Data collection is the act of collecting and aggregating quantified, reliable information from a multitude of sources to provide answers to questions, recognize trends, predict future developments, and facilitate decision-making. We often use the term “data” in today’s technological world to describe a source of information; if “information” is knowledge and “knowledge” is power, “data” is power in any organization.

However, you are required to collect data before you can convert that into strategic action. And, for data to become a strategic action, collection has to take place.

So, if you’re ready to get started, let’s jump in and learn about data collection; what does it mean? (Hint: It’s a lot more involved than typing a few words into Google!) What are the forms and types of data collection, and what tools and technologies are relevant or helpful to use to collect meaningful data?

If you want to understand the data collection process from the ground up, you are at the right spot. Let’s go!

Data collection is relevant for the purpose of collecting and analyzing any data from any source to help answer research questions, solve problems, evaluate outcomes, and predict possible trends. It is an essential part of interdisciplinary research and analysis, and decision-making in science, business, healthcare, etc.

To collect data efficiently, researchers must determine the type of data necessary to answer their research question, find reliable sources of data, and choose the methods necessary to collect the data. As we will discuss, there are many different types of data collection that can be done, and each type of data collection is suited well for certain purposes and situations. Data collection is valuable and the basis of sound decision-making in evidence-based academic research, private sector business, and public sector government.

Just as a judge has to establish evidence to render a judgment, or a general must gather intelligence before proceeding with a strategy, sound decision-making is based on information. In many instances, data is information, and informed selection starts with precise, relevant data.

Although data collection has been prominent in history for thousands of years, and in some ways represents the same techniques and notions, the volume, variety, and velocity of data have never eclipsed what we see today. In fact, in this new era of digital technology, the same data is no longer available, let alone the way we collect and manipulate it. To compound this effect, methods of data collection must evolve within the frameworks of the new digital era.

Understandably, whether one is conducting research in an academic environment or a business person is developing a marketing strategy, data collection forms an essential baseline for smarter decision-making and more efficient or effective strategies. Data helps researchers understand patterns and better measure the value of their information and performance, and it reduces the risk of using totally unrelated information to make decisions.

We now have a better idea of what data collection is and why it is important! Let us explore the various techniques and approaches to collecting data. Data can be collected by using telephone surveys, feedback forms, conducting online polls, conducting interviews in a one-on-one or group context, and many other methods. Data collection can be grouped and categorized in several distinctive ways.

Data collection methods usually fall into one of two major categories: Primary and Secondary. Both types are important in research, analysis and decision making. The following outlines each data collection method in detail.

Primary data collection occurs when data is collected from an original, direct source. It provides researchers and evaluators with direct, original information that is specific to their study objectives. Primary data collection is helpful when fresh or original data is needed.

Examples of primary data collection techniques are:

• Surveys and Questionnaires

Surveys and questionnaires are established sets of questions to solicited responses from individuals or groups. These can be done in-person, by telephone, through email or via online platform.

• Interviews

An interview involves direct interaction, whether in person or virtually, with the research participant. Interviews can be structured (with a fixed set of questions), semi-structured (some guided and some open), or unstructured (free flowing). Interviews can be done in a face-to-face format, or via telephone or video conferencing.

• Observations

Researchers observe participants or subjects under natural circumstances in order to collect real time data about behavior or event-based information without involvement or over-surveillance.

• Experiments

In an experimental study, one or more variables are altered in a controlled setting and the response(s) of participants is recorded as evidence in order to establish cause-and-effect relationships results.

• Focus Groups

Small groups of individuals are recruited to come together with a moderator to discuss specific issues, using a structure to record comments. This changes the form and allows the researcher to see shared insights, views, emotions and attitudes.

The process of data collection that uses secondary data is the process of using data that has been found and collected by other people that who likely collected it using different methods or purposes, where the researcher considers the relevance of those data to their own study. It is also a cost-effective and time-efficient method.

• Published Sources

This category includes books or other published literature, including peer-co-reviewed articles, magazine articles, blogs, and other media (e.g., newspapers, reports, government reports, etc.) containing credible and valuable information.

• Online Databases

Databases often refer to ones that house academic projects, such as related educational websites and databases such as online databases. Depending on the platform, an online database can include the literature and research available and identified as pertinent information before accessing the literature or studies that are also key to your research. It can often be a statistical record of some data collection, such as surveys or market trends.

• Government/ Institutional Record

Usually, government departments and research organizations that offer more detailed datasets and ongoing archives, such as even census data.

• Data from Websites

Public data from websites such as research blogs/publicly available information sharing, and communities that share data publicly

• Previous Research Studies

Previous research studies and a re-interpretation of their findings offer other studies rich in data that can potentially be recreated or built upon for a new related piece of research.

Having covered the methods, let’s now look at some individual instruments in data collection – in particular those of primary methods:

• Word Association

The respondent is provided with a stimulus word and asked to provide the first word that comes to mind (cognitively or emotionally). This will give the researcher a view into the respondent’s associations.

• Sentence Completion questionnaires

The participant is asked to complete a series of incomplete sentences, both of which would offer the researcher insights into the respondents’ attitudes, beliefs or preferences.

• Role-Playing

The respondent is given hypothetical situations and asked to respond to them, giving more insight into their behaviours.

• In-Person Surveys

With in-person surveys, there is an advantage to discussing clarification and following up, but it is also a lengthier process.

• Online/web surveys

An easy-to-distribute and analyse tool, however, their use relies on the user and their willingness to respond accurately.

• Mobile Surveys

Sent via SMS and/or mobile applications are equally effective for data collection as asking someone in person to assess their views quickly.

• Phone Surveys

Generally conducted by external company. However, response rates could be low based on unresponsive survey respondents or no call screening activities.

Direct observation

Observation of the subject’s behaviour in real-time is an observable behaviour within their environment – it yields unfiltered data as the researcher is observing an uncontrolled setting. Best approach for a small study where subjects are not aware of observation activities. Since subjects are unaware, they can act authentically.

Protecting the integrity and related processes of data collection is important to yield trustworthy, valid, and scientifically correct research results. Whether it is unintentional (random or systematic mistakes) or intentional (deliberate falsification), it is equally important that the processing data collection be protected. This is where quality assurance and quality control comes in.

These strategies are distinct in their purpose and stage of implementation in the research process:

Quality Assurance (QA) – Before data collection begins; prevention services.

Quality Control (QC) – During and after data collection; detection and correction services.

Let’s look at these two with more detail.

The goal of quality assurance is to prevent mistakes from occurring in the first place. QA ensures that the entire data collection process is clearly specified, and consistently documented, in advance, typically in a detailed procedures manual. An effective manual provides the research team with adequate knowledge and expectations throughout the experience, and can mitigate impactful mistakes from being made.

Badly designed QA procedures can result in the following:

• Unclear division of roles and responsibilities for training (and retraining) data collection staff.
• Incomplete or unclear lists of data items that are to be collected.
• No system for recording procedural changes during the study.
• Unclear description of the tools and methods employed, instead of providing detailed instructions ( i.e., in steps).
• Unclear understanding of who will review the data, when, and how?
• Poorly articulated procedures for data collection instruments, including the initial and ongoing calibration and maintenance of products.

Recognizing and addressing these omissions ahead of time, quality assurance puts you in the best position for ongoing, consistent quality data.

Quality control targets error identification as they happen and correct mistakes, if the process is done well, development of clear QC practices must be available in the procedures manual. Rapid and effective response when there are QC issues is dependent on the investigators and field staff developing a good communication way of working together.

Common qualities include:

• Staff observed collecting data.
• Regular reviews of data reports to catch inconsistencies, outliers, or invalid codes.
• Regular check-ins, visits, or video calls.
• Audits of qualitative and/or quantitative data records to verify compliance with protocols.
• If any problems are identified, corrective action must be taken quickly, retraining, omit, change procedures, rules, tools, or processes are possible examples.

Generally, data collection problems that are warranted immediate QC action can include:

• Fraud, or unethical behaviour.
• Repeated violations or systematic mistakes to protocols.
• Wrong or inconsistent data field entries.
• Inadequate performance, or dynamics with particular members of the data team, or data collection sites.

One way that researchers can help establish reliability when conducting research in areas such as social and behavioral sciences, where human subjects are involved, is through secondary or multiple validation measures. For instance, researchers do not use themselves in a survey of young adult risky behavior. Therefore, in order to validate their responses, they may ask cross-check questions or collect other subsequent measures for comparison to establish reliability.

Quality Control (QC) refers to the process of continuously observing, assessing, and remediating a situation during and after data collection. Effective QC dictates that all procedures be provided in detail in the data collection procedures manual, including specific steps for communicating between personnel and researchers, processes for escalation if there is an issue, and documentation on corrective action being taken. It is important to ensure that there is a communication structure that enables personnel to report when there is an issue as soon as possible during data collection. If the structure of communication between investigators and field staff is not formalized, there is a potential for weak monitoring and no action or awareness to troubleshoot or do a resolution for an issue or error in study follow-up.

Examples of Quality Control include:

• Observed personnel at data collection
• Attending the visit site/research participant via video calls to assess perform
• Routine audits of data for consistency, missing values, and invalid codes
• Routine review of the final data report, to highlight deviations or unexpected patterns.

While site visits may not be possible in all research domains, continued auditing of qualitative and quantitative data is necessary to assure the protocol was adhered to.

In addition to being able to identify problems, QC is also able to identify the timely and appropriate corrective actions to inform and improve procedures and avoid recurrence. Some data collection problems demand immediate follow-up when identified, including:

• Fraud or unethical behavior
• Systematic errors or capricious disregard of protocols
• Incorrect or inconsistent entries of data
• Underperformance of identified staff or site

In disciplines such as social and behavioral, when data often involves human subjects, it is common practice for researchers to incorporate a second act of validation to ensure self-reported data is accurate.

For instance, a researcher studying risky behavior in a young adult population may have clarified the fit of the follow-up questions for their continued trust after an individual being observed had admitted they may have provided erroneous and distorted responses. Or, a researcher may include a measure of cross-validation of data, which would allow them to unpack social influences behind reported behaviors.

After collecting the data, the next step is to turn that raw data into informed insights. This process is a necessary step to reach quality conclusions and allow sound decision-making.

1. Process and Analyze the Data

After data collection, the first step is to process and analyze. Processing can be an analytic method, such as statistical methods to discover trends or patterns (quantitative data), or to guide thematic analysis to investigate context and meaning (qualitative data). The goal is to turn raw data into informally interpreted data, which can lead to decisions, policies, or planning.

2. Interpret and Present the Findings

After data analysis, once data are analyzed, the next step is to interpret and present the findings. This can vary depending on the audience:

• Researchers may prepare academic papers.
• Monitoring & Evaluation teams may ask for longer analytical reports.
• Field teams may want a quick dashboard or would prefer the real-time data view.

The most important thing is organized clarity and make the products accessible to users with clear conclusions so that anyone can use the results easily/effectively–that is, present the results in a way that stakeholders may act on.

3. Effectively Store and Manage the Data

Data integrity is not achieved by just performing an analysis on it. Once you interpret the data, you need to consider data storage. This includes:

• Secure cloud storage platforms.
• Performing regular backups.
• Limitations on access to data with access protocols to prevent unauthorized access and unauthorized users accessing sensitive and protected information and data.

Overall, the use of these storage and data management practices can effectively store your data safely, keep it from being jeopardized or lost, and keep it secure from others accessing it if it is sensitive information as well.

Regardless of the level of sophistication in your plan, you may run into obstacles related to data collection, and it’s important to review some of the most common ones:

1. Issues of Quality

Low-quality data is one of the most significant impediments to practical analytics, and especially AI and machine learning. Ensuring that you control for dirty, unreliable, and inconsistent data should always be your most elevated concern.

2. Inconsistent Data

Each data source may contain the same data, but it could be recorded in different formats, units, and spellings. Inconsistency can happen for a number of reasons, particularly in the world of analytics where you have a number of data sources that come together, merge, merge back or shadows of once-mارية are altered due to business’s established procedures.

3. Data Downtime

Business and revenue opportunities may stall or decline due to the temporary inability to prepare, analyse or interpret your data accurately. When data is unavailable, unreliable or inconsistent for any length of time (known as data downtime) it can halt your business operations, slow down decision making, and specify your trust in predicted outcomes. Problems may occur  due to changes in schema, software, systems or simply a lack of care in maintaining your historical and current data pipelines.

4. Ambiguous Data

Ambiguous data in large analytics instances arises through spelling errors, a misspelled column or bad formatting. It is worthwhile determining whether there was ambiguity in your colony or known through data filing. Ambiguity presents itself in trusting accurate reporting and interpret data accurately.

5. Duplicated Data

Duplicated data occurs when we, good as SaaS provide the same data in various clouds (for example) for many different ways – just like when we duplicate across multiple sources (sources of truth). Duplicated or buffered data lends itself easily toward redevelopment and tests based on impacted business as a customer can lose everything they see in that zone and affect highly dysfunctional hiding in our learned behaviour process.

6. Too Much Data

The rapid growth of big data is a boon, but sitting through too many irrelevant data points can be problematic. Analysts are now spending 80% of their time on preparing some data points.

7. Data Accuracy

Some sectors such as healthcare or finance are unforgiving when it comes to data accuracy. The slightest mistruth can result from human error or system drift, but inaccurate data can result in poor decision making and regulatory non-compliance.

8. Data Hidden/Data Silo

Data that worth something, every organization locks up data/most of the current organizations’ data is hidden away in silos, unshared. For example, if the sales team does not share up-to-date customer data with the support team; we lose context which ultimately results in a negatively impacted customer experience and intelligence to the business.

9. Relevance

Relevance is very difficult. Not only do I need to find the right data; I have to consider:

• Domain
• Demographics
• Time
• Variable
• Rummaging through irrelevant data is inefficient, poor analysis always leads to frustration and wasted effort.

10. Data Don’t Know What Data

Most of the time when you don’t define what data you want to collect from the outset, you are usually doubling up on work, collecting irrelevant or useless data or worse just abandon the project.

11. Big Data Management

You cannot manage large amounts of data, using the same technology and processes you used when it was significantly smaller and simpler. Moreover, you should be moving away from traditional systems of data processing; towards new and more efficient technologies for storage, analysis, visualization.

These following best practices will help maximize and ease the data collection process:

1. Consider the total cost for each data point

If you’re asking a question—or asking for data—you should consider that the data point has a cost (in time and money) attached to it. Only collect data that is important.

2. Plan for accessibility and feasibility

Some data you identify will not be easily accessible. Some data will be sensitive or confidential, and if applicable, constrained by the type of data collection. Remember that some states will have barriers in the collection of certain data types! Consider how easy it would be to collect each type of data.

3. Consider mobile data collection

Some mobile data collection options may make your data collection process easier, such as:

• SMS – text surveys
• IVRS – automated phone calls
• Apps – field agent enters data on handheld

Choose the best for your audience and context.

4. Think relevance

Ask yourself a few questions:

• What do I need to know?
• What data do I have available?
• What is of direct use to my inquiry or strategy?

5. Don’t forget about the identifiers!

If you plan on using any data or hybrid data for future analysis, e.g., action planning, then make sure you have metadata (e.g., location, date, demography) that will be useful to “situate” the data and keeps track of its sources.

6. Consider using mobile data tools!

Smartphones with inexpensive apps allow for faster, cheaper, and even more accurate data collection. Low-cost smartphone proliferation and all of the available apps have made mobile data collection much more accessible and are a preferred option.

If you’re attentive to common pitfalls and receiving some guidance with best practices, you can adopt processes that are efficient, trustworthy, and bring powerful results for more meaningful endings in research, business, or policy.

Today, data collection is at the heart of data-driven decision making—whether you’re working in business, research, healthcare, or policy-making. From figuring out what data to collect and how to collect it, to maintaining data quality and dealing with data problems, every piece of the data collection process is important to the outcome of your project.

When we think about or actually incorporate quality assurance and quality control in our approach, we can ensure the reliability of our data, as well as its integrity. By confronting issues such as inconsistent data, data overload, and low response rates early-on, we can better save time, save money, and strengthen the credibility of our analysis considering our data is quality data.

Furthermore, as our tools and technologies have changed, so have the methods and approaches to data collection. In many cases, we can now increase the efficiency of our data collection methods (e.g., mobile data collection, automated data collection, and cloud-based approaches), and more importantly, expand opportunities for higher levels of scale and speed.

So, as we move forward with more and more data collection in the world, successful data collection ultimately comes down to more than just “data collection:” it is about collecting the right data, in the right manner, analyzing it in a meaningful way, and applying the information we have gleaned to improve the likelihood of a better outcome.

By following sourcing steps, best practices, and being aware of pitfalls, you and your organization now have the ability to intervene in your agency’s effective use of data.