Calculating an "activation" percentage using SQL - mysql

Question

So I have the following SQL schema (http://sqlfiddle.com/#!2/b366c) and what I'm trying to achieve is the % of companies that I can consider activated.
In the schema, you can see there are the following tables
organisations (otherwise known as companies)
competitions
competitionmembers
activity_entries
What I would like to do is, figure out the % of companies in the organisations (i.e. total users) that create a competition (competitions table), invite at least another person (competitionmembers table) and have completed at least one activity (activity_entries table)
This may be too complex, but what I'd like to do is also create a funnel - to visualise where most companies drop off. For this, I understand I should create a seperate query for each of the steps and then just stack them to see the flow.
Using the sample data provided here (http://sqlfiddle.com/#!2/b366c) you can see that:
1. 4 companies have registered
2. 2 companies have created competitions
3. 1 company has a competition with at least 2 participants (not just the admin)
4. 1 company has registered at least one activity
So 25% of companies became "activated".
I would really appreciate some help in building these queries and visualising percentages!

Answer 1

Maybe not the most efficient way, but the intermediate results ought to be small enough for this not to matter overmuch.
You can run the inner queries on their own to look at the different results:
SELECT COUNT(oid) AS organizations,
SUM(IF(competitions > 0, 1, 0)) AS competing,
SUM(IF(activations > 0, 1, 0)) AS activated,
100.0*SUM(IF(activations > 0, 1, 0))/COUNT(oid) AS actpercent
FROM (
SELECT oid,
SUM(IF(cid IS NULL,0,1)) AS competitions,
SUM(IF(aid IS NULL,0,1)) AS activations
FROM (
SELECT
o.organisationId AS oid,
c.competitionId AS cid,
a.id AS aid
FROM organisations AS o
LEFT JOIN competitions c USING (organisationId)
LEFT JOIN activity_entries AS a USING (competitionId)
) AS situation GROUP BY oid
) AS summary;
First we get the situation list of all organizations, competitions and activities; here you may add a WHERE condition to filter organizations of interest, removed competitions and so on.
From this we get a summary of organizations with the number of competitions and activations for each. Each competition can only count for one if it's activated (i.e., if you get three competitions, one with three activities and two with zero, you will retrieve three as the number of competitions, one as the number of activations).
Then we just get the total count of organizations, and calculate the number of activations as a percentage.
The output of the above would be,
ORGANIZATIONS COMPETING ACTIVATED ACTPERCENT
4 2 1 25
Addition
lserni would it be possible to add one more layer to your query, which
is the "inviting" aspect. i.e. if there are more than 2 users in the
competitionmembers table for a competition?
In this case for each competition we need to know how many members there are in another table. So we have to act on the query where the competitionId is available, and we modify situation:
SELECT
o.organisationId AS oid,
c.competitionId AS cid,
a.id AS aid
FROM organisations AS o
LEFT JOIN competitions c USING (organisationId)
LEFT JOIN activity_entries AS a USING (competitionId)
We just add the necessary GROUP BY existing-columns and the new aggregate field, and of course the necessary LEFT JOIN:
SELECT
o.organisationId AS oid,
c.competitionId AS cid,
a.id AS aid,
COUNT(m.id) AS members
FROM organisations AS o
LEFT JOIN competitions c USING (organisationId)
LEFT JOIN activity_entries AS a USING (competitionId)
LEFT JOIN competitionmembers AS m ON (c.competitionId = m.competitionid)
GROUP BY oid, cid, aid;
(which I think illustrates one of the advantages of nested "serialized" queries - they're easier to maintain. That at least is my opinion. Maybe the truth it's just that I can't wrap myself around the more complicated, all-in-one queries...).
Now that we have members of competition, we look to the query immediately external to the one above:
SELECT oid,
SUM(IF(cid IS NULL,0,1)) AS competitions,
SUM(IF(aid IS NULL,0,1)) AS activations
FROM v_situation GROUP BY oid
By the way: you can simplify the writing of these queries by offloading them to VIEWs. CREATE VIEW v_situation AS SELECT o.organisationId AS oid, ... GROUP BY oid, cid, aid; and you have a virtual table v_situation that you can use wherever you would a table).
...and rewrite it adding the number of competitions with one member and those with more:
SELECT oid,
SUM(IF(cid IS NULL,0,1)) AS competitions,
SUM(IF(aid IS NULL,0,1)) AS activations,
SUM(IF(members > 1, 1, 0)) AS withmany,
SUM(IF(members = 1, 1, 0)) AS withone
FROM ( ... ) AS situation
GROUP BY oid;
Then you just need to decide what to do with that information. You can pass it through and re-select the withone field in the parent query, or you can calculate its percentage. Only in this case remember that the number of competitions may be zero, so you need to arm against the case when
activations_with_many_members / activations
has a zero at the denominator, using IF to change the formula to 0.0 if no activations are present:
IF(activations > 0, <percent formula>, 0.0 ) AS percent_with_many
Also, if you only wanted members wherever an activation is also present, you should do so in the definition of members, so that a member is counted only if its id is not null (we have a member) and the aid is not null (we have activation):
SUM(IF(a.id IS NOT NULL AND m.id IS NOT NULL,1, 0)) AS members

Answer 2

select 1/ count(organisations.organisationId) * 100 *
(select count(distinct(org.organisationId)) from organisations org
inner join competitions cmp on org.organisationId = cmp.organisationId
inner join competitionmembers cmpm on cmpm.competitionid = cmp.competitionid
inner join activity_entries act on act.competitionid = cmpm.competitionid) as pct
from organisations