How to increase sales. Big Data – instruction on application

In recent years, to the surprise many began to talk about Big Data. It is strange to hear narrowly specialized term from salaried analysts of filling stations networks, shops, restaurants, hotels, electronic trading platforms.

I have never met a real business that has independently developed the competence of processing huge volumes of unstructured information. It is time to master the techniques and tools of Business Intelligence for the detection and implementation of strategic opportunities.

A considerable variety and high growth rate of Big Data allow to regularly receive useful results, effective in the condition of continuous growth or oppressive uncertainty. This technology enables to: provide 98% collection of parcels for online store in 50 minutes, speed up of goods passage in the warehouse by 11%, reduce the loss of fresh fruits and vegetables by 8%, raise the net margin of stock shops by 7%.

Internal staff thoroughly know details and peculiarities of the business, write heaps of clever reports, generate excellent deep hypotheses in the area of their competence, can recalculate mountains of accumulated values, but it has nothing to do with the well-known big data. When owners and top managers show expensive hardware and software platforms and the results generated by them, there is a candid desire to cry bitterly and loudly. Were spent substantial funds, recruited squads of specialists, proclaim bold promises, but practicality of transferring raw data into meaningful and applied form is discussed as a reason of future uncertain time.

Talking about Big Data makes sense, if you:
- understand what it is, and have them in reality
- are able to form hypotheses and know the methods of productive processing
- are willing to expend considerable effort to analyze and apply the results in real business.

Do you want to interpret large amounts of data, focusing on the key factors of efficiency, simulating the outcome of various options for action, tracking the results of the decision-making? I would recommend to start with acknowledgement that you do not have Big Data. Big data sets are not available to employees of the enterprises because companies do not spend efforts to collect information of dubious utility. Here are some real examples of initiation of collection and processing of large data with actuality no more than six months.

Known network of filling stations agreed to collect data on the climatic conditions at its facilities, conducted survey of twenty theses for employees, started to request from servicing bank lines of payment transactions by payment cards and authorization codes, to enrich directories of sold goods with indicators:
- energy value
- how many units can be carried in a hand
- fragility of packaging – does it break when falls
- sensitivity of product – resistance to compression
- the position of the weight/volume of packing series
- the percentage of packaging in the weight of product
- component composition of proteins, fats, carbohydrates
- opacity of packaging– such goods are associated with health
- sonority of packaging – thunders in the vehicle when is shaken
- the nearest place of possible consumption – at the station, in the vehicle at home.

Now the owners of the network are the happy holders of large data sets on the employees, products, routes of fueling vehicles, behavior and preferences of customers. The possibilities of analytics widened unprecedentedly – we were able to create a joint BI processing of external data with internal information sources, that created a more complete picture and formed “business-mind” – comprehension that cannot be derived from a simple data sets of standardized accounting systems. Trends observed in the first year allow to suggest an increase in margin by 4% as a result of the implementation of found strategies.

Online fashion boutiques on my request began to record:
- selected clothes and footwear, as well as the order of their fitting
- brands on brought in bags with earlier purchases
- announce reason and cause of purchases – for one-time event, for every day, to a trip.

It was possible to calculate the dependences, many of which with its irrationality caused a healthy skepticism among sellers, but nevertheless in total raised sales by 16%.

For example, early termination of women’s visits, that try on more than four pairs of shoes allowed to:
- gradually displace this type of customers from stores that unloaded sellers by 19.1%
- increase the likelihood of purchase by women around by 14.7%
- reduce stress on employees from failed sales.

For restaurant chain we collected the following information:
- treatments, that are used to call the waiter
- amount of tips and denomination of banknotes
- which places at the tables are occupied by guests
- agree to use freshly ground pepper to spice cooked meals
- sequence, which is requested for serving dishes or lack of wishes on this occasion.

By combining new data with already existed:
- dishes ordered by adjacent visitors
- tables occupied at the time the guests arrive
- previous orders tied to loyalty or payment cards,
managed to make more accurate predictions:
- Chef – the volume of products to optimize purchasing prices
- waiters – where to sit guests and what to offer from the menu to get more tips
- owners – the forecast return on the dining outlets and the need for opening new establishments.

Analysis of hotel services consumption provided personified recognition of visitors and helped to link them with left comments on popular sites for travelers. This is allowed to distinguish very small population, affecting which by the service of increased attention the hotel managed to significantly improve feedback left on websites, which raised a hotel in the rankings and made it possible to significantly raise prices.

The large online store implemented with our assistance mechanism of splitting screen into squares to track the route of the mouse pointer between the elements of a virtual grid with the measurement of time spent in each quadrant. For the quarter were collected and compiled hundreds of thousands of samples of “handwriting” moving in the visual field. Routes, subjected to the clusterization allowed to allocate 21 styles of behavior among site visitors and to develop for them specialized mechanics to increase the likelihood of purchase completion.

Involvement of external experts in Big Data allows to achieve results that exceed expectations:
- working with different businesses in different sectors, professionals have already made a sufficient amount of iterations for other people's money, and it is better to learn from others' mistakes
- making projects in the adjacent fields, learned and tested set of promising and not working hypotheses that will save time for brainstorming during tasks setting
- colleagues managed to accumulate vast collections of methods, algorithms and procedures of cleaning and normalization of data sets, which as a rule increases the accuracy of forecasts by 13.8%.

Do you think it is early to think about Big Data? Then at least provide yourself with Clean Data – unify text strings of directories. “Double Espresso” write in the same way, do not store the 11 spellings of city “St. Petersburg” and 7 variations of the profession “accountant”. Otherwise, your analytics has unacceptably rough errors, its conclusions make significant harm, and you are looking for the causes of failures in random areas of unreality.

Author: Oleg Braginsky
Translation: Daniil Shmitt
Source: New Retail