Stage 2

 Reference Sheet BB 

Triangulation and analysis

 Triangulation 


Triangulation is an important technique for processing collected data. It checks the validity of your findings and starts to build a knowledge base. Good risk managers constantly triangulate new data. Just as a triangle has three sides, at least three sources that converge on roughly the same finding are needed before you can conclude that information is strong, meriting the status of ‘knowledge’. Weight of evidence suggests that, if we examine a given issue from different points of view and independently reach the same finding in each case, it is reasonable to conclude that the information is more than likely to be valid. It is an industry standard in mixed methods research, and should be a fundamental part of your risk assessment.


The four types of triangulation to consider are described in the table below. The most common are data triangulation (comparing responses across key informants or respondents) and method triangulation (comparing findings across collection methods). Whatever the type, triangulation is a structured way to compare findings and identify divergences, convergences and gaps. 

Type
Example

If findings diverge (for example, when measured by different methods), you will need to follow up, to be certain you understand why, and to correct your results if they prove to be false. In contrast, if findings converge (and different methods repeat them), it strengthens confidence in the results. 

 

The triangulation matrix combines perspectives and characteristics, and compares them across collection methods to record a community’s story of risk and resilience. In this story, each characteristic of resilience is a chapter, and each method an actor with a compelling perspective. In each cell of the table, you record full sentences of rich detail that tell the story. In the survey and observation columns, you insert summary statistics (the quantitative dimension of the story).

Ref. Sheet BB - HIGH.png

 Processing and analysing data 
 

This provides guidance on how to process evidence the community has collected. 

 

Comparing what you see with what you hear. Every person involved in an assessment needs to individually nurture and continually employ their observation skills. These help the participants to process what they hear, and capture discrepancies and areas of convergence. During focus group discussions and surveys, for example, one team member should always be asked to observe and take guided notes on what they see: body language, interactions, relative positions, expressions of power and social mores, etc. These observations are qualitative evidence that provide context and contribute to the processing of assessment results.

Processing data at the end of each group session. You brought people together in a focus group, for example, to explore vulnerability, threats or a given characteristic of their resilience. If you have prepared well, you knew exactly what the aims were, and whether or not you met them. (And, if aims were not met, you should have a solution or back‑up plan ready.) 
Rather than summarize the results yourself, however, give participants the opportunity to draw their own conclusions. The last question you ask should be open‑ended: invite them to say what they remembered or learned from the session. Ask, for example: “If you tell your spouse or friend about this meeting, what will you tell them?” Even if they do not mention content (and some will), you will receive strong feedback on how participants perceived the process. Every group session needs to end by giving the participants a chance to express their own conclusions. This is a critical part of data processing in the community. 

  • Before you process and analyse your data, take stock of the findings. See where they converge or diverge before drawing any conclusions about trends. Fill in gaps that have been noted.

  • Allow ample time to process your data. Rushing through processing will always cause you to miss many important connections.

  • To properly process the data, designate one team member from the community to manage the evidence base. They will need to know where pieces of evidence are, and the format they are in, etc., and should obtain key pieces from team members once they have been discussed, to archive them carefully.

  • It is never too late to process more. Feel free to return to the data to test an idea that occurs to you later or query a conclusion. Do this even if data collection and processing have already taken place and actions have started. The most important thing is to learn from our errors, mistakes or wrong impressions. Admit when you go astray and take matters forward from there. Both quantitative and qualitative data require analysis. It is more challenging to analyse qualitative than quantitative information because it contains more words, which have multiple meanings and obey fewer rules.

  • Data disaggregation may not be feasible unless you have planned your collection process in a way that enables you to capture the different aspects of risk stories that you want to disaggregate. Where it is possible, go back and collect additional data if you lack evidence of the right sort. Disaggregation is a critical dimension of analysis, because it gives a voice to key groups that otherwise may be marginalized.

  • Reduce data to key findings. This is one of the hardest and most important steps of assessment. The challenge is similar to writing a one‑page summary of a 100‑page report. Don’t underestimate the time required. Finding a structure (like the structure of a triangulation matrix) is critical to successful summarizing.

  • Make concluding statements and keep notes, notably of the original ideas that participants expressed in their own words, because then the community can recognize themselves and their own thinking in the final result. Add interpretive qualifications (perhaps in italics) so that those reading these can see that they have not yet been reviewed by the community. What the community does not own or identify with should be discarded (or set aside for later work).

 Organizing the data 
It may be easier to organize and process some data at your branch office. If you do so, members of the community and volunteers should continue to participate fully.

  • After an intensive data collection process, organize and process your data. You will have handwritten notes of every session recorded by your assessment team. You will have the flip charts as well: these should be typed up in a format that reminds you of everything that was said and felt during the session. You may have survey and observation forms that need to be keyed into a computer. You may also have data that have already been entered or saved on cameras, audio recorders, tablets or telephones. At regular intervals, the team should also have completed a triangulation matrix for each community. Each of these pieces of evidence should be inventoried and their originals kept in a safe place.

  • Enter your quantitative data (if it is not automatically captured by a tablet, phone or other technology). Number your completed survey or observation sheets and create a data entry mask (for example, in MS Excel) in which to key them in. When this has been done, clean all the data (error check by looking for logic errors, outliers or empty cells, etc.). Check the numbered surveys or sheets if you see that an error was introduced during data entry or data capture. When you are comfortable with the quality of your numerical data, you can develop some initial summary statistics. Use a spreadsheet program to calculate the sums, frequencies and averages of your quantitative data, as appropriate. (MS Excel is proposed because it has easy‑to‑learn formulas and is globally the most accessible programme. More sophisticated statistical software packages (SPSS, SAS, STATA or EPI‑INFO) are able to go way beyond summary and descriptive statistics. Numerous sophisticated data analysis techniques using statistics exist, but are not the subject of this guidance.)

  • If you have not been able to carry these summary statistics into your triangulation matrix, do so now. This step will offer you a chance to compare the new facts with the qualitative findings, generating deeper insights. Present the numbers in full sentences to add quantitative findings to the triangulation matrix.

  • Qualitative data. When you triangulated (as described above), you processed mainly qualitative data. When you deliberately noted where it converged or diverged, you applied a technique known in qualitative research as coding. Coding is a process of grouping words or phrases (and assigning them a name or code) in a manner that allows their meaning to be counted and compared. When you noted that three out of four key informants or two out of three applied methods produced the same conclusion, you coded them ‘green’ to show convergence. You may also have concluded, for example, that “5 out of 6 sources reported that [adversity X] was the most problematic for this community”. In coding, any piece of qualitative evidence you collected can be counted, that is, converted into a quantitative form for logical analysis.

  • In the VCA, Methods Reference Sheet 3 (The Wall Method) offers further ideas on processing qualitative data using triangulation.

  • If you have time to process (and eventually analyse) more deeply, transcribe (type up) recorded interviews or focus group discussions into a document file. Such files can be coded electronically by qualitative data software. They use the same type of coding as the triangulation matrix, although it is more sophisticated and sometimes easier to quantify. You can also code with colours or symbols on flip charts or coloured Post‑its on a wall. The best processing technique is the one that works for you, in your context.