What is attribution?
Attribution is the specific approach used to compare a test group against a baseline, proving exactly how much extra revenue was directly generated by the Contextual Commerce campaigns.
Accurate attribution is essential for demonstrating the value of Contextual Commerce campaigns. Contextual Commerce attribution doesn’t predict individual engagement, but rather tracks conversion trends and patterns over time. This approach ensures we measure value fairly by crediting the campaigns only for additional sales, rather than taking credit for purchases that may have happened anyway.
How it works
We use a rigorous, multi-step data alignment and attribution process to establish an identical baseline between our systems. This ensures that you can measure the true incremental revenue generated by our campaigns. This process involves:
- Data alignment: We match six core metrics (like sessions, orders, and average order value) between platforms to ensure discrepancies are under 5%. Refer to the data alignment section to learn more.
- A/A testing (control vs. control): We run identical, non-campaign traffic splits to prove our tracking infrastructure is flawlessly synchronized. Refer to the A/A testing section to learn more.
- A/B testing (launch): We compare a control group with a test group to measure the actual revenue lift driven by Contextual Commerce. Refer to the A/B testing section to learn more.
Common questions
How do I report on attribution?
As Contextual Commerce integrates directly with your existing analytics platform (like Google Analytics 4 or Adobe Analytics), your group assignment data flows directly into your own systems. This allows your team to independently verify every data point and easily export customized A/B test reports. Alternatively, the Bazaarvoice Portal offers a visual "Incremental Revenue Year-to-Date" chart that is useful for executive summaries.
If a campaign in the test group is underperforming, what can my team do to fix it?
Contextual Commerce is an all-inclusive, fully managed service that optimizes itself. If a specific campaign is underperforming, our AI decision engine's Automated Campaign Optimization (ACO) instantly detects the drop in engagement and automatically adjusts campaign weights in real time to favor higher-converting messages. Bazaarvoice Strategy and Data Science teams continuously monitor this campaign-level data to design, test, and deploy fresh campaigns on your behalf at no additional cost. To make adjustments to the enabled campaign set, contact your Contextual Commerce strategist to create an action plan.
Where do I view attribution?
You can monitor your control versus test group performance through the Bazaarvoice Portal. The Portal features a dedicated Contextual Commerce dashboard that is continuously updated. You can review exactly which campaigns are active, track your key performance lifts in conversion rate and revenue-per-session, and see a visual breakdown of your performance across desktop, mobile, and tablet devices.
Data alignment
Before activating a campaign, Contextual Commerce checks whether the data from our platform aligns with your analytics. This ensures that billing is based on an accurate baseline and attributable impact.
The following metrics must be aligned between Contextual Commerce data and your analytics:
- Sessions
- Orders
- Revenue
- Conversion rate (CVR)
- Revenue per session (RPS)
- Average order value (AOV)
Contextual Commerce expects the following delta standards between our data and your analytics:
-
<5%
Acceptable alignment -
5%-10%
Requires further investigation -
>10%
Triggers a critical investigation
Bias detection
Bias can impact both control and test groups, affecting the perceived value of campaigns and the trustworthiness of attribution.
- Bias in the test group: Leads to overestimating the campaign's impact, resulting in inflated billing.
- Bias in the control group: Leads to underestimating the campaign's effectiveness when the control group shows more conversions than expected.
How uneven splits help detect bias
Bias is more easily detected with uneven splits (for example, 60/40 or 90/10) because they highlight data inconsistencies. A 50/50 split accelerates confidence in results, while a larger control group ensures greater data stability by reducing variance.
A/A testing (Control vs. Control)
During A/A testing, Contextual Commerce assesses the statistical significance of the control and test groups to ensure alignment. Since no campaigns are running, we expect no significant differences between the two groups. This test displays identical content to both the control and test groups to confirm that the traffic split is unbiased and that both groups perform similarly.
Contextual Commerce performs two A/A tests:
-
50/50 split
This test uses a standard 50/50 traffic split to ensure no bias in traffic allocation or performance differences between the two groups. -
Uneven split (for example, 60/40 or 90/10)
This second test magnifies any issues by skewing the traffic split. If there are any problems with the infrastructure, they are more likely to be detected with an uneven split, as it exaggerates the differences.
Data alignment is a collaborative process. Both Contextual Commerce and the client must review and approve the results before moving forward to the Proof of Value phase. While we aim for high-confidence results, Contextual Commerce allows flexibility for business sufficient results, reducing the opportunity cost of extending the A/A test duration.
Ongoing data alignment
Data alignment is not a one-time activity. While monthly A/B tests begin testing campaign performance after the initial alignment, Contextual Commerce will continue testing and maintaining data alignment to help detect any shifts over time and ensure that billing remains based on an accurate baseline.
A/B testing (Control vs. Test)
Contextual Commerce measures campaign performance using A/B testing that compares conversion trends between a control group and the test group that saw campaigns. Contextual Commerce monitors the differences in campaign performance between these groups to validate attribution.
- Control group
Establishes a baseline by measuring shopper behavior without exposure to any campaign. - Test group
Shows measurable differences in conversions and revenue in shoppers who saw a campaign.
By maintaining alignment in how the groups are split, we ensure that differences in performance are due to the campaign alone and not to other factors.
