Eight steps to your ESS IA
The internal assessment (IA) is the biggest single piece of work in the course, worth a fifth or a quarter of your grade.
It is also, genuinely, the interesting part. You choose somewhere real, find out what is happening to it, and argue something about what should be done. What follows is the order we suggest doing that in, and why.
Where is your data going to come from?
Both routes are marked against the same six criteria and both can reach full marks. What changes is where the difficulty sits. Pick one and the guide shows you that version of every step. You can switch at the top of any page.
Primary data
- What it looks like
- A place you can get to, repeatedly. Quadrats, transects, probes, a questionnaire. Usually one morning of fieldwork, and no second visit.
- What it asks of you
- A method a stranger could follow to get similar data, and enough readings to run the test you chose before you set off.
- Where the difficulty is
- Getting the design right in advance, because you cannot go back and add the variable you forgot.
- The scale you work at
- One hillside, one reach of river, one park, one morning.
Secondary data
- What it looks like
- Numbers somebody else already collected and published, which you download and work on yourself. Water quality records, weather stations, satellite measurements, national statistics.
- What it asks of you
- Instructions clear enough that a stranger could download exactly the same file you did, and a rule you fix in advance about which of the numbers you will use.
- Where the difficulty is
- Finding data that exists for the place and the years your question needs, and being honest about how you chose it once you can already see the answer.
- The scale you work at
- Fifty years, or fifteen cities, or a whole river basin.
The environmental issue. The real-world strategy and the people arguing about it. The statistics, the analysis, the evaluation, and the 3,000 words. Criterion B (Strategy) is worth the same four marks on either route, and it is still the part that could not appear in a Biology or a Geography investigation. Steps 3 and 8 are the same page whichever route you take.
Neither route is the easy one. Secondary data means doing the analysis yourself, on numbers somebody else collected. It does not mean writing about what other people found. A report that only reviews other people’s writing does not count as a repeatable method, and that alone holds Criterion C (Method) at the bottom band, however good the rest of the report is.
Doing both is allowed and fairly common: measuring a stream yourself while pulling ten years of flow records to set your morning in context. Take the primary route and read across to the secondary one for the parts about finding and citing a dataset.
The process
Each step produces something the next one needs. The word counts are what we suggest spending when you eventually write that section up, not what to write now.
The IB sets no section word counts and no required order. These figures are ours, and so is the order. The next section explains the one part of it that will surprise you.
Why the real-world strategy comes third
Most people find a dataset and then, somewhere near the end, add a section about stakeholders. I suggest it goes before you settle on the data, and it is the most consequential decision in this whole guide.
Here is the reasoning. A data report has a natural shape that assembles itself: download, filter, chart, statistical test, conclusion. That shape is what almost every example online looks like. What it never produces on its own is a real-world strategy and the people arguing about it.
So that layer is not something you can lose. It won't appear unless you deliberately put it there. Settle the argument first and your dataset gets chosen to serve it. Leave it until just before the draft is due and you will be hunting for a strategy that fits results you already have.
Choose the argument before you choose the measurements, and your investigation will have something to be about.
What makes this an ESS investigation
ESS is an interdisciplinary subject in the Diploma, and the investigation is where that has to be visible. Three ideas run through the course, and a strong report carries all three.
Your variable sits inside something with inputs, outputs, stores and feedbacks. Not a list of measurements.
People want incompatible things, and what happens to an environment depends on how that gets resolved.
Whether what is happening can carry on, and at whose expense.
One criterion asks you to name a real-world strategy addressing your issue and explain a tension between the people who disagree about it. Four marks out of thirty, and it is the only part of the report that could not appear in a Biology or a Geography investigation.
That is not a reason to avoid biological or geographical data. Monitoring records, satellite products and national statistics are all fine, and the guidance says data may be primary or secondary. It is a reason to make sure there is an environmental issue and a live argument wrapped around whichever data you choose.
How to use this guide
Work through the steps in order. Each one opens with what you should be arriving with and closes with a checklist, and the checklist is the point rather than the decoration. It is the only part that asks about your investigation rather than explaining things in general.
Tick honestly. An untickable box is not a mark against you, it is simply the next thing to go and do. If you are signed in, your ticks are saved and your teacher can see how far along you are.
The rivers that feed Lake Geneva
The investigation asks whether thirty years of investment in Geneva’s sewer network shows up in its rivers. One published cantonal dataset: E. coli counted at 184 monitoring stations, year by year, from 1995 to now.
It began as a study of bathing water at the lake shore and had to move. The only lake station in the file reads zero in 24 of its 28 years, so the receiving water is clean and has nothing to say; all the variation is in the tributaries feeding it. That was only discoverable by opening the data.
What makes it an ESS investigation rather than a spreadsheet exercise is what is happening around it. Three real strategies act on those rivers: a cross-border river contract with France, a cantonal drainage programme whose cost is shared across every commune, and a renaturation fund whose most visible project took land from farmers who said so publicly. Real plans, and people with genuinely incompatible goals arguing about them.
It runs through every step because it went wrong in useful ways, and none of that has been tidied up:
- A question the data could not answer · step 2The first version asked about bathing water at the lake shore. The only lake station in the dataset reads zero in 24 of its 28 years, so there was nothing to analyse and no rewording could have fixed it.
- Thirty-seven thousand rows, seven hundred observations · step 4Every station-year was repeated about fifty times. The means were fine and the sample size was destroyed, which is the worst combination: a test on the file as downloaded returns a significant result whatever is in it.
- A missing year nobody announced · step 52007 is simply absent. So is the reason. A chart that draws a line straight across it is asserting something the data does not say.
- Averages of one visit beside averages of twelve · step 4The file records how much sampling sits behind each value, and the range is 1 to 12. Comparing them untouched compares two different kinds of number.
- A growing network that looks like an improvement · step 6The monitoring grew from 19 stations to about 30, so early and late years describe different places. Restricting to the 49 present in both was the difference between a finding and an artefact.
- A headline that was false at a third of the sites · step 6The median improved and 15 of 49 stations got worse. Deciding which of those two sentences to lead with is most of Criterion E.
Every one of those is more use to you than a study where everything worked first time.
One thing before you start
You will get one round of written feedback on a full draft. Your teacher can comment on what is strong and what is weak, but they cannot correct it, and there is no second draft after that.
Which means two things. Hand over something complete rather than something polished, because feedback on three finished sections and five missing ones is feedback on three sections. And aim for 2,900 words rather than 3,000, because that feedback will almost certainly ask you to develop something, and at the ceiling the only way to do that is to delete something good.
Most advice tells you to choose a topic. We think you should choose a place, and step 1 explains why.
Start step 1The Geneva rivers investigation used on the secondary-data route of this guide is the author’s own analysis of a published cantonal dataset. Every figure quoted comes from the Canton of Geneva’s open data, downloaded on 10 September 2026; the duplication, the missing year and the varying campaign counts described in the guide are as found on that date and may since have been corrected. The selection rules, the analysis and the conclusions are the author’s and not the canton’s. It is written to show how an ESS investigation is built, not to report findings about water quality in Geneva.