Medical lab technician to data analyst: QC charts and TAT data

What you already handle that analysts are paid to handle

Most lab technicians work with more structured data than they realise. Before planning anything, list the records you touch in a normal week:

• LIS worklists and result exports — sample ID, test code, collection time, receipt time, result authorisation time. That is a timestamped event table.
• Internal QC logs and Levey-Jennings charts — control values against a mean and standard deviation. You already apply Westgard-style rules, which is statistical process control by another name.
• Sample rejection register — reason codes such as haemolysed, clotted, insufficient volume, mislabelled. That is categorical data with an obvious business question attached.
• Critical value call log — time the value was found, time the clinician was informed. Another duration metric.
• Reagent and consumable stock sheets — lot numbers, expiry dates, consumption.

The point of the list is not to impress an interviewer. It tells you which dataset to rebuild synthetically, because you already know its quirks: missing receipt times, results re-authorised after a repeat run, samples logged at the collection centre hours before they reach the lab.

Days 1–30: SQL on a lab-shaped table

Spend the first month on SQL only, using a table you understand. Here is a small synthetic TAT dataset and the first query worth writing: median-style comparison is hard in plain SQLite, so start with average TAT and the share of samples breaching a target, by test and shift.

CREATE TABLE samples (
  sample_id TEXT PRIMARY KEY,
  test_code TEXT,
  shift TEXT,
  received_at TEXT,
  authorised_at TEXT
);

INSERT INTO samples VALUES
('S001','CBC','morning','2026-09-01 08:10','2026-09-01 08:55'),
('S002','CBC','morning','2026-09-01 08:20','2026-09-01 09:40'),
('S003','CBC','night','2026-09-01 23:05','2026-09-02 00:50'),
('S004','LFT','morning','2026-09-01 09:00','2026-09-01 11:30'),
('S005','LFT','night','2026-09-01 22:40','2026-09-02 02:10'),
('S006','LFT','night','2026-09-02 01:15','2026-09-02 03:05'),
('S007','CBC','night','2026-09-02 02:00','2026-09-02 02:45'),
('S008','LFT','morning','2026-09-02 10:10','2026-09-02 12:00');

-- Hypothetical targets: CBC 60 minutes, LFT 120 minutes
SELECT
  test_code,
  shift,
  COUNT(*) AS samples,
  ROUND(AVG((julianday(authorised_at) - julianday(received_at)) * 24 * 60), 1) AS avg_tat_min,
  SUM(CASE
        WHEN (julianday(authorised_at) - julianday(received_at)) * 24 * 60 >
             CASE test_code WHEN 'CBC' THEN 60 ELSE 120 END
        THEN 1 ELSE 0 END) AS breaches
FROM samples
GROUP BY test_code, shift
ORDER BY test_code, shift;

On this synthetic data the night-shift LFT average is well above the morning figure, and CBC breaches cluster in specific rows. The analysis skill is what you do next: check whether the night figure is driven by one outlier (S005), whether receipt time was recorded when the sample arrived at the bench or at the front desk, and whether eight rows justify any conclusion at all. Write those caveats down. Interviewers notice when someone questions their own numbers.

By day 30, aim to be comfortable with GROUP BY, CASE, joins between a samples table and a test master table, and date arithmetic.

Days 31–60: turn QC logs into a chart someone can act on

Your second month is about visual explanation. Rebuild a Levey-Jennings chart in Power BI or Excel from a synthetic control-value table, with mean and ±2 SD bands, and add a simple flag column for a rule you already use (one value beyond 3 SD, or two consecutive values beyond 2 SD on the same side).

A usable Excel version, assuming control values in column B from row 2, mean in $E$1 and SD in $E$2:

C2:  =IF(ABS(B2-$E$1)>3*$E$2,"1-3s","")
C3:  =IF(AND(ABS(B2-$E$1)>2*$E$2,ABS(B3-$E$1)>2*$E$2,SIGN(B2-$E$1)=SIGN(B3-$E$1)),"2-2s",IF(ABS(B3-$E$1)>3*$E$2,"1-3s",""))

Fill C3 downwards. Then build one page: TAT breaches by shift, rejection reasons as a sorted bar, and the QC chart. Keep it to three visuals. A lab manager reads a crowded dashboard the way a clinician reads a crowded report — not at all.

Days 61–90: the gap you have to close deliberately

Lab work trains precision and procedure. What it rarely trains:

1. 1Framing a question for a non-clinical manager. "Night LFT TAT is high" is an observation. "Two of three night LFT breaches happened when one technician covered both haematology and biochemistry" is a finding with a decision attached.
2. 2Statistics beyond QC. Learn distributions, medians versus means for skewed durations, and why a small sample needs a cautious sentence.
3. 3Business vocabulary. Revenue per test, collection-centre throughput, repeat-test cost. You do not need finance depth, but you need to connect delay to money or patient experience.
4. 4Python basics — reading a CSV with pandas and reproducing your SQL result is enough at this stage.

Use the third month to write the project up: problem, data (declared synthetic), method, finding, limitation, recommendation. Run your resume through the free skill check to see which analyst skills it still does not evidence.

Diagnostics-chain operations analyst first

Aim first for roles where your context shortens the learning curve: operations analyst or MIS analyst at a diagnostics chain, hospital or health-tech company, then a general junior data analyst title. These roles value someone who knows why a receipt timestamp can be wrong. If you later want modelling work, read data analytics vs data science before deciding.

From the analyser bench to the analyst desk, in one answer

"For four years I worked the biochemistry bench and handled our QC and TAT registers. I kept noticing that we discussed delays by memory rather than data, so I learned SQL and rebuilt a synthetic version of our TAT log to show where breaches concentrate by shift and test. The part I'm proudest of is the limitations section — I found that receipt time was logged at two different points, which would have made one shift look worse than it was. I want to do that kind of checking full time, and a diagnostics operations team is where I understand the data best."

The answer works because it names a real record, a real method and a real caveat. If you want structured practice in SQL, Power BI and verified use of generative AI for this kind of analysis, the Data Analyst programme is built around weekly deliverables of exactly this shape.