Step #1. Master data collection
Regardless of what attribution tool you use, it’s essential to fuel it with enough quality data. Today’s consumer journeys are long, complex and nonlinear, so it’s crucial to connect all of your user touchpoints, events and interactions within your attribution tool. Measuring channels in silos is inefficient and will prevent you from gaining a holistic view of your consumer’s journey through all its interactions.
Data suitable for attribution tools include:
- Media and onsite data, such as tracking tags implemented into ads, Urchin Tracking Module (UTM) information (from an advertising landing page URL), website pixels, (which track on-site user behaviour), and mobile software development kits (SDKs – which measure in-app events).
- CRM and offline data, integrating in-store conversions and offline events like phone calls and shop visits.
- Cost data, including everything from online media events (like impressions, clicks, and views) to offline ones (like visits and purchases).
Step #2. Ensure data quality and consistency
After collecting your data, it’s essential to keep it clean. As data scientists often remind us: Garbage in, garbage out. Best practice tactics include:
- Creating consistent naming conventions
All the information displayed in your attribution reports will come from the different data sources discussed above so it is crucial to ensure naming conventions are consistent from one tool to another, to reconcile segments, events, placements and formats as well as to obtain a deduplicated and holistic view of all interactions.
- Assigning relevant categories
Channel grouping is a rule-based process that classifies traffic sources into different categories such as ‘paid’, ‘organic’, and ‘direct’. Attribution tools let you access by-default channel grouping reports, and create your own views and/or access detailed reports to reorganise your groups as needed. Non-relevant sources can therefore be deleted to avoid bias in your reports.
- Maintaining quality assurance
As for any industrialisation of data flows (in our case cost, CRM and/or offline data), you should ensure that quality insurance checks are implemented to raise alerts in case of degraded data transfers. Attribution tools often propose integrated diagnostics that continuously check the correct integration.
Step #3. Model successfully
Once all your data is collected and matched, it’s time to attribute weight in the conversion to each of the touchpoints your users interacted with. The allocation of weight will follow a specific model to explain which key factors drove the user towards conversion.
Two models exist:
- Rule-based or heuristic models follow a top-down approach where you decide which position of the journey has the greatest impact on final conversion. They are often the first step to attribution based on business hypotheses since they require little setup and are transparent. They are often used to start building a historical performance baseline and improve single-channel campaign optimisation.
- Data-driven models follow data science algorithms that assign credit based on estimated incremental value. This bottom-up approach tends to be closer to reality as the data identifies the weight of channels in the final conversion, independently from their position in the consumer journey.
Finding the best and most accurate attribution model can have a massive impact on budget efficiency. To find the right model for you, try running a conversion lift to identify incremental conversions linked to a specific channel. Additionally, test several attribution models and lookback windows, and compare which combination is closer to conversion lift results.
Step #4. Improve Analysis
Once your activation(s) has been running and you have been able to collect enough data, you can activate first insights and learnings. Before designing your campaign strategy, think about what you want to learn and identify how you will measure it (which methodology will you use? Which KPIs? Do you need control groups? etc).
If you are using attribution tools from platforms, native reporting capabilities will allow you to get advanced knowledge and answer questions like: What are the most frequent and effective consumer paths to conversion? Which channels are most often leveraged and in which order? And which channels have the most impact on final conversion?
You can select and play with many variables such as the lookback window, the attribution model, and the conversion KPI. Custom views will allow you to break down your reports by campaign, audience, or ad.
Step #5. Go further with raw data exports and dashboarding
If native reporting from attribution tools is not quite meeting your expectations, tools like Data Studio, Tableau and Klipfolio will allow you to access your attribution data by file extraction or through dedicated APIs.
This will let you reprocess and transform the data to build your very own KPIs (if not available in native solutions), and link data to specific visualisation tools to build your own breakdown and design your own views and graphs for analysis.
Most often, accessible data will be aggregated data matching the predefined taxonomy. To access the most detailed raw data you’ll need to enable linkage and use specific tools such as Ads Data Hub for Google data and Advanced Analytics for Facebook.
Bias and pitfalls
As a closing point, it’s important to realise that no matter what data you feed into it, no single attribution tool is perfect; you may find discrepancies when comparing and contrasting tools. Landing page visits and conversion rates might track differently in different tools, for example, due to conflicting conversion tracking in Facebook versus GA, for example, or cookie tracking problems.
However, attribution remains the holy grail for marketers, and following these five steps will ensure that you set up your attribution tools to give you the best chances of success.