads (id, campaign_id, campaign_name, objective, impressions, spend)
Actions (id, ads_id, action_type, value)
SELECT
f.campaign_id,
f.campaign_name,
f.objective,
SUM(f.impressions) AS Impressions,
SUM(f.spend) AS Spend,
SUM(a.post_engagement) AS "Post Engagement",
SUM(a.website_purchases) AS "Website Purchases"
FROM
favorite_ads f
LEFT JOIN (
SELECT
ads_id,
CASE WHEN action_type = 'post_engagement' THEN SUM(value) END AS
post_engagement,
CASE WHEN action_type = 'offsite_conversion.fb_pixel_purchase' THEN SUM(value) END AS website_purchases
FROM Actions a
GROUP BY ads_id, action_type
) a ON f.id = a.ads_id
WHERE date_trunc('month',f.date_start) = '2018-04-01 00:00:00' AND
date_trunc('month',f.date_stop) = '2018-04-01 00:00:00' --only get campaigns
that ran in April, 2018
GROUP BY f.campaign_id, campaign_name, objective
Order by campaign_id

Answer 1

Without knowing the actual table structure, constraints, dependencies and data, it's hard to tell, what the issue may be.
You already have some leads in the comments, which you should consider first.
For example you wrote, that this sub-query returns correct results:
SELECT ads_id,
CASE
WHEN action_type = 'post_engagement'
THEN SUM(value)
END AS post_engagement,
CASE
WHEN action_type = 'offsite_conversion.fb_pixel_purchase'
THEN SUM(value)
END AS website_purchases
FROM Actions a
GROUP BY ads_id, action_type
Is this one also giving correct results:
SELECT ads_id,
SUM(
CASE
WHEN action_type = 'post_engagement'
THEN value
END
) AS post_engagement,
SUM(
CASE
WHEN action_type = 'offsite_conversion.fb_pixel_purchase'
THEN value
END
) AS website_purchases
FROM Actions
GROUP BY ads_id
If so, then try replacing your sub-query with that one.
If you still have a problem, then I'd investigate if your join condition is correct, as it would seem, that for a campaign (campaign_id) you could probably have multiple entries with the same id, which will multiply the sub-query results - that depends on what is actually the primary key (or unique constraint) in the favorite_ads.