A drill at the end of every module
Not a quiz. A real task with a messy file: nine planted problems to find, eight business questions to answer in SQL, a dashboard to build and then defend.
Taught by Deola
Anyone can watch a tutorial. You came here to get good.
Four courses in data analysis, AI operations, virtual assistance and customer support, built backwards from what employers and clients actually ask for, and taught through work you do rather than videos you nod along to.
First three modules free · No card needed
Why this exists
I teach the way I wish I'd been taught: one idea at a time, with a file open in front of you. Every lesson ends with something you make, not something you watch.
The running order isn't my opinion. I took an analysis of 328 real data analyst job descriptions and taught the most-demanded skills first, so if you stop halfway you've still learned the things you're most likely to be asked about in an interview.
Everything lives in one place, lessons, exercise files, and the worked solutions, so you can come back to a technique two months later and find it in under a minute.
Excel81%of analyst job ads
SQL60%of analyst job ads
Power BI43%of analyst job ads
Python41%of analyst job ads
Communication showed up in 27%, more than three times as often as data visualisation. That's why it gets a module of its own rather than a closing lesson.
What you get
Built to be used, not just finished.
One dataset, all the way through
The file you clean in module 02 is the one you query in module 03 and visualise in module 06. You build one thing across the course, not ten disconnected exercises.
A checklist you keep
Each module closes with a one-page reference, the steps, the shortcuts, the mistakes to watch for. Print it, pin it, use it at work.
Three paces, one spine
Twelve weeks with primers if you're starting cold, six to eight self-paced if you're working, or an eight-week cohort using your own team's files. Nobody gets a cut-down version.
The courses
Four courses. All built backwards from the job.
Each stands on its own, and none assumes you've done the others. Start with the analyst course if you want the skills employers list, the AI course if you already have a job and a process that's eating your week, the virtual assistance course if you're building a client-services business from nothing, or the customer support course if you'd rather start with the queue almost every small brand needs covered.
No prerequisites
How to Be a Top-Notch Data Analyst
Excel, SQL, Power BI and Python, taught in the order employers ask for them, inside the loop an analyst actually runs: question, data, answer, decision.
10 modules58 lessons~11h+ capstone
View the full curriculum →
No coding required
AI for Business Operations
Gartner expects over 40% of agentic AI projects to be cancelled by 2027. This course is about not being in that 40%: pick the right process, take a baseline, put the human where the risk is, prove the saving.
8 modules42 lessons~7h 30m+ capstone
View the full curriculum →
No prerequisites
The Assistant They Can't Replace
Virtual assistance built for what AI has already automated and what it hasn't. The inbox, the calendar, the files, research, and running it as a real business — through one running client, start to finish.
10 modules50 lessons~10h+ capstone
View the full curriculum →
No prerequisites
The Ticket That Doesn't Come Back
Customer support built around the judgement AI can't fake: reading a policy, de-escalating a real complaint, and knowing when to bend the rules and when not to — through one running client, start to finish.
10 modules50 lessons~10h+ capstone
View the full curriculum →
Inside a course
Nothing hidden. Here's the whole thing.
The full outline, every module, every lesson, and the drill that closes it. Open one and see exactly what you'd be paying for.
Every lesson below is written out in full and live — this isn't a placeholder outline.
Open module 1 to start →
Every lesson below is written out in full and live — this isn't a placeholder outline.
Open the full course overview →
Every lesson below is written out in full and live — this isn't a placeholder outline.
Open the full course overview →
01 What the job actually is
1.1What a company is paying an analyst to do
1.2The four questions every analysis answers: what happened, why, what next, what if
1.3The request behind the request: how to run an intake conversation
1.4Scoping: what you will and will not answer, agreed in writing
DrillFive vague requests from real managers. Rewrite each as a question with a defined metric, a time window, and a decision it will inform.
Open module 1 →02 Excel that holds up under pressure
2.1Structuring a sheet so formulas do not break when data grows
2.2Lookups that survive an inserted column: XLOOKUP and INDEX/MATCH
2.3Conditional aggregation: SUMIFS, COUNTIFS, and multi-criteria logic
2.4Pivot tables aimed at a question, not at the data
2.5Power Query: an import that cleans itself every time
2.6Error handling, validation, and leaving an audit trail
2.7Handing a workbook to someone who did not build it
DrillA 12,000-row sales export with inconsistent regions and text-formatted dates. Produce a weekly summary that refreshes on new data.
Open module 2 →03 SQL for answering questions
3.1Reading a schema before writing a single line
3.2SELECT, WHERE, ORDER BY, and filtering honestly
3.3JOINs, and what a wrong join quietly does to your total
3.4GROUP BY and the aggregation traps that pass review
3.5Window functions: running totals, ranks, period on period
3.6CTEs: writing a query the next person can read
3.7Performance basics, and why your query hung the database
3.8The interview query: thinking out loud under pressure
DrillEight business questions against a sample retail database. Every answer comes with the query and a one-line explanation of what it counts.
Open module 3 →04 Cleaning data, and knowing when to trust it
4.1Profiling a dataset in ten minutes: row counts, ranges, nulls, oddities
4.2Duplicates that are not duplicates, and vice versa
4.3Dates, time zones, and the silent type error
4.4Deciding what to drop, what to fix, and what to flag
4.5Writing down your assumptions so the number can be defended later
DrillA dataset with nine planted problems. Find them, document each decision, and state how the headline number changes.
Open module 4 →05 The statistics you actually use
5.1The four ways an average misleads, with examples from real reports
5.2Spread and distribution: why one number is never enough
5.3Correlation, causation, and finding the confounder
5.4Sample size and significance in plain language
5.5Reading an A/B test result without overclaiming
DrillThree plausible-sounding claims from company dashboards. Say what is wrong with each and what evidence would settle it.
Open module 5 →06 Power BI: from dataset to decision surface
6.1The data model explained without jargon: facts, dimensions, grain
6.2Relationships, and why your totals are wrong
6.3DAX: measures versus calculated columns, and when each belongs
6.4Time intelligence: last month, same period last year, rolling twelve
6.5Designing one page a manager reads in ten seconds
6.6Publishing, scheduled refresh, and row-level security basics
6.7What to do when someone says the dashboard is wrong
DrillBuild a single-page operations dashboard from the module 03 database, then defend every number on it.
Open module 6 →07 Python where it beats the alternatives
7.1When Python is the right tool, and when a spreadsheet wins
7.2pandas: load, filter, group, and the shape of a dataframe
7.3Joining and reshaping without losing rows
7.4A cleaning script you can rerun next month
7.5Quick charts for exploring, not for presenting
7.6Automating a recurring report end to end
7.7Version control and file discipline, the minimum that matters
DrillTake the manual monthly report built in module 02 and turn it into a script that runs in one command.
Open module 7 →08 Working with AI without losing the plot
8.1What language models are genuinely good and bad at in analysis
8.2Drafting SQL and formulas with AI, then proving them correct
8.3The three-way reconciliation: a habit that catches invented numbers
8.4Using AI to explain someone else's query, model or macro
8.5What must never be pasted into a chatbot, and why it is your problem
DrillAn AI-drafted query that returns a believable number. Find its three errors and state the business consequence of each.
Open module 8 →09 Communicating the answer
9.1The one-sentence finding, written before the chart exists
9.2Charts that argue: choosing a form that carries the point
9.3Structuring an executive summary people finish reading
9.4Presenting a number you are not fully certain about
9.5Handling "that doesn't look right" in the room
9.6Saying no, and offering the smaller question you can answer
DrillOne analysis, three deliverables: a Slack message, a one-page memo, and three slides for a board meeting.
Open module 9 →10 The analyst's operating system
10.1Documenting work so future-you can rerun it in five minutes
10.2A request queue that does not drown you
10.3Building a portfolio that survives thirty seconds of a recruiter's attention
10.4The analyst interview: take-homes, case questions, and what they are testing
DrillWrite up the capstone as a portfolio case study: problem, approach, decision it changed, in under 400 words.
Open module 10 →11 Capstone: six days before the board meeting
·A written scoping note, turning one blunt investor line into answerable questions
·A cleaned delivery-log workbook with a documented cleaning log
·A SQL investigation naming the real driver, with a join-trap note
·A confidence-labelled verdict on a pilot the founder already believes worked
·A Power BI-ready model, a documented Python script, an AI-review memo, and a five-minute walkthrough
BriefA three-year-old delivery startup, six days from a funding review, and one line of an investor email as the only brief. You scope it, chase it, and defend what you left unverified.
Open the capstone brief →01 What a client is actually buying
1.1Nobody hires a virtual assistant. They hire back their week.
1.2The four kinds of client, and which two can actually pay you
1.3What AI already does better than you, and what that leaves
1.4The trust ladder: from a one-hour test task to a monthly retainer
1.5Your service, in one sentence a client can repeat to someone else
DrillFive real VA job posts. Name what the client is actually losing, which parts AI would already handle, and what's left that needs a person. Pick the one you'd apply to and say why the other four are traps.
Open module 1 →02 The inbox
2.1Getting access without getting your client hacked: delegation, aliases and shared inboxes
2.2Triage: the four piles and the ten-minute pass
2.3Writing in someone else's voice, and the file that makes it repeatable
2.4Which replies need a template, and which need a human decision
2.5The end-of-day summary that makes a client keep you
DrillSixty messy messages, one hour. Clear the inbox, keep a log of every judgement call, and write the summary the client reads at 6pm.
Open module 2 →03 The calendar
3.1Time zones, and the single mistake that costs people the contract
3.2Booking links, buffers, and protecting someone's focus
3.3The reschedule chain: cancelling without damaging a relationship
3.4Travel, prep time, and the day-before brief
3.5Owning a calendar when you cannot see the reason behind every meeting
DrillA double-booked week across three time zones with two people who both think they're the priority. Fix it, then write the note explaining what you moved and why.
Open module 3 →04 Files, documents and the client's second brain
4.1A folder structure a stranger can navigate in ten seconds
4.2Naming things so that search actually finds them
4.3Documents: templates, formatting, and version control without extra software
4.4Building the client's process library as you go, one page at a time
4.5The handover file that makes you hard to lose and easy to trust
DrillA client's dumped Drive: 400 unsorted files, six near-duplicate contracts, no naming convention. Restructure it and write the one-page map explaining where everything now lives.
Open module 4 →05 Meetings, and what happens after them
5.1The agenda that makes a meeting shorter
5.2Notes: what to capture, and the 80% you should ignore
5.3Turning talk into owned actions with names and dates on them
5.4Transcription tools, and the specific errors they make
5.5Chasing people without becoming the person they avoid
DrillA forty-minute recorded call with three people talking over each other. Produce the notes, an action list with owners and dates, and the three follow-up messages.
Open module 5 →06 Research and data you can be trusted with
6.1Turning a vague request into a research question with a stopping point
6.2Where to look, and how to know when you have enough
6.3Verification, and labelling how confident you are in each row
6.4Getting it into a spreadsheet the client can actually use
6.5Presenting a finding in five lines
Drill"Find me thirty podcasts I should be a guest on." Deliver the sheet, the criteria you used to include and exclude, and the five-line summary.
Open module 6 →07 Working with AI as your junior
7.1What to hand it, and what must never leave the client's account
7.2The brief is the skill: prompts that produce a usable first draft
7.3Checking work you did not do: the three-way check
7.4No-code automations that save real hours, and the ones that break silently
7.5Confidentiality, client data, and the line you do not cross even once
DrillAutomate one recurring task end to end and document it so the client could run it without you. Then write the paragraph explaining why you're still worth paying.
Open module 7 →08 Communication and discretion
8.1Writing a message that gets a decision, not a conversation
8.2Tone across cultures, seniority and time zones
8.3Saying no, saying "not yet", and offering the smaller thing you can do
8.4Escalation: what you decide, what you ask, and how fast
8.5Discretion: what you know, and what you never repeat
DrillThree situations that could each lose you the client: a mistake you made, a request you cannot meet, and something you were told that you should not have been. Write the message for each.
Open module 8 →09 Running yourself as a business
9.1Hourly, retainer or package: the arithmetic behind each, and when to use which
9.2What your hour actually costs you: data, power, transport, tax, unpaid admin
9.3The one-page agreement and the deposit rule
9.4Getting paid across borders: rails, fees, FX, and surviving the long wait
9.5Tracking time and invoicing without losing your evening to it
DrillBuild your rate card from your real costs, write your one-page agreement, and set up a payment route end to end — proven with a live test transaction, not a screenshot.
Open module 9 →10 Getting the first three clients
10.1Choosing a niche narrow enough that someone remembers you
10.2The profile: a LinkedIn page and a one-page service page that do the same job
10.3Proof of work when you have never had a client
10.4Outreach: where they actually are, and the message that gets a reply
10.5The trial task and the interview: how to be the one they choose
DrillTwenty real prospects, twenty messages sent, all tracked in one sheet. Report what came back, and what you'd change about message twenty-one.
Open module 10 →C Capstone: the unreasonable week
·A triaged inbox with a written log of every judgement call
·A repaired calendar, plus the note explaining what moved and why
·A restructured drive and its one-page map
·Meeting notes with owned, dated actions and the follow-ups sent
·One research deliverable with confidence labels on every row
·One automation, documented well enough for someone else to run
·A five-minute walkthrough, marked on how you handle being challenged
BriefYou are dropped into one fictional client's genuinely bad week: ninety emails, a broken calendar, an unsorted drive, a recorded call and a vague research request, all at once, with more arriving than you can finish. Marked on judgement and on what you chose to leave undone — not on tidiness.
Open the capstone brief →01 What a support rep is actually paid to keep
1.1Nobody hires a support rep to close tickets. They hire one to keep customers.
1.2The four kinds of support work, and which two actually pay
1.3What AI chatbots already do better than you, and what that leaves
1.4The trust ladder: from a trial shift to owning the whole queue
1.5Your service, in one sentence a founder can repeat to someone else
DrillFive real support job posts. For each, name what is actually being protected — a rating, a repeat customer, a refund cost — which parts a chatbot would already handle, and what is left that needs a person. Then pick the one you would apply to and say why the other four are traps.
Open module 1 (in course overview) →02 The ticket queue
2.1Getting access without becoming the breach: shared inboxes, permissions, and what never gets shared
2.2Triage: severity, channel, and the ten-minute pass through a stacked queue
2.3Writing in the brand's voice, and the file that makes it repeatable
2.4Macros versus a human reply: when a template is the right call, and when it insults the customer
2.5The end-of-shift handover that makes the next person's morning easy
DrillSixty stacked tickets across email, chat and WhatsApp, one hour. Clear the queue, keep a log of every judgement call you made, and write the handover the next rep reads.
Open module 2 →03 The angry customer
3.1Why customers escalate, and the three complaints that are never really about the product
3.2The de-escalation script: acknowledge, own, act — without over-promising
3.3When policy says no and the relationship says be careful
3.4The public complaint: same fire, different stage
3.5Knowing when you are the wrong person to keep talking, and handing off well
DrillThree real angry-customer threads — one email, one live-chat transcript, one public review. Respond to each within policy, and keep the customer.
Open module 3 →04 Channels: one problem, five sets of rules
4.1Email: the paper trail, and writing for someone who will reread it later
4.2Live chat: speed without sloppiness, and the canned reply that reads canned
4.3WhatsApp and social DMs: informal tone, real boundaries, still logged
4.4Voice notes and calls: what changes when there is no undo button
4.5Choosing the right channel to move a conversation to, and why switching defuses more than it seems
DrillThe same complaint arrives on four channels in one afternoon. Reply appropriately to each without contradicting yourself.
Open module 4 →05 Refunds, returns and judgement under policy
5.1Reading a policy like a lawyer reads a contract — what it actually permits
5.2The refund conversation: verifying without accusing
5.3Where discretion lives: the calls you make alone, the calls you escalate
5.4Goodwill gestures: when they retain a customer, and when they train one to complain
5.5Documenting the exception so the next rep is not stuck explaining it
DrillTen borderline refund requests against a real policy document. Decide, write the reply, and flag which ones needed escalation.
Open module 5 →06 The knowledge base
6.1Turning a repeated question into a documented answer with a stopping point
6.2Auditing an existing help centre for what is outdated, missing, or wrong
6.3Writing an article a stressed customer can actually follow
6.4Keeping macros and docs in sync as policy changes
6.5Reporting the pattern back: when five tickets are actually one product problem
DrillAn outdated help centre with a stale return policy. Find the gaps, rewrite three articles, and flag the pattern worth escalating to the founder.
Open module 6 →07 Working with AI as your junior
7.1What to hand the chatbot, and what must always route to a human
7.2Briefing a bot on your knowledge base without letting it invent policy
7.3The three-way check: catching a hallucinated promise before the customer does
7.4Reviewing AI-suggested replies at speed without rubber-stamping them
7.5Confidentiality: what never goes into a prompt
DrillTen AI-drafted replies, three of which invent a policy that does not exist. Find them, fix them, and explain the business risk of each.
Open module 7 →08 The metrics, and what they hide
8.1CSAT, first response time, resolution time — what each is actually measuring
8.2When chasing a metric makes the service worse
8.3Reporting a bad week honestly, with the reasons, not just the numbers
8.4Saying no to a target that would mean doing the job badly
8.5The one-page weekly report a founder actually reads
DrillA messy week of ticket data. Compute the real numbers, and write the report that explains the dip honestly.
Open module 8 →09 Running yourself as a business
9.1Hourly, retainer or per-ticket: the arithmetic behind each, and when to use which
9.2What your hour actually costs you: data, power, transport, tax, unpaid admin
9.3The one-page agreement and the trial-shift rule
9.4Getting paid across borders: rails, fees, FX, and the wait that makes people quit
9.5Tracking shifts and invoicing without losing your evening to it
DrillBuild your rate card from your real costs, write your one-page agreement, and set up a payment route end to end — proven with a live test transaction, not a screenshot.
Open module 9 →10 Getting your first three clients
10.1Choosing a niche narrow enough that someone remembers you
10.2The profile: a LinkedIn page and a one-page service page that do the same job
10.3Proof of work when you have never had a support role
10.4Outreach: where small brands actually are, and the message that gets a reply
10.5The trial shift and the interview: how to be the one they choose
DrillTwenty real prospects, twenty messages sent, all tracked in one sheet. Report what came back, and what you would change about message twenty-one.
Open module 10 →C Capstone: the backlog
·A cleared sixty-ticket backlog across three channels, with a written judgement log
·A de-escalated public complaint thread that kept the customer
·Three rewritten knowledge-base articles, plus the pattern flagged upward
·Ten AI-drafted replies reviewed and corrected, with the risk of each explained
·One honest weekly metrics report
·A five-minute walkthrough, marked on how you handle being challenged
BriefYou are dropped into one fictional new brand's genuinely bad week: a backlog across three channels, a public complaint gaining replies, a policy grey area, and a chatbot that has been quietly promising a discount that does not exist — all at once, with more arriving than you can finish. Marked on judgement and on what you chose to leave undone — not on tidiness.
Open the capstone brief →
Questions
The things people ask before buying.
Do I need any experience to start?
No. The analyst course assumes you can open a spreadsheet and type in a cell, and nothing beyond that. Every module opens with a short primer covering the knowledge it assumes, which you can skip if you already have it.
Why is Excel taught before SQL and Python?
Because that is the order employers ask for them in. Across 328 analyst job descriptions this year, Excel appeared in 81%, SQL in 60%, Power BI in 43% and Python in 41%. If you stop halfway through the course you have still learned the two things you are most likely to be tested on.
What software do I need?
Excel or Google Sheets, Power BI Desktop, and Python. All three are free to use for the course, and the sample database is hosted, so there is nothing to install to follow the SQL modules.
How long does it take?
It depends which track you take. Twelve weeks at three hours a week if you are starting cold, six to eight weeks self-paced if you already work with data, or a fixed eight-week cohort for teams. Same lessons in all three, different pacing and assessment.
Is the AI course only for technical people?
No, and it involves no coding. It is aimed at whoever owns a process that is eating the week. The hard parts are choosing the right process, measuring a baseline before you build, and deciding where a human stays in the loop.
Do I get feedback on the capstone?
On the all-access and team plans, yes, marked against the published rubric. On a single course you get the rubric and the worked example, and you can mark it yourself.
What if it isn't for me?
Email within 30 days and I will refund you, no forms and no questions. The first three modules are free, so watch those before you buy and you probably will not need to.
Start today
The first three modules are free.Do the drill, then decide.
Modules 01 to 03 cover the two skills that appear in most analyst job ads. Work through them, attempt the drill, and judge the teaching on the result rather than on this page.
Start module 01