Tips on data analysis – Samples

Data analysis is only as good as the quality of the raw data you put in. To get meaningful results, you need to make sure you are analysing the right information. The following questions should help you make sure you are on the right track.

Do you have the right sample?

If you’re looking for information about a specific group or product, make sure your sample is relevant. For example, if your market is retirees, a poll of university students probably won’t yield much relevant information. If you are looking at broader trends, you probably want a true random sample.

Is your sample really random?

This is a common stumbling block for everyone from students to professional pollsters. Many factors can bias a sample—perhaps you’re only collecting data in one postcode (people in any given neighbourhood tend to have similar lifestyles), or are soliciting responses from a magazine whose readers have strong political views in a certain direction.

Is your sample big enough?

Thirty data points is the bare minimum you can use for a significant result. A larger set will return more reliable information. However, the benefits of increasing your data set size decrease after you’ve collected several thousand data points, so you don’t need to go overboard.

Metapack Delivery Conference Event Summary

Metapack’s Delivery Conference was held at the Business Design Centre in London February 7th.

Clear Returns’ highlights of the day included an insightful introduction from Andrew Starkey, IMRG ’s Head of E-Logistics. Some key statistics came to light from his presentation including:

– 671 million online retail orders were made in 2012
– Click & Collect use may increase to as much as 50% of orders by 2015
Mike Newnham, Chief Customer Officer from the Royal Mail then revealed the results of their survey conducted over the Christmas delivery period. Some interesting insights into customer behaviour during this time surfaced from their findings.

– 15% of online shoppers return Christmas items
– 71% of customers think free delivery would encourage them to shop more

Towards the end of the day we heard from Clare Gilmartin, VP Marketplaces Europe, Ebay. Her insight into changes the e-commerce industry will go through was also a terrific presentation. Her predictions include:

– Mobile connections will surpass PC connections globally
– China will surpass the US in terms of e-commerce growth in the next 3 years

Clare also highlighted that flexible returns play a role in purchasing decisions. She stated that 63% of online shoppers look at return policies prior to purchasing and 50% shop more often with a retailer and are more likely to recommend a friend to that retailer if they have a lenient return policy. This seems like a rather risky message to send.

Nick Gomersall, Innovation and Communications Director at Metapack , delivered another brilliant presentation. His focus on back-end fraud enlightened the audience to the news that this cost is currently not being measured, which can be as much as 1-2% of total turnover.

Retailers throughout the day were being encouraged to offer free, fast, convenient delivery and lenient, flexible return policies to satisfy customer’s demands. They were also told that if customers overbuy they often end up keeping more items than they intended to. Furthermore fraudulent customers who claim their parcel wasn’t delivered tend to get the benefit of the doubt and receive another parcel immediately.

Our research shows that customers often exploit lenient returns policies to overbuy and use their bedroom as a fitting room or engage in ‘wear and return’ behaviour. If free delivery is added into the mix this kind of behaviour will almost certainly be encouraged.

Retailers need to be extremely careful and cautious before they deploy these types of strategies. Even though free delivery and returns may encourage greater sales, the true cost of around 30% of these items being returned, repackaged, reshipped and restocked must be considered.

Tips on data analysis

The latest addition to our team and our resident expert in data analysis and predictive modelling, Shaylon Stolk, shares her knowledge of statistics. By understanding these basic analysis tools data can seem much less complex!

Mean

Most of us are familiar with the arithmetic mean or average, or a set of data. This is simply the sum of the data divided by the number of points. If you have a dataset with a normal distribution, this will be a representative typical value for your dataset.

Median

If you have a data set with lots of outliers (numbers which fall far outside the expected values), a dataset with lots of variability, or a non-normal distribution, it may be better to take the median than the mean. Arrange your data points in ascending order by value, and take the middle number in the sequence.

Mode

Taking the mode of a group of values identifies which value occurs most frequently. In a normally distributed data set, this will be the mean value, or extremely close to the mean value.

Standard Deviation and Error

Identifying the level and significance of variation in data trends is not particularly intuitive. To make these determinations, you’ll want to understand both the standard deviation and the standard error.

The standard deviation is a measure of ‘normal’ variability. A data set with a lot of variation will have a larger standard deviation than a data set with less variability. 70% of your values should fall within one standard deviation of the mean.  A value which falls within 2 or 3 standard deviations can be considered an outlier.

Error describes the level of uncertainty in a set of results. Typically, the larger the sample size, the smaller the error. An error of 5% or lower is usually seen as indicating statistically robust results.

BRC report on the increasing cost of retail crime

A recent report from the BRC (British Retail Consortium) found that the cost of retail crime has continued to rise significantly, with the overall cost in 2012 totalling £1.6 billion.

What is particularly worrying about this increase is that e-crime is costing retailers the most, as it now accounts for 37% of all retail crime totaling almost £600 million. Although it was not made clear what their definition of e-crime consists of. Part of Clear Returns’ own research involved identifying issues of fraud such as wear and return. As the online shopping channel grows,  Drapers announced last week that online sales in 2013 are expected to reach £87 billion, this type of fraudulent behaviour could increase in frequency.

Retailers are increasingly being pressured to enhance their multi-channel strategy in order to survive and compete successfully, therefore security measures and protection must now become one of retailers top priorities in light of this new information. Perhaps we will see more companies investing in technology such as the Clear Returns system to help effectively manage this problem throughout 2013.

Vicky Brock delivers Google Analytics Workshop

Clear Returns’ CEO and Google certified partner Vicky Brock gave a compelling presentation on Google Analytics this week to a group of eager attendees.

Discussing several issues surrounding data analysis such as cookies and privacy laws Vicky provided some useful tips to make the most of the tool. She also discussed the tool from several different business perspectives, such as an ecommerce or a lead generation viewpoint.

Audience members left with key pointers to take away and apply to their own businesses including:
-View mobile and non-mobile traffic to understand how people are using your site
– Introduce your own custom activity e.g. social media
– Exclude internal users of the site to avoid confusing the data
– Viewing the search terms used on your site can show unmet demand from customers

Clear Returns hope to hold more events like this in future, if you would like to keep up to date with all our events please sign up to our newsletter.