Advanced web & store attribution in Insight
The most advanced online & offline Attribution for Retail, in Adobe Insight
My friend & colleague Derek Tangren recently wrote about whether you need an advanced attribution model. As he proposes, some of the important considerations are What are you doing with your marketing attribution data and What will you do with an advanced model?
In addition, the recent Adobe Digital Index Report “How social stacks up: Why marketers aren’t giving social the credit it deserves” made a very strong case for moving beyond the long-used, long-maligned “last click” attribution, illustrating the gap in value attributed to social simply between first and last click models.
I have spent much of the last few years of my career working with advanced attribution modeling with our retail & travel clients, as attribution is one of the main reasons that most new clients are adding Adobe Insight to their analysis toolbox. As a result, I have some fairly strong opinions based in solid experience – and we’ve developed some solid services and methodologies around helping you understand and use your attribution data. Many of those ideas were further validated by feedback and direct conversations with many digital marketers as a result of several attribution-related presentations I was able to give at the recent Adobe Digital Marketing Summit in Salt Lake City.
To understand the concept of attribution, consider a typical path to purchase. Let’s imagine a customer who sees a display ad for Adobe Photoshop, performs a search for “Photoshop” a couple of days later and clicks on anatural search result, performs a search for “Buy Photoshop” the next day and clicks on a paid search result, and then the next day experiences the Adobe Photoshop page on Facebook, clicks through, and purchases Photoshop:
Attribution seeks to provide answers to which marketing efforts had what share of influence on the conversion. Many digital marketers in retail are very familiar with attribution of online clicks to online conversions, and most exclusively use “last click” attribution, giving full conversion credit to the very last click.
How about going beyond the click? And how about going beyond just the online conversion?
A key difference about Adobe Insight, as it stands among the rest of the Adobe Digital Marketing Suite, is its ability to integrate event-level log data from practically any data source. Because of this, Insight is frequently used to combine not only Web traffic data (from Adobe SiteCatalyst, Adobe Insight Sensor, or other commercial or proprietary logging systems), but also event-level data from store sales (POS) systems and event-level data from marketing vendors like ad servers (DART, Atlas, etc.) or email service providers (Responsys, ExactTarget, Cheetahmail, etc.)
All Events in a Customer View in Insight
Via Adobe Genesis, many retail and travel marketers already integrate aggregate-level marketing data with their digital data in SiteCatalyst and the rest of the suite. But Insight takes this a step further, allowing you to correlate precise events within a single view of the customer and understand how each display ad view, email send, direct mail piece – in short, every addressable marketing touch point – influences conversions.
In Insight, all of these individual events can be associated and grouped by Customer, then organized by experience (some experiences are Web visits; others are store purchases; others are call center sessions; others are marketing sessions). All events are ordered by time, presenting fast, multi-dimensional analysis on any level of the data.
These conversions don’t have to just be cart checkouts. Attribution can be made to multiple conversion points, both online and offline: retail store counter purchases, hotel check-ins, airport kiosk events, rental car check-outs, B2B lead form submissions & Webinar attendance. You name it. If it’s an addressable event, your attribution can be configured to it in Insight.
This is important when you consider that “conversion” for most types of businesses — retail and otherwise — happens in both online and offline channels.
Baseline Insight Attribution Solution
Because we’ve found that many retail and travel marketers seek first to know what’s possible, then to understand how it applies to their business, our philosophy around attribution modeling has “grown” to a point where we initiate an Insight attribution engagement by turning on a set of pre-configured models. We’ve developed our best practices around configuring these few key models, then spending time with your data in them:
First Touch — Pretty simple: The first marketing touch gets the full credit for the conversion.
Last Touch — Also pretty straightforward: The last marketing touch gets the full credit for the conversion.
Even (Linear) — Every marketing touch within a defined period of time receives an equal share of credit for the conversion.
Starter/Player/Closer — The “Starter” (initiating) touch receives a set percent of credit for the conversion, the “Closer” (last) touch receives a set percent of credit for the conversion, and the remaining “Player” touches each receive an equal share of the remaining credit for the conversion.
Latency Score — Every marketing touch is scored with a numeric value reflecting how many days it fell prior to the next conversion.
Pathing — Leverage several advanced visualizations in Adobe Insight, once all marketing touches and conversions are organized for analysis, to understand both the direct and indirect paths between customers’ experiences with various touch points.
Engaging: “Touches”, time frames, conversions & more
Note that I used the term “touches” a lot above, not necessarily clicks. That’s because, with impression-level, or other touch (email sends, email opens, etc.) data integrated in your Insight dataset, you can go beyond standard Web “click only” attribution and also give credit to all views/impressions.
When we kick off an attribution engagement in Insight, we first ensure you’ve organized all of the event data you want in a customer-centric dataset.
We then discuss a few key questions:
- What to consider a marketing touch (what logic to use to flag each of your event rows that represent marketing touches in the data)
- What to consider a conversion (is it an online order? a store purchase? a lead submission form? other major or micro conversions?)
- How far back to look for attributing value to the marketing touches (5 days? 15 days? 30 days?)
- Any exceptions to consider (ignore branded search? ignore banner impressions if they’re immediately followed by a click?)
Next, we turn on the baseline models above.
Finally, we go initiate a series of workshops with you and your analysts. Insight-focused business consultants and predictive/statistical consultants from Adobe Consulting work with you to understand the intricacies of your data, provide initial results reporting and recommendations based on the initial baseline models, and provide recommendations for further customization of the models, before turning the solution over to you and your team to use.
Once these models are in place, you’re empowered as a marketer to have much deeper analysis on the impact of your digital marketing on your online & offline conversions. You can:
- Gain a stronger understanding of how certain marketing channels are “closers”, contributing directly to conversion and other marketing channels are further up the marketing influence funnel
- Test and understand the results of shifts in your digital marketing efforts
- Perform more informed testing of various creatives in specific placements, seeking to boost their effectiveness
- Build upon these models with further custom models in Insight, including models informed by statistical and user engagement inputs
The real results are evident in the potential for more efficient marketing efforts, avoidance of marketing cannibalization, and direct impact on online & store conversions. And what modern retail marketer wouldn’t want the potential for those results?
Next week, my Insight peer Jeremy King will be sharing some more detail around how these models are built and configured in Insight for the existing retail Insight architecture geeks who might be reading. Stay tuned!