Web Analytics in the Bigger Picture of Cross-Channel Analytics

Web Analytics In the Bigger Picture of Cross-Channel Analytics

Eric TobiasDirector, Analytics ServicesUnilytics [email protected](416) 441-9009 x228

Topics We’ll Cover

• What We’re Seeing• What is Cross-Channel Analytics?• Common Goals• Challenges Encountered• Key Terminology• Components• Best Practices

What We’re Seeing

500+ engagements in North America across numerous verticals:• Self-service• E-commerce• Government• Consumer packaged goods (CPG)• Professional & association organizations• Service providers• Consulting• Intranet• Legal

Web Analytics Adoption Phases

We categorize analytics adoption in five phases:1. Implementation – Software acquisition and installation2. Basic Analysis – Reporting and monitoring of page views, visits, visitors, etc.3. Optimization – Campaigns, visitor segmentation, multivariate testing4. Automation – Dashboards and alerts5. Integration – Core business systems, cross-channel analytics

We have observed a marked increase in phase five implementations in the last year.

What is Cross-Channel Analytics?

Cross-channel analytics is the collection, analysis, measurement, and reporting of customer interaction with a company, product, service, or brand.

It is based on a hierarchy:Company

↓Channel

↓Touchpoint

↓Customer

What Channels?

We generally work with four types of channels. Those channels, along with touchpoints in each, are:• Digital – Web, e-mail, chat, online advertising, web 2.0, surveys• Phone – IVR, phone support, telemarketing• Print – Forms, publications, mail, coupons• In-person – Service counter, point of sale

Benefits of Cross-Channel

• Obtain a consolidated view of customer interactions• Optimize customer interaction across channels• Achieve more holistic view of customers• Correlate data from various channels• Extract trend and growth metrics across channels• Identify drivers that cause cross-channel “churn”

Challenges Encountered

There are a few challenges to these projects:• Political issues when working with managers from each channel• Frequent lack of common identifiers (e.g., customers, activities,

topics) requires translation infrastructure• “Time” has mixed meanings between systems• Integrated data is large, tends to require BI approach• Huge volume of measures and metrics requires classification and a

degree of automated handling

Examples of Cross-Channel Projects

A few examples of our clients currently executing cross-channel projects:• An e-commerce client is reducing customer service costs by transitioning

customers from phone-based support to web-based self service.• A major government agency is reducing their annual costs for forms and

publications by providing web-based versions to the public.• A telecommunications client is ensuring customers receive the same corporate

message and experience in each channel and touchpoint.• A CPG client is implementing a system for measuring effectiveness of print-based

promotional campaigns to drive traffic to their brand sites.• An IT consulting company is proving ROI for a self-service web site by comparing

development costs to cost savings in transitioning customers to web-based self-service.

Standard Components

Most cross-channel implementations will use the following:• KPI Paradigm• Cross-channel customer segmentation• Time standardization• Metric scoring

KPI Hierarchy

Goal

KPI

Critical Success Factors

Metrics

Measures

High level company goal

Special metrics that tell you how you are doing

Items that are vital for a strategy to be successful

Relationship of measures - ratios, averages, rates, or percentages

Raw numbers and data (web analytics, off-line touch-points, customer

databases, email marketing)

KPI ParadigmGoal

KPI

Measures

KPIs are driven by company goals

BUT…

KPIs are constructed from Measures

KPI Paradigm Example #1

KPI Paradigm Example #2

Cross-Channel Customer Segmentation

An example from one of our CPG clients:Customer Segment

Coupons Redeemed

Top Recipes Printed

Engagement Demographics

Personal Demographics

•Young Mother •Brand X•Brand Y

•Recipe 254•Recipe 786•Recipe 990

•Frequent visitor•Visited 8 times•Within last 3 wks•Registered user

•Female•Married•21-25 y/o•2 children

•Male college student •Brand M •Recipe 123 •Infrequent visitor•Visited 2 times•Within last 6 months•Guest status

•Male•Single•18-23 y/o•No children

•Female Retiree •Brand G•Brand L•Brand X

•Recipe 822•Recipe 890•Recipe 992•Recipe 1022

•Frequent visitor•Visited > 10 times•Within last week•Registered user

•Female•Over 65 y/o

Time Standardization

It is not generally feasible to store real-time data from cross-channel systems, therefore it is necessary to roll measures and metrics up to a predefined level when integrating cross-channel systems.

Time standardization also handles the discrepancies that exist in different channels for standard “time” definitions.

Standardizing time requires a survey of the various systems being integrated and assembling a master list of “time”.

Metric Scoring

Key Concepts & Terms :• Goals – Target values the metric should achieve with a timeframe in

which it should be achieved.• Valuation – An assessment of the value of the metric at any given

time.• Change classification – A means for classifying the degree of

change in the metric.• Impacting factors – A historical perspective on changes that have

had an affect on channels and touchpoints.

Metric Scoring Example #1

Example metric: Average knowledgebase searches per visit Current value: 2 knowledgebase searches per visitGoals: Short-Term = Reduce by 1 within six months

Long-term = Reduce by 2 within one yearValuation: 0 – 2 = “Excellent”

2 – 4 = “Acceptable”> 4 = “Critical”

Change classification : 0 – 100% = Negligible, do nothing100 – 150% = Minor, notify assigned analyst150 – 200% = Noteworthy, notify analysis team> 200% = Excessive, notify analysis team and channel manager

Impacting factors: Change = New FAQ added to search pageEst. Impact = Reduce metric by 0.5, starting four weeks after releaseAct. Impact = Metric reduced by 0.25 within four weeks and then stabilized

Metric Scoring Example #2

Example metric: Average daily transfers from web to phone for “Change of Address” transactionCurrent value: 450 transfers per dayGoals: Short-Term = Reduce by 100 within six months

Long-term = Reduce by 400 within one yearValuation: 0 – 200 = “Excellent”

201 – 500 = “Acceptable”501 – 600 = “Warning”>600 = “Critical”

Change classification : 0 – 5% = Negligible, do nothing5 – 20% = Minor, notify assigned analyst20 – 30% = Noteworthy, notify analysis team> 30% = Excessive, notify analysis team and channel manager

Impacting factors: Change = Fix intermittent bug that interferes with submit actionEst. Impact = Reduce transfers by 50 per day, starting one week after releaseAct. Impact = Transfers reduced by 100 within two weeks

Metric Scoring Example:Dashboards

                               LT Goal ST Goal                                                                                         0 100 200 300 400 500 600 700                             

Example #2

Example #1

0 – 2 = “Excellent”2 – 4 = “Acceptable”> 4 = “Critical”

0 6

2 4

3

Goals:Short-Term = Reduce by 100Long-Term = Reduce by 400

Valuation:0 – 200 = “Excellent”201 – 500 = “Acceptable”501 – 600 = “Warning”> 600 = “Critical”

Advanced Cross-Channel Components

Some cross-channel implementations will use the following:• Model scoring• Text mining• Topic cross-referencing• Automated metric handling• Forecasting

A component that compares known characteristics of customers with predefined archetypes. The purpose is to identify the “type” of the customer.

In this example, the customer is recorded as belonging to the “Young Professional Female” archetype.

Model / Archetype CRM Web Analytics Phone Support Point of SaleYoung Professional Female 75% 100% 80%Male College Student 10% 28% 9%Female Retiree 63% 67%Middle Age Father 21% 33% 42% 44%

Model Scoring

Text Mining

A component designed to extract “meaning” from large amounts of free-form text found in many Web 2.0 technologies.

It is designed to find thegeneral “buzz” about acompany.

For example, is this a good endorsement? Is it an isolated opinion, or is it representative of others’ views?

Topic Cross-Referencing

A component that allows for comparison and correlation of customer motivation (a.k.a., “driver”) across channels.

Topic Channel Attribute Value

Installation Troubleshooting

Phone

IVR Prompt 1 – 3 – 2

Call Topic InstIss004

Digital

KB Article CKB223108, CKB233211

Print

ISBN 978-3-16-148410-0

Automated Metric Handling

A component for sifting through the hundreds, and thousands, of possible metrics and focusing analytic teams on the most important metrics.

Metric Change Classification Action

Ratio of enquiries to claims

Between 0.60 and 0.75 Notable E-mail analyst

Greater than 0.75 Excessive E-mail team

% of first call resolutions

Between 40% and 50% Problematic E-mail Channel

Between 51% and 75% Notable Daily Report

Greater than 75% Verify E-mail analyst

Forecasting

A late-stage component that uses a history of experience with metrics to forecast values at future dates.

“Based on our prior experience with adding FAQs on top phone call drivers to the web site, how do we expect the web site traffic will be affected?”

“What is the current trend for our knowledgebase searches per support visit, and based on that where will our search volume be in six months?”

Wrap-Up

• Cross-Channel has been around for years, but was mainly used by large companies with physical stores and e-commerce. It is being implemented in a variety of verticals where companies are entering phases four and five of web analytics adoption.

• Many benefits to be gained: holistic view of the customer, cost optimization by channel, company-level view of behavior instead of in isolated silos, and trend and growth data.

• Many of the challenges encountered by early adopters have been identified and solutions derived.

• We are consistently receiving calls on how to determine KPIs. The KPI Paradigm is a best practice to determine the critical metrics.

• Standard components consist of KPI Paradigm, cross-channel customer segmentation, time standardization, and metrics scoring.

• Advanced components consist of model scoring, text mining, topic cross-referencing, automated metric handling, and forecasting.