booth branding business buy
Photo by Pixabay on Pexels.com

Retail analytics involves making sense of data to augment decision-making processes that impact profitability. Retail data collected typically comes from operations, and this data can come in a structured or unstructured format. At times, internal data sets can be combined with external datasets to augment the analysis. The goal of retail analytics can be to increase sales revenues or cut cost (or both).

With the proliferation of sensor and cloud technologies, it is now easier than ever to collect and analyse data from various sources. Unfortunately, this competitive advantage has also left many behind.

About this guide

This guide is targeted at the small and medium retail businesses who are thinking through an analytics implementation and where to start. We wrote this as a crash course for those interested in learning more about retail analytics and some of the thought process that goes behind selecting some tool.

This guide will not…

  • Go through highly technical formulas
  • Make you an expert on the topic
  • Market a specific tool for you to use

But, if you get through it, you will be able to…

  • Speak confidently about the subject
  • Better weigh the pros and cons of the tools out there
  • Select the one that best suits your workflow

 

Why we wrote this: Full Disclosure

Haste is a data science platform for small and medium-sized retail businesses.

Our mission at Haste is to provide a platform for retailers to write their next chapter of growth. To us, this means providing the tools so that retailers can be at their best all the damn time.

 

You don’t need an analytics tool…

You need a solution to your current problem. Analytics is a tool that is meant to help, but if you don’t have the problem statement figured out, analytics is not going to help. Don’t even start searching for tools until you have decided on what problem in your retail business you want analytics to help solve.

Stay open-minded in your research on this topic. Don’t just listen to vendors who want to push their retail analytics solutions to you and don’t get catch up in the stories of your peers where they tried using analytics and it failed. Figure out why it didn’t work for them and whether that applies to you.

 

Once you have a problem statement in mind…

person writing on white book
Photo by rawpixel.com on Pexels.com

Don’t rush out into the market searching for an analytics software just yet. It is going to be expensive because research without preparation just burns time and you are going to feel jaded after that. Think through this framework-

  1. What problems do I want to solve? (done)
  2. What sort datasets do I have on hand?
  3. What is the general level of expertise of my team?

The next few sections are solely dedicated to explaining these pointers and you can skip them if you already know the different types of data sets and have a good gauge of the analytics level in the market.

 

Data Sets

Before jumping the gun and researching the best analytics solutions for you, take a quick look at what you have that is helping you store and manage data. On the most basic level, you should have a point-of-sale system that logs your transactions and merchandise information. Other sources include-

  • Customer Relationship Management Software: Shopper
  • In-store Sensors (Offline): In-store Traffic and Behaviour
  • Inventory Management System: Product
  • Google Analytics (Online): Website Traffic and Behaviour

Each one gives you a different dataset which you can then use for analytics. We will run through some examples in a later section but there are two important points to highlight before moving forward. You have to ensure that the data is accessible and clean.

 

Accessibility-

The data that you generate and collect for your retail operations belongs to you. Regardless of the vendor, you should have been given the ability to extract the raw files for your own work. Take the point of sales system as an example, the data stored there might be required for accounting, inventory management and analytics.

Word of caution- there are some third-rate vendors out there that will not allow you access to your own data. Stay away from them.

Be careful when you are selecting a point of sale provider. Find out a provider that allows you to either extract the raw data (in a format that you want) or one that has an open API allows you to cheaply integrate into the multiple tools that you will want to use. If you are have already committed to a vendor and they are not all that you thought they would be, change. Now. Before it is too late. There are lots of cloud-based solutions that don’t cost more than $100 a month.

 

Cleanliness-

We have all heard phrases like “Data is the new oil”. An analytics platform is like a refinery that helps you make use of it. A refinery isn’t going to work as well if your data set isn’t as well maintained. If the dates are messed up or the products are mislabeled, the data is completely useless. You can either spend some time cleaning it or just trash it because (depending on the state), it is not usable.

Word of caution- there are some third-rate vendors out there that do not have a proper database structure. Run as far away as possible.

It commonly happens because they didn’t come from your specific retail domain and thus might not understand the complexities behind it. Take for example if they have been catering to bookstores. If they were to sell the same system to a specialty meat grocer, they will suddenly have to capture data like the price per weight which isn’t something that their database was engineered for. This will likely mean some shoddy patchwork to get a system up and running which can lead to complications down the line. Run away. Early.

 

Team Expertise

Now that you have the data set prepared and the raw data ready, it’s time to take a good hard look at the team. This guide was written for the small and medium retailer and if you are one of them, you know that good talent is hard to come by.

You might have an amazing team that you love to work with but we all have to face the reality that individuals who possess strong analytical skills are in hot demand and hard to recruit and keep. So, how committed is your retail business?

No one is committed full-time to analytics…

If this scenario sounds like your business right now, you are not alone but you should start questioning the sustainability. You might be allocating half a day a week to this but realistically retail analytics is a full-time job.

You will need time to research and learn the relevant formulas that can apply to your specific domain. We know that business analysts are hard to come by and if you are stuck doing the work, consider purchasing some simple software tools to help you get by. (Microsoft Excel doesn’t count but should be good enough for the first three-nine months of operations)

 

One-two business analyst with a retail background…

Even though you feel strapped for time and you don’t have enough manpower, you already have more resources than most. Having analytics be one person’s full-time job gives you a HUGE advantage over your competition.

Notes about management-

We know the next thing on your mind is probably how to justify their presence. It isn’t easy because analytics and big data is probably not your area of expertise. It doesn’t matter. Smart analysts will be able to figure out a solution but you have to help.

You have to be able to clearly articulate the problem you want them to solve.

Listen to them and get tools to help them out in their job. At this juncture, I know you are probably against making another software purchase but think of it this way…

If you are paying between $60,000-90,000 per year for the business analysts on your team, how much does it really cost to invest in tools that can make them just 5% more efficient in their job?

 

Full data team, data engineers to data scientists, the full suite…

Out of all the retail businesses out there, you are probably within the top 5% but you know that this comes at a cost. With the average salary of $120,000 a year, the only way you can justify maintaining a team of 8 is that you are already very profitable and their objective is to find ways to grow the business 10% every month. If you are in this state, (1) Buy all the tools you can get your hands on to enable this team, even if they cost half a million every year and, (2) stop reading this guide because you are probably already making over $250 million a year.

How do you choose the best tools to enable the team?

The next thing on your mind after reading all this is probably, “So, how complex should the tools be?” Our advice is to let the end users choose because they are more aware of

  1. Limitations of the data available
  2. Their own limitations

They will not recommend a tool that is too complicated for them to use and they will know not to purchase software solutions that cannot work with the business data.

 

Different Types of Retail Analytics

We are going to assume that you are not a domain expert on analytics. So this section is going to be a brief description of what the heck is the different TYPES of retail analytics and where you should be focusing your attention on.

What you need to know is that they are not completely sequential but they tend to be in the order of increasing difficulty so there are advantages in first setting the foundations right.

 

Descriptive AnalyticsWhat has happened?

This is a reporting function that conveys to you historical data in a digestible format. It is like a health report for your business. In the retail context…

  • How did your sales perform in the last quarter?
  • Which products or categories didn’t hit their target?

Notice that at this level, some level of benchmarking will be required for you to understand whether what has happened is actually good or bad.

Pro Tip: You need to have internal targets set.

If the numbers showed you that you were growing your business 10% month over month, is that a good thing? If there was a slow down in one store but an increase in sales on your online affiliate channels, is that good news?

This is the most basic form of analytics and you shouldn’t move forward without it

 

Diagnostic AnalyticsWhy did it turn out that way?

On some level, this is like detective work. Figuring out what was out of place and searching for clues that can lead you to an answer.

A spike in sales can be attributed to a promotion that you ran, a new product release or maybe a seasonal holiday, you won’t know for sure unless you do some digging.

Say you discounted a particular item for a period of 3 months, sales skyrocketed- great! So you want to try the same trick for the next 3 months- this time it doesn’t work- why?

Maybe you were initially discounting a winter jacket and now its summer time- nobody wants that product anymore.

This may seem like a tedious process at first but once you learn the reason, you become capable of replicating a success (or avoiding a mistake).

Pro Tip: You might not find the answer on the first inspection so record your findings and test again later. BE VERY SPECIFIC.

Use a company Google Sheet or a group chat to record all these findings so that someone can pick up where you left off. If you have the suspicion that the X’mas break was the cause of a surge in sales, you can only test the same hypothesis next year so remember to record something like this-

  • Fashion line sales expected to increase by 450%, 2 weeks before X’mas possibly due to the promotional campaign launched during the last week of November.

When the holidays come around next year, you can try to run some advertisements 3 weeks before X’mas to see if sales really to spike up by that much.

 

Predictive AnalyticsWhat is going to happen?

Right…forecasting again, time to talk about inventory.

Well yes and no.

The most common response we get when talking about predictive analytics is managing and forecasting inventory requirements. That is just the basic and you can do so much more than just forecast inventory requirements.

If you did a historical analysis on your marketing conversion rates, you might be able to gauge how much sales you can drive with next quarter’s marketing budget.

Say if you knew that in the past spending $200 a month on paid advertising through channel A got you about 1,000 visitors. 2% of them converted and made an average purchase of $50. So, that’s $1,000 is sales generated from this channel.

If you are able to replicate this effect on sales with double the budget, this should be accounted for within your forecasting models because you will need to prepare your inventory for an influx of demand and ready your staff for the extra work they will have on hand.

 

Prescriptive AnalyticsHow can we make it happen?

This type of analytics pieces together different learnings from the others to offer suggestions or advice as to how retailers should act in order to influence an eventual outcome. This is really difficult to get right because of the multiple variables that can influence the outcome. Let’s say you have enough data and the right resources to execute if you knew…

  • Rainy days increase the likelihood of impulse purchases, you could stock more merchandise closer to the point of purchase.
  • That rap music made shoppers more likely to purchase snacks, you could move the snack section to a more noticeable location
  • Shopper are more likely to shop during Monday evenings, you could dynamically increase prices to capture more margins

Unfortunately, although prescriptive analytics has the most direct impact on business alignment, the investments required to achieve this has many retailers convinced that it might not be worth investing in.

Our opinion as of 2018 is that technology and cost structure will eventually improve but we agree that it might not be worth the investment especially if you are a small business.

You can still run experiments can gather enough results from the other types of data analytics to prepare yourself for the future.

 

So, when am I supposed to invest in analytics?

grey metal hammer
Photo by Pixabay on Pexels.com

If you have read this far, you probably have some idea of whether you are ready. Here is a guide to help you think through it.

Stage

Brick & Mortar E-Commerce Level of Analytics Requirement

0

1 store <100 monthly trx Not really required

1

1-3 store

100-5,000 monthly trx

Basic Business Intelligence Tool

2

3-10 stores

5,000-10,000 monthly trx

At least some Descriptive Analytics

3

10-15 stores

10,000-20,000 monthly trx

At least some Predictive Analytics

4 >15 stores >20,000 monthly trx

Small Team working full-time on this

*Notice the bar for e-commerce is set much earlier because data is much more readily available

 

Stage 1

You have gone past the seed stage and now you are a real business with real headaches. You probably are so strapped for time that you can no longer connect with all your shoppers and you can’t afford to be spending your whole day in front of spreadsheets.

Use some form of business intelligence tool to keep you updated about how the business is doing every day.

 

Stage 2

This stage is all another setting the infrastructure in place. With business growing steadily every month, you are going to want to have the right people and tools in place. Setting KPIs and metrics only make sense if everyone can be aligned to some business metric and every manager is working towards these goals

Use these tools to help your team get the work done

 

Stage 3

There are way too many things for one manager to handle at this stage. You are going to need to give them the right reports for them to make decisions else profitability is going to be heavily compromised. If you are still the central point of contact for decisions, you should be very worried. Set up alert systems and processes in place for contingencies when things don’t go according to plan.

Use tools that help you set alerts and plan for the future to get the work done beforehand.

 

Stage 4

If you don’t have all the analytics tools in place, it might already be too late. At this juncture, analytics is no longer a “good-to-have”. The benefits of optimisation compounds.

If at 20,000 transactions they find a way to save $1 or get an extra $1 per transaction, that’s $20,000 in profits. If they can repeat that process and your business grows 5% month-on-month, that’s approximately $300,000 added to the bank account in twelve months.

 

Great! I’m sold. Where do I begin?

We mentioned this WAY above. Problem statement first.

What problem do you need to address? Add in some details

  • Is this a generalised problem or domain-specific?
  • What is the ideal outcome that you are hoping for?

Here are some examples of how we broke it down-

Problem Domain Ideal Outcome Solution
Inventory is constantly in a mess because of over/ under ordering Operations Merchandise goes to the right place and time Demand Forecasting & Inventory management tool
Unsure about the payback period of a marketing campaign Marketing Ability to justify marketing spend with a fixed payback period Tool to track attribution from different channels and conversion calculator
Unable to see how different shoppers are behaving Marketing Engage different shoppers based on their different behaviours Use CRM data and form k-means cluster based on factors like when they shop and what they buy
Can’t figure out how to price a new product Merchandising Able to gauge sales based on price for new products Product similarity matrix and forecasting engine for different price points
Can’t figure out what products go well together Merchandising Display products together to drive upsells Basket analysis of compliments and possible substitutes

There are dozens of techniques to solve the same problem and you have to pick a solution that best suits you. Many make the mistake of falling into the technology hype and purchase a tool that doesn’t actually fit their specific workflow.

Say that you came across a brand new way of doing basket analysis that can sequentially predict what items a shopper is going to add to their basket.

We all think it is cool that you can use Bayesian techniques and machine learning to predict the order of purchase but unless you have the manpower to design your shop to match the order, you might be wasting your time with this tool.

body of water at cloudy sky
Photo by Nina Bogomaz on Pexels.com

Final Words.

Thank you so much for spending the time to read this entire guide, and if you have any questions or feedback, we would love to hear from you.

 

Good luck!

To measure is to know. — Lord Kelvin

 

Posted by:Fu Fei

Data Geek, Passionate about enabling entrepreneurship Figuring out how to gamify retail operations Twitter: @SymbolOfFlight

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s