By: Pavan Korada, Group Vice President of Analytics & Data Science
Marketers and advertisers are plagued by attribution challenges. They lack the tools needed to provide adequate transparency about media mix and channel strategies.
A recent eMarketer report on U.S. digital marketers found:
- 57% want to improve their cross-channel measurement and attribution
- 52% want to improve cross-channel audience identification/matching
Marketers need an accurate and transparent approach to measurement; something simple, scientific and unbiased. The best way to achieve this is with an experimentation framework that removes subjectivity and provides true reads.
We call this approach Deterministic Lift Analysis.
Deterministic Lift Analysis Builds Clear Path to Marketing ROI
Deterministic Lift Analysis is an accurate, simple measurement tool for marketers. Key advantages over more traditional measurement tactics include:
- Experimentation framework captures the true impact of a campaign
- Bias-free measurement removes subjectivity, assumptions and lack of transparency
- Personally identifiable information (PII) accurately quantifies across display and social
Marketers benefit from experimentation approaches like determinist lift analysis because it removes bias, improves understanding and translates to the real world. Our unique approach allows advertisers to bring unbiased measurability to all digital channels and cross-channel campaigns.
Practiced for a long time on the direct mail slide, deterministic measurement is fast becoming favored in the digital world. It is essentially an experimentation scheme that quantifies the true impact of marketing spend.
Most digital marketers are using one of the four most common measurement and attribution approaches. None of these approaches are necessarily bad. Each has its own pros and cons.
Here’s a quick rundown of these four common attribution approaches:
- Click-Through Rate: The most commonly used metric, CTR gives a good idea of engagement; but engagement is not the same as purchase.
- Last Click: Near-purchase clicks get credit, but brand awareness activities and other campaigns get none. Channels undervalued.
- View Through: Assigns a portion of credit to the last advertised impression before the conversion event. It was designed to overcome shortcomings of last click, but the problem is it is easily gamed. It has little bearing on media placements and overvalues retargeting.
- Algorithmic: Model-driven methodology calculating channel-specific credit. Examines historical relationships between channel spend and conversion to allocate resources between channels. However, it lacks industry standard and confuses correlation with causation.
These four approaches may dominate the marketplace, but none of them truly serve marketers effectively. Deterministic attribution overcomes two primary weaknesses of these solutions:
- Channel variances in targeting. Targeting can be quite varied across channels. And channels may end up artificially looking better purely because of differences in targeting and not because of channel efficiencies. This difference in audiences is usually not accounted for by current solutions.
- Results don’t match reality. Model-based approaches are very good at describing a theoretical view of the ecosystem, but the interpretability is subjective, and findings aren’t directly actionable.
Determinist Lift Measurement in Action
Here’s how we use deterministic lift measurement to calculate campaign impact:
- Build the target universe for the campaign from an audience of only email addresses.
- Carve out 20% from the initial universe for a control/holdout group, leaving 80% for the treatment group.
- Control and treatment groups must be completely homogeneous.
- The treatment group is exposed to the campaign, while the control group is not.
- For an email-driven campaign, the treatment group email addresses are served up campaign-specific offers.
- For a display campaign, the treatment group email addresses are mapped to cookies for targeting via AppNexus, GoogleDBM, etc. via Zeta’s identity resolution capabilities.
- For a social campaign via Facebook, the treatment group is onboarded via Custom Audiences.
- A deterministic match is done upon campaign completion comparing email addresses between the advertiser’s customer records and the treatment and control groups.
- The difference between the treatment and control match rates is the true impact of the campaign.
Targeting email addresses to start the campaign and then ending the campaign using email addresses for impact measurement enables us to measure through cookie-based and walled-garden ecosystems. The ability to use PII to kick start the campaign and measure impact is what makes this possible.
This measurement can be extended to optimize the media mix for a multichannel setting:
- Identify KPIs of interest for the campaign (can be brand or performance KPIs).
- Identify a target population of prospects with known email identifiers.
- Split the target population into treatment (80%) and control (20%) groups.
- Carve out the treatment model into equal–budget, homogeneous groups. Each group corresponds to one channel.
- Run channel-specific campaigns with similar creatives and messaging (calculated for each advertiser).
- Calculate lift in conversion and brand KPIs for each channel by using an incrementality framework.
- Normalize incrementality measurements for KPIs of interest to calculate optimal media-mix.
Your Next Step Toward Deterministic Attribution
If you’d like to learn more about deterministic attribution, check out our webinar on the topic, request a demo or contact us for more information at 1-212-660-2500. It’s time for you to keep it simple and scientific. Start measuring attribution the right way, today.