The Analyst Process

In his book on data analytics, Digital Marketing Analytics, Kevin Hartman explains the analyst process. The analyst process is entirely applicable to the preparation step in the Enterprise sales process.

ART & SCIENCE:

Platform to drive business results, analytics to measure those results and prove out ROI. Analysts analyze analytics to influence business decisions (ie. invest to drive ROI).

NATURE OF THE ANALYST:

Strategist - How data drives business goals (ROI)

Techie - Data platforms & tools to analyze data

Storyteller - Data stories (the ROI story)

Data to express ideas or concepts

ANALYSIS PROCESS:

Descriptive Analytics

Predictive Analytics

Prescriptive Analytics

Dataviz

Visualizing data into a concept - data visualization to aid in decision making

DATA ANALYTICS:

● Analytics to get insight into consumer behavior and to measure (attribution) of marketing efforts using analytic frameworks

● Machine learning to serve up personalized ads (relevance)

● Strategist - linking business process (ie new customer acquisition) to business goals and the KPIs that are the value drivers

● Techie - using the platforms

● Buying funnel and the consumer journey

● How to isolate undecided buying intent? By bottom of the funnel connection

● Data -> collect data -> framework to organize data

● Test and learn mentality in marketing

DIGITAL MEASUREMENTS:

● Analyst’s challenge is to measure total marketing effectiveness by determining which media investments drove sales, calculating returns with attribution and optimizing future investments

● Use the data analytics tool to deliver actionable insights (dataviz) by showing patterns in the data

● Machine learning to unlock and lock in a lifetime value of a client

● Data -> analyzing for patterns in the data (knowledge) -> develop frameworks or methodologies

● Marketing Analytics Process:

● Plan - Objective, key questions related to objective, data needed for KQs, and tools/sources of data

● Collect - Mine data from platforms, data management system

● Analyze - Clean up data into tables, apply analytic methodologies to determine patterns in data, compress learnings into sound bites

● Report - Dataviz, recommendations based upon takeaways from the analytic methodologies

THE ASK

● The analyst is a strategist who applies a set of analytic methodologies and their tools/skills on data to solve problems and find new opportunities.

● Campaign objectives (CPC) drive media objectives (conversions) which drive marketing objectives (demand generation) which drive business objectives (units sold)

● Biz objectives results from marketing activities

● Analyze for one marketing objective at a time

● The development of insights from analysis

● Gather data and then structure data (into tables)

● Chart the data from a tabular form to see patterns

● Benchmarks make numbers have a context

● Understanding change through time in data

● Reducing large data into valuable insights and dataviz to make them actionable in order to drive decisions

● Data designer - visual information displays insights (dataviz) - good dataviz uses contrast

CONNECTING DATAVIZ TO THE ANALYST JOURNEY

● Strategist - Goal - plan - scope the goal, KQs, data needed and platforms to obtain data

● Techie - Information - collect - pull and store the data

● Quant - Story - analyze - format the data sets, apply analytics methodology to get analytics, characterize the data as sound bites

● Designer - Visual form - report - Dataviz charts, graphs and infographics to create actionable insights

THE MCCANDLESS METHOD OF PRESENTING DATA VISUALIZATION

● Patterns in the data to produce an effective story

● Contrast to display the insight

● Subtitle is the takeaway of the graphic

● Present the data story

● Order of the story:

Insight first

Evidence second

MCCANDLESS METHOD:

Introduce the graphic by its name (Title)

Answer the obvious before being asked (ex. What we are looking at?)

State the insight of your graphic (Subtitle)

Call out data to support the insight (Interesting examples from the data to substantiate your insight)

Close and transition to the next topic

McCandless Method for every graphic

Analyst’s graphics

TELLING A STORY

What’s our performance? (descriptive data)

What does that mean? (insights)

What should we do? (prescriptive)

DATA SCIENCE & ANALYSIS

Data analysts – examine information in order to find out actionable insights (patterns that lead to business outcomes)

DATA SCIENCE & ANALYTICS DONE IN EXCEL & ACCESS – create actionable insights to drive C level decision making to improve business outcomes

Need to clean, organize, model, analyze data to turn data records into meaningful Dataviz

Clean dataset and model data, identifying patterns in data using analytic methodologies and Dataviz to display this pattern that affects business outcomes

Data Sceince is on a project by project basis

PIVOTING DATA – CROSSTAB, BUILDING A PIVOT

● Field headings tables

● Insert pivot table

● New worksheet

STARTING WITH DATA ANALYSIS

Discover need

Determine the business rules

Find the data that fits the need and the rule

Build the dataset

Analytic methodologies to use

Dataviz to show pattern or actionable insight that affects business outcome

WHERE DID THE DATA COME FROM?

Access tables file in MS Office

WORKFLOW & PROCESS DIAGRAM

Finding data within a workflow (ie process)

Diagram the workflow

SYNTAX

Programmers must know the language of the programming platform they are using

How to write formula and functions (ie the language of Excel) ex. =EDate ( , )

Knowing how to calculate data you don’t have with formulas and functions is half the battle.

DATA SCIENCE:

1. Find data

2. Prepare data

3. Select and apply analytic methodologies

4. Present results in compelling dataviz

EXCEL FOR DATA ANALYSIS

Analytic methodologies to answer certain types of questions (typically related to a business issue)

DESCRIPTIVE STATISTICS– facts about the data to make estimates at a known confidence level. (ex. Mean median and mode are facts about the data)

Sample more conservative than a population (everyone), subtract 1 to account for not complete data in a sample

Patterns emerge as you gather more data

Using data distributions to determine the probability of a certain value occurring:

WHAT DISTRIBUTION DOES THE DATA REPRESENT?

Normal

Exponential

Uniform

Binomial

Poisson

BUSINESSES RUN ON DATA

z-score – how likely the results were due to chance (below 5% then significant results) – test for significance

ONE WAY TO ANALYZE DATA IS CALLED HYPOTHESIS TESTING.

When you form a hypothesis, you are making an educated guess about 2 sets of data

Null: A has no affect on B

Alternative: A has an affect on B

Test your alternative hypothesis

Does data support your hypothesis or reject it?

Guess about the relationship

HOW TWO SETS OF BUSINESS VALUES RELATE TO EACHOTHER:

Correlation is a more complicated version of covariance

0 to 1 with 1 being a perfect correlation moving together

-1 to 0 with -1 being a perfect negative correlation moving opposite

Known y is dependent variable as a result of the independent x variable (x causes y)

When you own or operate a business, you are always wondering what happens next. Use current trends to forecast (can fill handle down, forecast function linear if you aren’t told otherwise)

TRENDLINE TO SHOW RELATIONSHIP BETWEEN DATA

Click chart

Chart element

Trendline – more options, format trendline, always assume linear unless told differently, can forecast with trendline

DATAVIZ

Ribbon – insert – statistical chart (histogram) to see central tendency

Common goal of data analysis in business is to determine relationship between values:

Scatter charts (insert scatter x y)

Trendlines to charts (ex. Correlation?)

Typically 90 to 95% confidence interval used

Margin of error: calculate standard error = standard error x z score

Sources of error:not using random samples, basing decision on a small or early sample

As you have more values, the central limit theorem starts to take hold.

ANALYZING DATA USING SAMPLES:

Gather large samples as possible

Estimate population standard deviation

Determine confidence level

NORMAL DISTRIBUTION:

Within 1 standard deviation of mean, expect to see 68% of values

Within 2 standard deviation, 95% of values

Within 3 standard deviations, expect to see 99.7%

The more data points you have, the more measure of central tendency

INTERPRETING DATA ANALYTICS (RESULTS)

Ex. Correlation value is the correlation significant or not?

Correlation Lookup table for the two tailed test or one tailed test

Confidence level table # of sample and correlation value -> if significant, how significant

INTRO TO EXCEL

Excel is an electronic spreadsheet program (application)

Store, organize, manipulate data

Data goes into cells organized into rows and columns to then apply analytics methodologies and dataviz after that

Each cell has its own address ex. A1

Formulas for custom calculation and functions for prebuilt analytics

Highlight data and headers applicable then recommended charts

Pivot tables to view data from different perspectives, each sheet has its own tab in the workbook, Designing applied analytics to data sets and formatting for data viz to understand trends and relationships in the data

DATA ANALYTICS IN EXCEL:with new data set, discover the facts about it by determining:

Mean

Median

Mode

Variance – one measure of error (distance of individual values from mean or average)

Standard deviation – sq root of the variance as a measure of polarity around mean

How two sets of data vary in relationship to another, covariance of two data sets

0 means two data sets don’t vary together

Positive means move in same direction, more is stronger

Negative means moves in opposite directions

What is the covariance of that data pair (=COVAR), is the Covariance significant?