Author
altocloud
View
179
Download
4
Embed Size (px)
The true meaning of dataData Science meets Marketing
Maciej DabrowskiChief Data Scientist, Altocloud
Altocloud
Real-time analytics Real-time for us is under 1-5s
Q: How many customers are currently on my website?
Q: How many customers are looking at the new article?
Q: How many people from Dublin who spent over 20 minutes on a star wars product page end up spending over 100?
Analytics
aggregations - slicing a bit like
Predictive AnalyticsQ: Which customers currently on my site are likely to convert?
This talkWhat is Data Science?
Common traps in data analysis
Data Science and Marketing
Data Science
Data ScientistHuman (storytelling) vs. Machine analytics (Machine Learning)
Type A (analytical/statistician) vs. Type B (builder/engineer)
Data ScienceSelect a question and a metricWho is likely to convert? (purchase/conversion rate)
Collect relevant dataUser behaviour (page views) and demographics (device)
Analyse the data and discover patterns10% of returning customers who visit my website on their iPhone after 8pm and spend over 20 minutes end up buying.
Common problemsAm I using correct metrics to answer my question?
What is the quality/accuracy of my data?
Do I use correct visuals and draw the right conclusions?
Metrics
MetricsCommon metrics:number of sessions/visitsnumber of unique visitorstotal salestime on site
Other metricsconversion rate (percentage)
Is the metric accurate?Monthly visits
Is the metric accurate?Daily visits
MetricsMake sure that you understand how your metric worksHow are the visits counted?
Always challenge the quality of your dataWhat events can influence my metrics?
Use the right metric for the job absolute value vs. percentage
PresentationLabel your axes!
PresentationLabel your axes correctly!
Tricks to make your data look better
Less is moreOverloaded dashboards may hide important facts about data.
Focus on what you want to knowUse charts when you care about trendsUse numbers when you care about absolute valuesUse pie charts when you care about percentages
Simplicity allows you to understand data quicker and easier.
Correlation vs. causation
Correlation vs. causationConclusion: Science is depressing!
Correlation vs. causationConclusion: Cheese makes you more likely to get killed by your bedsheets
Correlation vs. CausationConclusion: Eating margarine will get you divorced!
Data Science for MarketingContent marketingWhich content has the potential to go viral
Marketing successPredict the success of marketing campaigns
Customer analysisPredict churnSegment your customers
Amazon Machine LearningEasy to start
Does not require complex knowledge of Machine Learning techniques and algorithms
Require to move your data to the cloud
Big ML
R ProjectFree desktop tool
Very powerful for advance statistics
Can work with Big Data platforms (Spark)
Requires more knowledge about stats
SummaryMake sure that you understand your data and metrics
Less is more in analytics dashboards
Correlation is not causation
Data science does not require very complex tools!