The deck I co-presented with Ran Cohen of Upfront Digital Media (a Legolas Company) at IRI's Summit in April, discussing the evolution of digital advertising from direct response to brand building
27. August 30-31, 2010
How It Works…
Identify & Create purchase-based
audience segments from Panel
Legolas passes exposed
panelists to IRI
IRI compares purchases between
test and control groups
$2.00
$2.50
Buying Rate
Non-Exposed Exposed
Ad tags identify when panelists are
exposed to ads.
Legolas models panel to
larger audience and
activates campaign
Define objectives and targets Deliver Panel IDs to Legolas
IRI identifies control group
Heavy Users
Competitive Users
High Category Users
28. Case Study: Baked Goods
• Campaign Summary
– Initial, directional results are positive
• Dollars per household are up 6.9% vs. prior year.
Driven by an 11.3% increase in dollars spent per
occasion.
• Targeting Performance
– Females 21 – 49 with kids and who bought competitive
products indexed higher than total campaign in dollars per
household & penetration
Target
Dollars/
HH
Penetration Occasions
$ per
Occasion
Buyer
Count
Target Source
Total Targeted 109 106 101 102 509
F 26-55 w kids 117 107 106 103 242 Target Source 1
Brand X switchers & heavy sweets buyers 111 105 99 107 68 Target Source 2
F with kids 117 108 114 95 59 Target Source 3
F 25-49 w kids who buy sweets 107 100 107 99 96 Target Source 4
F 21 - 49 with kids who buys competitive 126 123 100 102 175 Target Source 5
F 25-54 w kids 98 101 100 97 122 Target Source 6
F 25-49 w kids 97 102 100 96 109 Target Source 7
Interim Period - Index to Total Campaign
29. Case Study: Frozen Pizza
• Campaign Summary
– Initial, directional results are positive
• Dollars per household are up 13.3%, penetration was
up 17.5% offset a small decrease in occasions vs. the
control.
• Creative Performance
• ―Soccer‖ only creative clearly outperformed
Creative
Dollars/HH Penetration Occasions
$ per
Occasion
Buyer
Count
# IRI HHs
Exposed
44047 - Soccer (Only) 148 140 103 103 41 626
44049 - 14 Min (Only) 72 90 88 92 28 644
44051 - Obstacle (Only) 119 134 88 101 39 647
Interim Period - Index to Total Targeted
30. Case Study: Soup
• Campaign Summary
‒ Initial, directional results are positive
• Dollar per household sales: +$0.02 during the 6-week interim period
• Purchase occasions and dollars spent per occasion are driving the
increase.
• Targeting Performance
Target
Dollars/HH Penetration Occasions Dollars/Occasions Buyer Count
No Profile 73 88 99 84 183
Purchse-Based
Targeting
128 112 102 112 204
Target Source 1 119 129 79 118 35
Target Source 2 133 113 102 115 148
Target Source 3 107 103 89 116 82
Target Source 4 150 121 104 120 67
Interim Period - Index to Total Campaign
31. Case Study: Soup (cont’d.)
• Publisher Performance
Publisher
Dollars/HH Penetration # Occasions $/Occasions Buyer Count
Publisher A 118 131 91 90 89
Ad Network A 96 111 82 105 37
Publisher B 103 95 104 104 26
Publisher C 162 123 95 139 30
Portal A 92 91 100 102 281
Interim Period - Index to Total Campaign
37. Panel Analytic Tree – Drivers of
Dollar Sales
Dollar Sales
% of HH Buying
―Penetration‖
Dollars Per Buyer
―Buying Rate‖
$ Per Purchase Occasion
―Purchase Size‖
Purchase Occ Per Buyer
―Purchase Frequency‖
38. August 30-31, 2010
Frequently Asked Questions
• How can I best use the optimizer?
The indices show sales from households exposed to your campaign. We can segment sales lift across
purchase-based targets, creative, publishers and messaging tactics (e.g. contextual vs. behavioral)
The indices directionally demonstrate which tactics may have an effect on household purchases.
• What’s the difference between the optimizer and an ROI analysis?
The optimizer indexes raw data from our panel to quickly review which digital media tactics may have
an effect on household purchases. It’s a snapshot in time that shows product sales during the
campaign. It uses only one cell (exposed/test).
An ROI analysis uses a statistical regression model that demonstrates advertising impact from buying
households on product sales. We adjust the model to ensure the exposed (test) and non-exposed
(control) cells look alike in their buying behavior (price paid, retail outlet shopped, etc.) and
demographics. The analysis provides the most likely (average) short-term sales lift that we would
expect to see at least 80% of the time given the same tactics.
39. August 30-31, 2010
Frequently Asked Questions
• What’s the difference between raw and modeled data?
Raw data includes purchase information pulled from our panel database as uploaded by our
participating buying households.
Modeled data excludes outliers and attempts to ensure the test and control sample sets are similar
in their buying behavior, in-store promotional activity, pricing considerations and demographics.
Additionally, influencing factors outside of the digital ad campaign are adjusted for, allowing us to
provide a statistically significant result.
• Can the optimizer predict our total campaign sales lift and ROI?
No, the optimizer is a snapshot in time that illustrates sales during your campaign. Directionally, it
can demonstrate if sales are moving in a positive or negative direction. It does not demonstrate
impact to sales based on exposure to advertising.
40. Target Optimization Analysis (vs
Total Targeted)
• Directionally, audience targeting has had a positive impact on the campaign
– Most targets performed similarly, but Target 2 under-performed compared to the rest.
• Target groups are mutually exclusive
• Sample Size = 572 raw exposed targeted buyers (out of 11,377 exposed Panel households)
• Please note – raw data, for diagnostic purposes only, not predictive of final results.
Target
Dollars/HH Penetration Occasions
$ per
Occasion
Buyer
Count
# IRI HHs
Exposed
Target 2 (Only) 79 90 93 94 46 1,059
Target 3 (Only) 104 109 96 99 70 1,288
Target 4 (Only) 103 93 103 107 82 1,541
Target 9 - National L&L or BC (Only) 100 95 103 102 129 2,696
Target 10 - National Demo (Only) 104 102 95 108 56 1,091
Super Target 1 - General (Only) 89 92 96 100 221 4,542
Super Target 3 - National (Only) 99 96 99 104 189 3,943
Interim Period - Index to Total Targeted