According to the famous quote by Bill Gates, Founder of Microsoft. He says: “If your business is not on the Internet, soon your business will be out of business.” This statement by Bill Gates summarizes the importance of every business, regardless of its size or composition, the need to establish an online presence.
The need to establish business footprint on the internet and numerous data generation channel gave rise to stupendous data generation statistics.
These large data come with large reservoir of insight for business.Are businesses leveraging on these data? How can small businesses tap into these bank of opportunities?
Why should entrepreneurs utilize data science?
For organizations to succeed nowadays, they need to nail the personalization piece, and business owners fully grasp this a lot more than most. When business owners start businesses, among the things that are first consider is how they are able to provide their customers, however as it pertains time for you to scaling just what the customer wants, it gets very difficult to conceptualize what sort of individual would accomplish that. This is where machines are available in. Personalization at scale is exactly what we see most from entrepreneurs. With retail, as an example, we are trying to create models that predict what individuals want considering the way they’re browsing items. So we take a product feed and look at the characteristics of what folks are put and browsing that together with what others have actually browsed with comparable choices. With that, we can make recommendations that are really informed. How many various characteristics can find out from a shirt or from an accessory? How do we put things together and [feature] auto-tagging and all sorts of of these different things that businesses need and want?
How do you explain the protocol of personalization from a data-science perspective?
Personalization is about identifying a product or service that we think someone needs and displaying it on the dashboard of the individual. How possible can that be achieved? What are the procedures required to achieve this? It has been observed that most organizations that claim to provide personalization services are only providing customer segmentation. They are found to be segmenting the customer base into smaller segments. Although it’s a mean to the means but not the target intelligent use of data.
We need to be in the predictive model, not the descriptive model when dealing with personalization. The predictive model involves identifying the historical traits in identifying what the future trends will look like. The is the right mode where the really interesting things are discovered as the precursor to recommender systems.
Why are small businesses not leveraging opportunities in data science?
When it comes to the usage of Data analytics, SMEs are often left far behind. According to a study by Coleman, S. et al (2016) conducted in the United Kingdom in 2012 SMEs have an adoption rate of 0.2 % while the same rate was about 25 % for businesses with more than a thousand employees – Coleman, S. et al (2016). Similar studies conducted in Germany and Austria by Seufert, A(2014) revealed similar results. In Germany, 46 % of all interviewed SMEs indicated that they are not even going to use Big Data in the future – Seufert, A(2014)
What are the constaints preventing small businesse from adopting data science?
The negative disposition of most small business to the adoption of big data analytics was found to be one or more of the outlined reasons. This ranges from security considerations, financial restrictions, lack of knowledge, and lack of prioritization for business issues – Michael Dittert & Christopher Reichstein (2018)
Especially the categories lack of knowledge and financial restrictions are closely interrelated. Many of the SMEs are sheer domain specialists, so there is no awareness of new concepts like Big Data . As a result, there is no deeper understanding of Big Data techniques. There is a lack of support that could help developing skills in Data Analytics[3]. In addition, there still is a demand for case studies that can help to exemplify the benefit of Big Data to them. On the labor market, there is a shortage of data analysts. Hence, most of the SMEs are not able to recruit experts who can help to use Big Data. As there is no in-house expertise SMEs must rely on external expertise. Both, recruitment and external consulting are very expensive. Due to their financial restriction, SMEs often cannot afford to take advantage of Big Data – (Coleman, S. et al, 2016)
Cross Industry Standard Process – Data Mining (CRISP-DM) Framework
In order to ensure smooth adoption of data science by businesses a group of data mining experts from Integral Solutions Ltd (ISL), Teradata, Daimler AG, NCR Corporation and OHRA, developed an open standard process framework, Cross-industry standard process for data mining(CRISP-DM), which describes common approaches used in data mining by data mining experts,
The CRISP-DM consists of the following six steps
- Business Understanding: The aim is to establish and define a concrete business objective, it is necessary to gain some expertise about the business
- Data Understanding: This step comprises the collection and analysis of the data. The analysis can be supported through visualization.
- Data Preparation: This involves the preparation of the data in the format acceptable for data modeling.
- Modeling: A model is applied to solve the Data Mining task. The best model and its parameters are chosen in a trial and error procedure
- Evaluation: The model must be evaluated in quantitative and qualitative ways. Therefore it must be asked, how accurate the results are and if they are suitable to fulfill the defined business objective.
- Deployment: The results must be used in a profitable manner. The easiest way of deployment could be, simply to report the results to the decision-maker, – Michael Dittert & Christopher Reichstein (2018)
The steps need not be followed in a rigid format.
SME Data Science Adoption Process Frameworks .
The SME process farmework is closely related to the CRISP-DM process.The SME data science process is stated below:
- Define a task: Ideas for an appropriate Data Mining application should be generated by the SME’s top management. This can be done internally based on workshops or externally based on consulting support. The collected ideas have to be evaluated in terms of cost-benefit considerations.
- Collect and analyze the data: The activity of collecting the data is stressed by putting it in the first place. Unlike big enterprises SMEs often do not have well-organized databases or even data warehouses themselves. This makes it necessary to set up a data basis for each Data Mining task.
- Choose and set up a model: Depending on the task defined in the first step, a model must be chosen. For some tasks, like for class prediction, there are a lot of models available.
- Format data: First of all, it must be checked what types of data the chosen model can work with. While models like a decision tree can work with nearly every kind of values, a neural net for example only accepts numerical values. If the values can’t be changed into a numerical scale, they have to be excluded.
- Evaluate results: there are several possibilities to evaluate the model’s quality. They depend on the kind of task that was performed. A classification model can be evaluated via split-validation, where the dataset is randomly divided into two parts.
- Report to decision-makers: The easiest way of deployment is to support the decision-makers with the knowledge gained from the Data Mining activities
What’s next with data science for SME?
The above-stated data science implementation process caters for most of the identified data science adoption constraints earlier stated. When these are followed appropriately, small businesses can tap into opportunities currently available or being used by big enterprises in uplifting the performance of their SMEs.
