Your first project

Welcome! This page walks you through your first Skimle project from an empty screen to a coded, explorable analysis. It takes about fifteen minutes of your time, plus a short wait while Skimle reads your documents.

There are four steps:

Step 1: Create a project

On the home screen, click New project in the top right corner of Your projects.

Skimle creates the project straight away and takes you to its data sources page. It starts life as "Untitled project": click the name in the top bar to rename it to something you will recognise later, such as "Hybrid work interviews".

Step 2: Add your data

A new project opens on Let's add some data!. A data source is one batch of material that you add to the project and analyse together, and a project can hold several of them (say, interviews and a survey export).

First, check the project language in the top right. It sets the language of everything Skimle writes for you: category names, summaries, reports. Your documents can be in any language, and you can mix languages within a project. The project language cannot be changed once documents have been added, so it is worth a glance now.

Then choose the kind of data you have:

For a first project, Text documents is the easiest start. Drag your files into the upload area (or click it to browse), and add as many as you like in one go.

Your files now wait in a list to confirm, showing their length and the credits each one costs. Before you confirm, two settings are worth knowing:

• Anonymise documents removes names and other identifying details before analysis. It is off by default; see anonymising documents for the options.
• Automatically analyse runs your project's analyses on the new documents as soon as they are processed. It is on by default and applies to every data source in the project. It only has an effect once the project has at least one analysis, so on a brand new project nothing runs yet: you choose your first analysis in the next step. Switch it off if you prefer to pick by hand which documents go into an analysis.

When you are happy, click Confirm files. This is the moment credits are spent (see pricing). For most projects these are the only credits you will ever need for those documents: once a document is in your project, you can analyse it as many times and in as many ways as you like at no extra cost.

No data to hand? Try the demo dataset, or browse our list of free datasets for qualitative research, from interview archives to parliamentary speeches.

Step 3: Choose an analysis

Open Analyses in the left menu. On a new project you will see Add the first analysis to your project with five choices.

Each one answers a different question about your data. Here is the short version:

Not sure? Start with identified themes. It needs no setup and gives you a good map of any interview-style material. You can add any of the others later on the same documents, at no extra cost.

Identified themes: let the data speak

This is Skimle's "zero prompt" option. Skimle finds the questions and headings your documents share, turns them into a parsimonious set of themes, extracts insights for each one from every document and groups them into subthemes. It works best when your documents follow a similar structure, as interviews from one interview guide do.

A project has one identified themes analysis, which is why its card disappears once you have added it. Read more in automatic thematic analysis, or see how it fits a full method in how to do thematic analysis with AI and how to find themes across interviews.

Predefined categories: decide what comes out

Here you name the categories you want, and Skimle codes every document against them. Click Add your own category, or pick from the suggestions Skimle draws from your data. Each category can carry instructions about what to include, and subcategories form from what Skimle finds.

That makes it work both ways round: bring an existing coding frame and the analysis is deductive; start from one broad category and let the subcategories emerge, and it is inductive. Our guide to inductive, deductive and abductive coding helps you decide, and predefined categories covers every setting.

Targeted analysis: point each category at the right place

Targeted analysis works like predefined categories, with one addition: you choose where each category reads from. That matters when only part of each document is relevant. If your interviews have a question about improvements, a category called "Improvement suggestions" should read the answers to that question, not every passing complaint in the transcript.

Skimle works out the sections of your documents for you: interview questions, document headings, or the columns of a spreadsheet. Next to each category, the Sections read setting opens a picker listing the sections your documents share and how many documents contain each one.

Tick the sections the category should read, or leave them all unticked to read whole documents. The preview at the bottom tells you exactly what the category will cover: here, all 24 interviews, but only about a third of their text. A document that has none of the ticked sections is skipped for that category.

A few more things targeted analysis can do:

• Limit which documents it reads, for example only the interviews and not the survey. Edit document scope under Optional settings filters by data source, by metadata or by excluding single documents. Every analysis type has this setting.
• Set up a literature review in one click. The Set up default literature review button adds six ready-made categories (research gap and motivation, concept definitions, key findings, original contributions, methodological choices and empirical setting), each pointed at the matching parts of academic papers, such as the introduction, methods or discussion. It recognises those parts whatever language the papers are written in. See our guide to qualitative evidence synthesis for the method behind it.
• Rebuild the section structure if you have added many documents since the first upload, so Skimle reconsiders the sections across everything you have imported.

Agentic analysis: start from a research question

Agentic analysis is Skimle's most thorough mode. You type a research question, such as "why do some customers stop visiting after one purchase?", and a short setup chat clarifies it with you and proposes a framework of categories and broader questions, which you can edit. Skimle then codes the evidence across your documents, looks for counter-evidence and writes analytical themes and a report that answer your question.

Choose it when the "why" matters as much as the "what". Read agentic analysis for the details, and how to use agentic analysis in Skimle for a worked example.

Skip analysis: code by hand

Prefer to read and code everything yourself, or just want to browse the documents first? Skip analysis creates an empty category for manual coding. You can still add an AI analysis at any time later. Many researchers mix the two; see combining Skimle's AI analysis with manual coding.

Running your analysis

Once you have chosen and set up an analysis, start it with the Analyse button. If Automatically analyse was on when you confirmed your documents, Skimle starts analysing them by itself as soon as the analysis exists.

The analysis runs in the background and usually takes a few minutes, depending on how much text you have. You will get an email with a link when it is ready (check your junk folder if it does not arrive), and you are free to close the tab in the meantime. When many analyses are running at once, yours may wait in a queue for a little while, but once it is done, browsing the results is instant.

Step 4: Explore your results

When the analysis is ready, the Explore & edit section of the left menu comes to life. The Categories view is the best place to start.

On the left is your category tree; click any category to read its AI-written summary, see how its insights are spread, and open every insight with its verbatim quote. Each quote links back to the exact place in the source document, so you can always check the context. You are in charge of the structure: rename, merge, split and move categories as you like (editing in the categories view).

Visualisations gives you the whole project at a glance: which categories are biggest, how they differ between groups and how they change over time.

From here, a few good next steps:

• Read documents in the documents view, with every coded quote highlighted, or see everything at once in the table view.
• Compare groups by adding metadata, such as role, site or customer segment, and using metadata analysis. Our worked example on discovering themes with metadata variables shows how.
• Ask questions in the chat: "What frustrates customers most? Give me three quotes." See how to use Skimle's chat for ten things it can do.
• Mark things up with tags and notes that cut across categories.
• Share and export your results to Word, Excel, PowerPoint or REFI-QDA for NVivo, ATLAS.ti and MAXQDA, from the export centre.

Where to go next

• Bring in more material: audio and video, spreadsheets or a Skimle Ask interview.
• Work with colleagues: sharing and collaboration.
• Understand what happens under the hood: how analysis works.

Questions or stuck somewhere? Use Send feedback in the left menu, and a real person will get back to you.