D²I methodology
We connect the question, the data, the method and the communication to produce rigorous, transparent and useful results.
D²I (Data to Insights) is the framework we use to approach every biostatistics and data science project, from understanding the context to interpreting and transferring the results.
What is D²I?
D²I is Azahar Data Insights' own methodology for keeping the project's objective, the available data, the methodological decisions and the intended use of the results connected. Its purpose is to prevent the analysis from becoming a series of isolated tasks or a black box that is hard to review.
The methodology combines three capabilities that have to move forward together: scientific and technical knowledge, a strategic understanding of the problem and clear communication. None of them can deliver a good result on its own. A correct method can answer the wrong question; a sound interpretation can lose its value if it is not documented; and a solid result may never be used if it is not communicated for the right audience.
A methodology that adapts to each project
Every project starts from a different situation. Some begin before the data exist, during study design, the definition of variables or the planning of data collection. Others need to review an existing database, test a hypothesis, develop a model, evaluate an intervention or build a tool for exploring the results.
D²I does not impose a fixed sequence or a set repertoire of techniques. The phases can overlap, repeat or carry different weight depending on the objective, the maturity of the project and the information available. Interim findings may also make it necessary to reframe a question, revisit a variable or go back to an earlier stage.
The methodology does not guarantee an identical path for every project. It ensures that the question, the data, the method and the intended use of the results stay connected throughout the process.
Want to apply this approach to your project?
Tell us about it and we'll work out together how to approach it.
Book a sessionFrom data to knowledge
The diagram gives an overview of how we connect understanding the problem, preparing and analysing the data, and communicating the results. It is an overall view, not a contractual sequence: the specific path and the outputs it produces are adapted to what each project needs.
Understand · Model · Communicate
D²I is built around three complementary dimensions. They are not three separate services or watertight compartments: each one creates the conditions for the others to make sense.
What do we need to answer, and how far can the data take us?
Before choosing a technique, we need to understand the context, the objectives and the decisions the analysis should help to inform. That means knowing how the data were generated, what their variables really represent, what quality they have and what limitations the study design introduces.
Understanding may involve reviewing sources, inclusion criteria, definitions, collection processes, missing values, inconsistencies and initial patterns. The aim is not to pile up checks, but to build a shared view of the problem and to determine which questions can be answered on solid grounds.
This dimension avoids one of the costliest mistakes in a data project: answering the wrong question precisely, or drawing conclusions that the available evidence cannot support.
Which analysis can deliver a valid, proportionate answer?
Modelling means turning a well-defined question into an analytical strategy suited to the design, the data and the objective. It does not always mean using complex algorithms, nor does it necessarily mean building a predictive model.
The approach may draw on descriptive or inferential statistics, explanatory or predictive modelling, survival analysis, longitudinal methods, simulation, causal inference, Bayesian approaches or machine learning, among other options. A technique is chosen because it helps to answer better, not because it is newer or more sophisticated.
The choice has to come with assumption checking, diagnostics, validation, performance assessment and sensitivity analysis when the project requires it. Modelling also means interpreting honestly how far the results go and what uncertainty remains.
How should the results be presented so that they can be understood and used?
An analysis does not end when an estimate is calculated or a model is fitted. The results have to be interpreted in context, prioritised and presented appropriately for those who need to review them, apply them or make decisions based on them.
The audience and the intended use shape the work from the start. A research team, an executive board, a clinical committee and a non-specialist audience need different levels of detail, narratives and formats.
Communicating clearly does not mean simplifying until rigour is lost. It means making the logic of the analysis visible, explaining its assumptions and limitations, and making it easy for each audience to find the information it needs.
What underpins the quality of the analysis
D²I includes five working conditions that help maintain methodological quality throughout the project: starting from the question rather than the technique, justifying complexity, assessing robustness, preserving traceability and reviewing the results together with the team.
The question before the technique
Each method has to serve the project's objective and be compatible with the design and the quality of the data. We don't pick a technique first and then look for somewhere to apply it. We start from the question and assess what answer the available evidence can support.
Justified complexity
The most complex solution is not necessarily the best. We choose the level of complexity that adds value, taking into account validity, interpretability, the resources available and how the results will be used afterwards.
Robustness and uncertainty
A result is not reliable just because a model has converged or a test has reached statistical significance. When the project requires it, we review assumptions, stability, sensitivity and reasonable alternatives, and we report the estimates together with their uncertainty and limitations.
Traceability and reproducibility
Every number, table or visualisation must be traceable to the data, the transformations and the decisions that produced it. Depending on the agreed scope, the delivery can include auditable code, methodological documentation, version control and instructions for regenerating the results.
Shared review
The analysis is not carried out behind the team's back until the final delivery. We check the context with the team, share relevant findings and propose adjustments when we spot weaknesses, risks or sounder alternatives.
We think with the team, not just for it.
The methodology grows out of a way of understanding science
D²I explains how we approach projects. Our principles explain why we work this way and which commitments guide our relationship with data, with people and with the consequences of each analysis.
Read the principles that guide our work →If what we explain isn't understood, it might as well not exist.
Results built to stay useful
The output of a project is not necessarily limited to a single document. Depending on the objective, the audience and later use, the delivery can combine analytical evidence, reproducible assets and different formats for communication or exploration.
Analytical evidence
Analytical evidence is the knowledge generated during the project: descriptive and inferential results, estimates with their measures of uncertainty, explanatory or predictive models, evaluations, simulations, scenarios and sensitivity analyses.
It also includes the interpretation of the findings in context, the assumptions that underpin the analysis and the limitations to bear in mind before using the results to guide decisions.
Reproducible assets
When they are within scope, the results can come with processed and documented databases, data dictionaries, organised and auditable code, a log of transformations, methodological documentation and the instructions needed to reproduce or extend the analysis.
These materials make it possible to review how each result was reached and make it easier for the work to continue after delivery.
Formats for communication and use
The results are presented in the formats that best suit the audience and the intended use. These may include:
Common formats
Statistical report
Complete, stable documentation for studies, projects and review processes.
Web report
Results organised by section, with navigation and interactive elements.
Dynamic dashboard
Exploration of dimensions, scenarios or levels of aggregation through filters and visualisations.
Infographic
A visual summary of the main findings to help communicate them.
Depending on the project, the delivery can also include executive or technical presentations, publication-ready figures and tables, materials for committees or reviewers, and presentation or knowledge-transfer sessions. There is no one-size-fits-all deliverable: we define the right combination so that the results can be understood, reviewed and keep adding value after delivery.
Tell us where your project stands
We can help whether you are still defining the study or already have data, results or an analysis that needs reviewing. We will look at the context, the information available and the intended use to propose a rigorous, proportionate approach.