Data analytics training on your own data

Data analytics training gives your team the skill to change and check its own reports, so reports stop depending on one person. Courses in Power BI, SQL and GA4 run on your own tables and end with a task each person does alone, reviewed in writing.

Last reviewed

One exercise from the SQL course, example

The join reported net sales of £670. The right figure is £470.

Example data
orders (6 rows)
order_idorder_dateorder_value
10012026-10-03120.00
10022026-10-0785.00
10032026-10-12200.00
10042026-10-1860.00
10052026-10-24150.00
10062026-10-2995.00
refunds (4 rows)
refund_idorder_idrefund_daterefund_amount
500110022026-10-1020.00
500210032026-10-1540.00
500310032026-10-2130.00
500410052026-10-28150.00

Before

SELECT
     SUM(o.order_value) AS gross_sales
    ,SUM(COALESCE(r.refund_amount, 0)) AS refunds
    ,SUM(o.order_value) - SUM(COALESCE(r.refund_amount, 0)) AS net_sales
    ,COUNT(*) AS row_count
FROM orders AS o
LEFT JOIN refunds AS r ON 1=1
    AND r.order_id = o.order_id
WHERE 1=1
    AND o.order_date >= '2026-10-01'
    AND o.order_date < '2026-11-01'
Result
gross_salesrefundsnet_salesrow_count
910.00240.00670.007
6 orders in, 7 rows out: order 1003 has two refunds, so it appears twice and its £200 is counted twice.

After

WITH refunds_by_order AS (
    SELECT
         order_id
        ,SUM(refund_amount) AS refund_amount
    FROM refunds
    GROUP BY order_id
)
SELECT
     SUM(o.order_value) AS gross_sales
    ,SUM(COALESCE(r.refund_amount, 0)) AS refunds
    ,SUM(o.order_value) - SUM(COALESCE(r.refund_amount, 0)) AS net_sales
    ,COUNT(*) AS row_count
FROM orders AS o
LEFT JOIN refunds_by_order AS r ON 1=1
    AND r.order_id = o.order_id
WHERE 1=1
    AND o.order_date >= '2026-10-01'
    AND o.order_date < '2026-11-01'
Result
gross_salesrefundsnet_salesrow_count
710.00240.00470.006
Refunds summed to one row per order first: 4 refund rows become 3, and the join returns 6 rows, one per order.
Example data: six invented October orders and four refunds, with net sales as order value less refunds against those orders, by order date. The first query joins every refund to its order, so order 1003 comes back twice and gross sales read £910. Summing refunds per order before the join gives one row per order, gross sales of £710 and net sales of £470.

What this fixes

One person changes the reportEveryone else only reads it, so when that person is away, the questions wait.
The total doubled and nobody knows whyA query copied from an old file joins one table too many, and the figure looks plausible enough to ship.
GA4 and the order system disagreeLast month's purchases differ between the two, and no one on the team can say which to report or why.
Everything ends up in ExcelAny question the report doesn't answer becomes an export, a pivot table and a figure only its maker can rebuild.

The four courses: Power BI, SQL and GA4 training

Power BI for report buildersIt's for the people who maintain your reports. They learn to write a DAX measure with a description and publish a page that uses it, and the final task adds one measure from your backlog to your own model.
SQL for analystsIt's for people who already pull data from the warehouse. They learn to join tables without multiplying rows, and the final task answers one backlog question on your own tables, with the row count beside each join.
GA4 for marketersIt's for people who read GA4 and have to explain it. They learn to build explorations and say why GA4 purchases and real orders differ, and the final task reconciles last month's purchases to your order system.
Reading the numbersIt's for managers who decide from reports they don't build. They learn to ask for any figure's definition, source table and owner, and the final task writes up one figure's formula, as in the KPI library, and traces it to its table.
The materialsEvery exercise, the copy of your tables it runs on and the worked answers stay in your drive after the course.

How data analytics training runs

  1. 1

    Scoping

    Scoping settles the course, who attends and which reports they look after, and your team's backlog of unanswered questions supplies the final tasks.

  2. 2

    Build

    The exercises are written against a copy of your tables in your own environment, with personal details removed.

  3. 3

    Sessions

    Each exercise runs on a report or table people will open again the next day, with mistakes like the one in the exhibit left in for them to find.

  4. 4

    Final task and handover

    Each person does the final task alone, and the written review goes to them and to you.

Why skills from training fade

Practising on someone else's dataA sample dataset has clean tables with one row per order, and your refunds table doesn't, so the exercises are written on a copy of your own tables, mess included.
Nothing to use it onPeople go back to their usual work and the first chance to use the course comes later, so the final task comes from your backlog.
A join that looks rightA plausible total goes unchecked, so from the first exercise people learn to write down the row count before and after every join and stop when it grows.

In the example, both queries return refunds of £240, so a check on refunds alone would pass, and the error is in gross sales.

Training by industry

EcommerceThe SQL and GA4 exercises run on orders, refunds and purchases, like the exhibit above.
Subscription and mediaThe exercises use sign-ups, renewals and cancellations, where one subscriber has many rows.
Financial servicesThe Power BI exercises use account or policy tables, copied with customer details removed before anything is written.

If the GA4 gap is broken tags, fixing them is tracking work. To have the model built first and the team trained on it after, use the Power BI consultant service. For a standing analytics function once the course ends, see fractional analytics.

Questions about training

Do you run Power BI training for finance teams?

Yes. The exercises use finance's own model, or a copy of it, and the final task adds a measure finance has asked for and checks it against the ledger. The finance dashboard in the KPI library shows the kind of pack it ends up in.

Can the training use our real data?

Yes, as a copy in your own environment, with names, emails and anything else personal removed. The copy keeps the awkward parts, such as orders with two refunds, because those are what the exercises are about.

Do you run SQL training for ecommerce teams?

Yes. The SQL course runs on your orders, refunds and order lines, and the exercise above comes from it. People learn to check the row count at each join, so a refund can't double a total without someone seeing it.

Is the Google Analytics training for marketers or developers?

It's for marketers who read the reports and have to explain them. The course ends with last month's GA4 purchases reconciled to your order system, and setting up tags is work for the tracking service.

How do we know the training worked?

Each person's final task is reviewed in writing, and you get the reviews. The task comes from your team's own backlog, so you see the result in a report you already use.

Is there a report only one person can change?

On the call we look at that report and who reads it, then suggest a course and the tables its exercises would use.

Book a call

Free 30 min call