PushON hosted the first Online Marketing Meetup in Manchester on 3rd April at Manchester Digital Tech Incubator. Speakers included Ceri Francis from Music Magpie, Denise Maskew from United Utilities, Ben Wilkins from Regital and Gareth Hoyle from Marketing Signals. The meetup was sponsored by Run 2, with Richard Gregory compering,
1. OnlineMarketingMeetup
Ecommerce. Delivered.
3rd April, 2019
MSP, Manchester Technology Centre
Hosts:
Richard Gregory
https://www.richardgregory.co.uk/
Elena Williams, Sarah Wilding, Jonny
Mangas, Lee Mason
PushON
Sponsored by
#OnlineMarketingMeetupMcr
@PushOnltd
#OnlineMarketingMeetupMcr
@PushOnltd
2. #OnlineMarketingMeetupMcr
@PushOnltd
Agenda
Ecommerce. Delivered.
17:45-18:15 Attendee registration, welcome drink
18:15 Introduction from Richard Gregory
18:30 Ceri Francis, Music Magpie: Measuring Facebook
18:55 Denise Maskew, United Utilities: A Look At Marketing From Inside A
Corporate Organisation
19:15 Food/Drinks
19:30 Ben Wilkins: The Revolution In Out Of Home
19:55 Gareth Hoyle: Driving Digital Clicks To Your Physical Bricks
20:15 Closing the event and Informal gathering, drinks
Sponsored by
4. Summary
Defining the problem
Briefly, on the Facebook Pixel
Measurement options overview
• Attribution window comparison
• Promo codes in Ads
• Regression analysis
• A/B Testing
• Facebook Marketing API
• Bayesian structural time series
Questions
10. Facebook Pixel: Common Problems
No transaction IDs – and against TOS to send them?!
Can’t validate the accuracy of pixel implementation
Agencies with poor understanding of measurement add their pixel to
the site
Multiple pixels duplicate event based conversion reporting
Non-specialists try multiple methods of implementation (e.g. dual
implementation in Shopify, + a tag manager)
Duplicate pixel triggers cause duplicate conversion reporting
11. Resource Cost Robust Results
Regression Analysis
“Look at the Pixel” & compare attribution windows
Facebook Lift Analysis (Marketing API)
Machine Learning (e.g. Bayesian approach)
Audience Based A/B Testing
Measurement Approaches
Reliance on Facebook Platform
Promotional codes in Ads
12. “Look at the pixel” – Attribution Comparison
How it works
1. Customise columns in “Ad Manager” or “Ad Reporting”
2. Click window comparison and select as appropriate
13. “Look at the pixel” – Attribution Comparison
Summary
Advantages Disadvantages
Easy Wholly relies on Facebook platform data
Allows closer comparison to other Display
platforms attribution windows
If using for optimisation, more time consuming/
less efficient measurement
No cost
14. Promotion Codes in Ads
1. Setup promotion codes
2. Setup reports and calculate CoS, CPA etc.
3. If possible, automate
How it works
Campaign Spend Code Redemptions Revenue CPA CoS
Campaign 1 £250 26 £1170 £9.61 21.4%
Campaign 2 £150 21 £1203 £7.14 12.5%
15. Promotion Codes in Ads
Advantages Disadvantages
No reliance on Facebook platform to track results Dependence on customer recall of the code
Allows daily optimisation Difficult to factor into multi-touch attribution
models
Efficient and reliable method of measuring FB-only
promos
Campaigns are incentivised so results cannot be
generalised to non-promo activity
May require marketing budget for voucher codes
Summary
16. Audience Based A/B Testing
1. Allocate significant portion of marketing spend over 1+ months to Facebook
2. Define two randomised audiences
3. Run campaign: pick which group to serve ads to, and exclude the other from targeting (“Treatment”
below).
4. Evaluate results (typically frequentist t-test unless you have a statistician) once significance is reached
How it works
Test Group Facebook Reach Sales Revenue AOV
Reach Estimated
Conversion Rate
Control 163,567 20,300 £701,568 £34.56 12.4%
Treatment 158,392 18,700 £649,825 £34.75 11.8%
17. Audience Based A/B Testing
Advantages Disadvantages
Analogous to CRM marketing testing when
targeting existing customers
New customer acquisition less reliable as may
require some reliance on Facebook reach metrics
Gives a view of the incrementality of Facebook
beyond PV/PC attribution
Some problems including results in direct-response
based attribution models
Summary
18. Regression Analysis
1. Allocate significant portion of marketing spend over 1+ months to Facebook, in a low-variance period
2. Define two randomised audiences (optional, useful for high variance periods)
3. Analyse results once significant
How it works
19. Regression Analysis
Example
Significant predictor @ 99%
Explains 16% of variance
Correlates strongly with orders
Not significant predictor (<90%)
Explains 7% of variance
Correlates minimally with orders
20. Regression Analysis
Summary
Advantages Disadvantages
Good solution for brands with fairly “static” sales
trends
Difficult to isolate Facebook performance when other
variables at play, e.g. seasonality
Gives a view of the incrementality of Facebook
beyond PV/PC attribution
Some problems including results in direct-response
based attribution models
Goes further than other methods to answer:
“what should the FB budget be?”
21. Marketing API
1. Randomise customers into “cells” (test groups)
2. Include customers eligible to see ads in both groups
3. Facebook will measure outcomes in both groups
4. Significance and incremental lift are calculated
How it works
22. Marketing API
Summary
Advantages Disadvantages
Best measure of eligible reached customers Wholly activated within the Facebook platform
Combines statistical validity with accurate source
data
High resource cost if done without assistance from
FB
Easy reporting as all data held in Facebook Can be run by FB AMs if spend is high enough, but
then no view of result calculations
Gives a view of the incrementality of Facebook,
including for offline sales
24. Bayesian Approach (Causal Impact)
1. Add time-series predictors of measured KPI and determine time-series predictors of performance
2. Specify control group for use as a predictor (optional + dependent on situation)
3. Run test and pass results through the determined model to estimate the impact of the
intervention(marketing campaign) in the post-period
How it works
1. Adrian Colyer, https://blog.acolyer.org/2016/03/03/inferring-causal-impact-using-bayesian-structural-time-series-models/
1
Campaign Period
25. Bayesian Approach (Causal Impact)
Summary
Advantages Disadvantages
Strongest prediction power of all listed methods Resource cost very high
Additional insight into predictors for your business Complex to setup and run
Requires a strong signal (often high spend) to be
valid
26. Bayesian Approach (Causal Impact)
Application of the method
https://blog.acolyer.org/2016/03/03/inferring-causal-impact-using-bayesian-structural-time-series-models/
R Library by Kay Brodersen @ Google
https://google.github.io/CausalImpact/CausalImpact.html
Python Port (incomplete) by Jamal Senouci @ dunhumby
https://github.com/jamalsenouci/causalimpact
Further Reading
31. How best to support each other
⊗ Live and breathe the brand
⊚ Take the time to understand the brand and the industry
⊚ … and take the time to teach us yours
⊗ Agree on deliverables, time frames, outcomes etc up
front – and how to communicate when things change
(on both sides
31
32. ⊗ Help us brief properly and understand the results we
may see (with budget options)
32
33. ⊗ Understand how seemingly small changes can have
big operational impacts
⊗ There is an art to keeping on top of us to deliver what
we need to versus driving us up the wall with
requests
⊗ When you pitch… don’t just send the A team, we
want to meet the people we will be working with
⊗ Give us the confidence to be brave and less risk
averse
33
34. And of course we are massively jealous you can wear
jeans and use Macs…not just on Fridays
34
54. 01.
02.
03.
Contents.
Why the high street is still
important. And always will be.
Driving targeted footfall to your
physical locations.
Measuring success from your
digital marketing spend.
#clickstobricks | @mrketingsignals | @search_magician
55. Why the high street
is still important.
#clickstobricks | @mrketingsignals | @search_magician
56. In November 2018, pre-Christmas
internet sales accounted for only 20%
of retail spending in the UK.
https://ons.gov.uk/
57.
58. Figures for September 2018 show
UK in-store sales declining 2.7%
compared to September 2017.
BDO’s latest High Street Sales Tracker (HSST)
59. The way we shop
has changed.
#clickstobricks | @mrketingsignals | @search_magician
60. The high street
has changed.
#clickstobricks | @mrketingsignals | @search_magician
61. Yet many activities still
involve visiting a
physical location.
#clickstobricks | @mrketingsignals | @search_magician
80. Reach a relevant audience by
identifying interests and overlaying
with demographics.
Demographic &
interest targeting.
Interests
Gender
Age Range
21-42
Awareness Consideration Preference Buy Loyalty Advocacy
81. Reach a relevant audience by
targeting specific job titles, functions
and skills.
Job title targeting.
Awareness Consideration Preference Buy Loyalty Advocacy
Job Title
Industry
Skills
82. All customers are NOT equal.
20% 60% 20%
Non-Profitable
Customers
Profitable
Customers
Very Profitable
Customers
Numberofcustomers
Time
#clickstobricks | @mrketingsignals | @search_magician
84. Location extensions on your search ads.
Show them where the
location is whilst they are
searching online
Awareness Consideration Preference Buy Loyalty Advocacy
85. Awareness Consideration Preference Buy Loyalty Advocacy
Show adverts in the local map pack.
Increase the visibility of
your local business within
Google Maps.
The easiest way to the
top of the map pack.
86. Awareness Consideration Preference Buy Loyalty Advocacy
Drive store visits direct from PLA ads.
Label adverts when the
user can visit a store and
see the physical product.
87. Awareness Consideration Preference Buy Loyalty Advocacy
Local inventory listings from PLA ads.
Let local users know that
your shop has the items
that they're looking for, at
the moment they search.
88. Physical customers can be
encouraged to mention the store they
visited in their online review.
Adds another layer of trust to
influence others to visit.
QR code on a receipt and a voucher?
Online reviews via
physical customers
Awareness Consideration Preference Buy Loyalty Advocacy
89. Whilst they
are at the shops.
#clickstobricks | @mrketingsignals | @search_magician
92. Direct the user straight to a phone
call, skipping the website visit.
Also show the physical location to
drive footfall
Click to call ads.
Awareness Consideration Preference Buy Loyalty Advocacy
93. Encourage footfall into
your local business with a
specific message or offer
within your Google Maps
listing.
Promoted pins.
Awareness Consideration Preference Buy Loyalty Advocacy
94. Local awareness ads.
Reach Facebook and
Instagram users who are
within a certain radius to
drive footfall to your physical
locations.
Awareness Consideration Preference Buy Loyalty Advocacy
95. Reach Facebook and Instagram users
who are currently at a certain location
by dropping a pin on the precise
coordinates of the physical location.
One of the most under used tactics in
most campaigns.
Social pin drops.
Awareness Consideration Preference Buy Loyalty Advocacy
97. Minimum targeting = 1KM
Add exclusion zones to the mix...
Minimum targeting = your house!
Or
A competitor's business.
Or an event you are attending
where your potential clients may
be.
Laser target users.
Awareness Consideration Preference Buy Loyalty Advocacy
98. Here is the beautiful end result.
Run special offers - Maybe use a
rival business name as the
voucher code.
Perfect to steal competitor and
drive local footfall
Laser target users.
Awareness Consideration Preference Buy Loyalty Advocacy
99. Give users a reason to
visit your locations.
#clickstobricks | @mrketingsignals | @search_magician
105. Ad platforms provide us with tools.
Offline Conversion Tracking /
Store Visits Report (GA)
Customer data must be shared.
#clickstobricks | @mrketingsignals | @search_magician
Offline Event Tracking
107. Ad platforms provide us with tools.
1. A logged-in
Google/Facebook user
clicks on a search/social
ad before visiting store.
2. Customer
information gathered
via loyalty scheme,
product activations or
e-receipts.
3. Transaction data
with associated
customer data is
uploaded to Google
Ads/Facebook.
4. Transaction
data is matched
with search/social
ad click from
logged-in user.
Instore Purchase
Attribution
#clickstobricks | @mrketingsignals | @search_magician
108. Customer match needed
We only get a positive result if
the data we collect matches
what the ad platforms hold.
Most people only have 1 email
address and phone number.
110. Key identifiers to collect.
Email Address Telephone Number Customer Number
#clickstobricks | @mrketingsignals | @search_magician
111. Voucher generated online to be used offline.
#clickstobricks | @mrketingsignals | @search_magician
112. Some food for thought…..
Would it be worth offering a
voucher code in store in order to
track those customers who are
“showrooming” you?
#clickstobricks | @mrketingsignals | @search_magician
113. Loyalty cards work. People love free stuff.
#clickstobricks | @mrketingsignals | @search_magician
114. e-Receipts.
Use the “green” angle and
help save the planet.
Easy and non-obtrusive
method of data collection.
Can be used in a variety of
different high street settings.
No guarantee they give you
their correct email address.
115. Follow up contact.
More convenient for the
customer to be reminded.
Grants permission to contact them
with special offers and discounts.
Easy and non-obtrusive
method of data collection.
Very high chance they will give
you the correct details.
116. You are giving the customer
something in exchange for data
Grants permission to contact
them - cross sell products.
Easy and non-obtrusive
method of data collection.
No guarantee they give you
their correct email address.
Offer free Wifi.
117. Attribute footfall in your real
world locations to online
marketing activity.
Too many privacy concerns for
this to work at the moment.
Beacon project.
Awareness Consideration Preference Buy Loyalty Advocacy
118. It is not an easy
journey.
#clickstobricks | @mrketingsignals | @search_magician
119. It is not an easy
journey.
#clickstobricks | @mrketingsignals | @search_magician
120. I hope I have taught
you all at least 1
thing today.
Download this presentation:
http://seo.im/clickstobricks
@mrketingsignals | @search_magician