Professor Karen Nelson-Field PhD’s presentation at Mumbrella MSIX 2018
- 2. We build our own technology to directly improve the validity of the marketing research we do. PhD Trained Marketing Researchers Media, Content diffusion, Modelling, Consumer behaviour, Advertising effectiveness, Mathematics + Computer Science Engineers Software engineering, Mathematics, Machine Learning , Computer vision, Product design, App development, System architecture, VR
- 4. Why the attention on ATTENTION?
- 5. Simple. There’s a genuine and logical argument need for a comparable cross platform impression measure. Attention is a universally applicable, cut-through metric. Reflects vast modal differences. Scalable and passively collected.
- 6. “ “Attention is the allocation of mental resources. Before consumers can be affected by advertising messages, they need to first be paying attention. Thales Tiexeira Professor Marketing Harvard Attention lets information in, memory holds it in place. Without the former, the latter doesn't exist. Dr Jared Cooney-Horvarth Cognitive Neuroscientist Uni Melb
- 7. “ AdAge: Fox Wants 'Attention' To Be New Language For TV Ads 2018 Ad Attention Research: Effectiveness hinges on more than just reach 2017 Do not pay for impressions, pay for attention. 2002 Attention is rising to attention as the trading future. The Economist introduces multi platform attention buy for its clients 2015
- 8. But there is attention and then there is attention ▪ Eyes on screen, eyes off screen, eyes on screen AND on ad ▪ In-situ viewing (natural) ▪ Scalable for sample size ▪ Anchored to Sales/Product choice ▪ Passive collection ▪ Dwell time/time spent in view ▪ Goggles or Lab settings ▪ Restricted sample size ▪ Anchored to Recall or Likeability ▪ Self declared attention VS
- 9. We see vast differences in attention declared by HUMANS versus attention measured by MACHINE We found half over-reported attention by 42% and half under-reported by avg. 36%
- 10. Biased and Subjective measurement can mean the difference between ‘any result’ and the ‘right result’.
- 11. 1. Download Collection App. Collects facial footage and viewing metrics. Intercepts natural ad load. Tracks test ad viewability metrics. 2. View Test Platform or Programming. Viewing occurs in a natural environment. 3. Go to Virtual Store. Discrete choice used as an anchor to attention. 4. Data Pushed to Analytics Framework. Sec by sec attention and viewability metrics at view level. Our collection process is natural and in real-time
- 12. What we mean by in-situ viewing We capture attention across BOTH media platform and viewing device …...PLUS
- 13. We also capture inward facing information COVERAGE % of screen covered by ad PIXELS % of ad on screen TIME time spent in view SCREEN SIZE size of viewing screen
- 14. And Brand Quality Purpose built annotation tool for objective measurement Metric 1: Brand Frequency Metric 2: Brand Entry Timing Metric 3: Total Brand Duration Metric 4: Brand Prominence (size)
- 15. Machine based annotation allows us to rate executions on their branding quality at scale and quickly. And minimises any potential error of human coding of this information.
- 16. Strong emotion drives forwarding of emails (Berger and Milkman 2012). Longer life span of ‘memes’ (Bell and Sternberg 2001). Emotions are a driver of brand favourability (Heath 2009). Greater concentrated attention (Teixeira, et al. 2011). And emotions: long noted as a conductor for further cognitive and behavioural reactions.
- 17. Positive High Arousal Low Arousal Hilarity Amusement Inspiration Calmness Astonishment Surprise Exhilaration Happiness Negative High Arousal Low Arousal Disgust Discomfort Sadness Boredom Shock Irritation Anger Frustration But for now, defaulted to tested Arousal/Valence Grid (2013 Nelson-Field)
- 18. - Products with Ekman 6 training data available. - Paul Ekman 1971 – Emotions constants across cultures. - Research not about behavioural outcomes (sales). - Ekman 6 not balanced across arousal and valance. - We are not alone: “Silicon Valley Thinks Everyone Feels the Same Six Emotions”. Why didn’t we integrate an ’off the shelf’ emotions SDK? Because measurement issues aren’t restricted to attention.
- 19. Working towards our own ML trained emotions model. (watch this space)
- 20. Just because it’s tech, doesn’t mean it’s the best measurement or construct. Look for both, good construct and good collection method in copy-testing.
- 21. Device Media Platform Ad Coverage Ad Pixels View Time Screen Size Product Choice Brand Quality Emotions NEW: Sound on/Sound Off NEW: Search and online buying ALL Considered with Attention. Research Questions - Copy Testing
- 22. SCALE SO FAR 70,000 TEST AD VIEWS, 44 STUDIES, 3 COUNTRIES 125,310 mins of viewing 37.5 million facial data points to train our attention model
- 24. Media choice matters a lot to attention Visibility differs significantly by platform and device. Attention can vary sig cross platform, regardless of ad quality (creative held constant). Coverage Variation avg. 10-100% Pixels Variation avg. 51-100% As such huge difference to the amount of attention advertising attracts.
- 25. High Arousal content helps too LOW AROUSAL HIGH AROUSAL STAS Avg. Attention per second STAS Avg. Attention per second 128 50 167 58 Prevalence 78% Prevalence 22% - High Arousal content delivers more ATTENTION and SALES than La content. - Although High Arousal content is rare. - No real surprise, this is all consistent with previous work.
- 26. “ BUT the influence emotion has on attention is SIGNIFICANTLY less than the influence of platform choice (about half).
- 27. At 5 frames per second we can dig deeper into the exact attention grabbing moments. What does this detail tell us about the factors that drive attention and its ability to produce a sales impact?
- 28. “ We see attention spikes are typically triggered by emotion, but not all high attention translates to a sale.
- 29. -5 5 15 25 35 45 55 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Second Lipton Active Attention (%) SalesDifferential Transposed our aggregated STAS to a second by second Sales Differential against attention. % uplift propensity to buy at active/non attention. What was happening when this uplift was greatest? Attention and Sales second by second When attention spikes, sales also go up here.But why not here?
- 30. This is a high arousal, high attention moment, but what brand is this? A common potential reason:
- 31. Mere presence of branding at attention peaks increases the chance of buying. Attention without branding still increases the chance of buying. And branding without attention helps, but significantly limits opportunity.
- 32. Sales are amplified when attention peaks and branding are aligned.
- 33. But we can take it one step further. - BRAND SIZE MATTERS MOST: 100% of higher performing ads (median split by STAS) showed the brand twice the size. - BRAND EARLY ALSO IMPORTANT With sec by sec analysis you can brand at the RIGHT TIME not simply ALL THE TIME.
- 34. E! NEWSFLASH An over-reliance on emotions hits advertisers where it hurts. 1- Emotions have less of an impact on attention than media choice. 2- Emotions have less of an impact on the sale than good branding. Obsess less on whether content will ‘go viral’ and start obsessing over whether the RIGHT content is VISIBLE.
- 35. An Attention Hierarchy formed EFFECTIVE ATTENTION = ((Ad Visibility + Ha Emotion) = Attention Spike) + Quality Branding -> Sales
- 36. Others are starting to recognise that logos at high attention points are valuable.
- 37. Ha Emotions Ad Visibility Quality Branding Only Ad Visibility + Ha Emotions = misattributed to competitor. Effective Attention Model Only Quality Branding + Ha Emotions = ad hardly visible. Only Ad Visibility + Quality Branding = limited cut through. Beware the dysfunction – all 3 maximise the attention opportunity
- 38. 1. Beware, not all attention measurement is equal. 2. More attention can be gained by choosing the right media upfront than by building emotive creative, but emotion does help. 3. Branding at high attention points significantly improves the chance of a sale. Do sec by sec copy testing. 4. Brand prominence is the most important brand quality element. Make every attention opportunity count, they are rare.