Are you an online retailer? Does your team have a handful of dashboard tools that display colorful KPI charts but aren't able to answer the questions that drive your business forward? For example, how do we sell a higher volume of our inventory? How do we increase customer retention? How and where do we recommend products to increase sell through rate? The list goes on.
While your dashboard tool isn't able to answer these questions, a data scientist can. We put together a list of key analyses for online retailers as a checklist for certain models that can be applied to identify new revenue opportunities.
3. PRODUCT RECOMMENDATIONS
WHY THESE ARE VALUABLE:
By understanding which products are purchased
together we can recommend products to
individuals that they are most likely to purchase
WHICH KPI THIS IS GOING TO EFFECT:
Increase average order value (AOV) by better
matching customers to the products they want
to purchase
EXAMPLE OF A MODEL THAT IS APPLIED:
Collaborative filtering, association rule mining
TYPICAL DATASET TO WORK ON THIS:
User sales data
We’re here to help
Contact sales@datascience.com
5. PERSONALIZATION
WHY THIS IS VALUABLE:
Tailoring search results to individual browsing
habits can provide visitors a more streamlined and
enjoyable shopping / browsing experience
WHICH KPI THIS IS GOING TO EFFECT:
Increase conversate rate of purchasing
users by creating experiences tailored
to individual preferences
EXAMPLE OF A MODEL THAT IS APPLIED:
Bayesian multivariate optimization
TYPICAL DATASET TO WORK ON THIS:
Browsing history, sale, favorite, and cart data
We’re here to help
Contact sales@datascience.com
7. SEARCH/RANK OPTIMIZATION
WHY THIS IS VALUABLE:
Optimizing the set and order of items displayed for
search results and department pages can help users
find what they want faster and increase conversion
WHICH KPI THIS IS GOING TO EFFECT:
Increase conversion rate and site traffic
EXAMPLE OF A MODEL THAT IS APPLIED:
Conversion optimization
TYPICAL DATASET TO WORK ON THIS:
Impression and click analytics
We’re here to help
Contact sales@datascience.com
9. INVENTORY AND
SALES MODELS
WHY THESE ARE VALUABLE:
Knowing the likely amount of sales to occur in the
future can inform ordering and stock decisions
WHICH KPI THIS IS GOING TO EFFECT:
Increase sell-through-rate by better aligning
inventory with sales
EXAMPLE OF A MODEL THAT IS APPLIED:
Yield management
TYPICAL DATASET TO WORK ON THIS:
Sales transactions
We’re here to help
Contact sales@datascience.com
11. SEASONAL SALES
TRENDS ANALYSIS
WHY THIS IS VALUABLE:
Invariably, most stores carry some products that
ebb and flow in popularity. These seasonal changes
can be automatically detected
WHICH KPI THIS IS GOING TO EFFECT:
Achieve better planning and forecasting by
being able to understand the natural cycles of
your business
EXAMPLE OF A MODEL THAT IS APPLIED:
ARIMA and other time series approaches
TYPICAL DATASET TO WORK ON THIS:
Sales transactions
We’re here to help
Contact sales@datascience.com
13. PRODUCT TREND ANALYSIS
WHY THIS IS VALUABLE:
All inventory goes through spikes and drops in
popularity. How similar items have performed in
the past, and how an item has performed recently,
often provides strong insight into how it will
perform in the near future
WHICH KPI THIS IS GOING TO EFFECT:
Improve your sell-through rate by being better
able to understand, stock, and promote the
items that will sell the best
EXAMPLE OF A MODEL THAT IS APPLIED:
ARIMA and other time series approaches
TYPICAL DATASET TO WORK ON THIS:
Sales and impression data
We’re here to help
Contact sales@datascience.com
15. CUSTOMER COHORT DESIGN
AND SEGMENTATION
WHY THIS IS VALUABLE:
An evidence based construction of personas
describing customers can inform product
development about how people use a store and
what features would and would not benefit them
WHICH KPI THIS IS GOING TO EFFECT:
Increase average order value (AOV) by
more accurately positioning products to the
right segments
EXAMPLE OF A MODEL THAT IS APPLIED:
k-nearest neighbors, cosine similarity,
jaccard index
TYPICAL DATASET TO WORK ON THIS:
Browsing history, sale, favorite, and cart data
We’re here to help
Contact sales@datascience.com
17. EMAIL CAMPAIGN SEND
FREQUENCY OPTIMIZATION
WHY THIS IS VALUABLE:
Gain a better understanding of when to send email,
whom to send them to, and how to send them
WHICH KPI THIS IS GOING TO EFFECT:
Increase customer retention rate by maximizing
your ability to keep customers engaged
EXAMPLE OF A MODEL THAT IS APPLIED:
Multivariate A/B testing
TYPICAL DATASET TO WORK ON THIS:
Email open and click-through logs
We’re here to help
Contact sales@datascience.com
19. CART ABANDONMENT
ANALYSIS
WHY THIS IS VALUABLE:
When consumers add items to their cart but fail to
checkout, the inventory they leave behind can tell
us a lot about them
WHICH KPI THIS IS GOING TO EFFECT:
Decrease cart abandonment rate and increase
net sales
EXAMPLE OF A MODEL THAT IS APPLIED:
Price elasticity, ad re-targeting, email
promotion optimization
TYPICAL DATASET TO WORK ON THIS:
User events on site and purchases
We’re here to help
Contact sales@datascience.com
21. FRAUDULENT TRANSACTION
DETECTION
WHY THIS IS VALUABLE:
Identify indicators of fraudulent transactions so
that you can put preventative measures in place to
stop them before they happen in the future
WHICH KPI THIS IS GOING TO EFFECT:
Decrease chargebacks and recover more revenue
EXAMPLE OF A MODEL THAT IS APPLIED:
Bayesian networks, logistic regression
TYPICAL DATASET TO WORK ON THIS:
Sales transactions, web logs
We’re here to help
Contact sales@datascience.com
23. CHURN ANALYSIS
WHY THIS IS VALUABLE:
Understand what events indicate a customer will
churn and which types of customers are most likely
to leave your service
WHICH KPI THIS IS GOING TO EFFECT:
Incrase customer retention by being able to
idenitfy users that are most likely to cancel
their subscriptions
EXAMPLE OF A MODEL THAT IS APPLIED:
Machine learning techniques such as
random forest
TYPICAL DATASET TO WORK ON THIS:
Customer log and event data
We’re here to help
Contact sales@datascience.com
25. SUBSCRIPTION LIFETIME
FORECASTING
WHY THIS IS VALUABLE:
Forecasting the subscription length of
different segments provides a more accurate
understanding of future revenues and your most
valuable segments
WHICH KPI THIS IS GOING TO EFFECT:
Increase overall Customer LifetimeValue by
understanding what segments of your user base
are the most lucrative
EXAMPLE OF A MODEL THAT IS APPLIED:
MCMC simulated survivial analysis on
censored data
TYPICAL DATASET TO WORK ON THIS:
Customer log and event data
We’re here to help
Contact sales@datascience.com
27. We’re here to help
Contact sales@datascience.com
LONGITUDINAL IMPACT
ANALYSIS
WHY THIS IS VALUABLE:
Small changes can have long term consequences.
A few too many sales can shift customer opinion to
think of you as the “wait for a discount” provider.
With properly constructed experimental design,
smart sellers can measure and watch out for
such pitfalls
WHICH KPI THIS IS GOING TO EFFECT:
Increase customer retention by ensuring a
sustainable and quality product over time
EXAMPLE OF A MODEL THAT IS APPLIED:
Multivariate hypothesis testing and
time series analysis
TYPICAL DATASET TO WORK ON THIS:
Marketing event logs and sales transactions
29. REFERRAL / LOYALTY
PROGRAM EFFECTIVENESS
WHY THIS IS VALUABLE:
The best advertising is word of mouth advertising.
Programs to encourage these activities often have
impressive ROI. Proper tracking and optimization
can ensure the greatest yield
WHICH KPI THIS IS GOING TO EFFECT:
Increase customer acquisition by optimizing
referall rewards and channels
EXAMPLE OF A MODEL THAT IS APPLIED:
A/B testing, simluation, and general
machine learning approaches
TYPICAL DATASET TO WORK ON THIS:
Program tracking
We’re here to help
Contact sales@datascience.com
31. SURVEY RESPONSE ANALYSIS
WHY THIS IS VALUABLE:
Most online shoppers are very willing to share
their thoughts with stores. Extracting the most
value from this feedback can enable an experience
better tailored to your best customers
WHICH KPI THIS IS GOING TO EFFECT:
Increase NPS by better understanding what
satisfies your customers
EXAMPLE OF A MODEL THAT IS APPLIED:
Crosstab significance testing, sample
balancing, survey design, open ended analysis
TYPICAL DATASET TO WORK ON THIS:
Survey response data
We’re here to help
Contact sales@datascience.com
33. DELIVERY DATE ANALYSIS
WHY THIS IS VALUABLE:
Returns, delays, and lost packages all spell disaster
for stores. When delivery plays a roll in this, an
optimized shipping schedule can often reduce
these costs
WHICH KPI THIS IS GOING TO EFFECT:
Increase customer satisfaction scores by
ensuring packages are delivered quickly,
on-time, and with minimal hassle for users
EXAMPLE OF A MODEL THAT IS APPLIED:
Historical analysis and simulation
TYPICAL DATASET TO WORK ON THIS:
Shipment tracking data
We’re here to help
Contact sales@datascience.com