Accurate multi-touch attribution reporting with Looker, Segment, and McGaw.io: a case study
Founder and CMO, McGaw.io
Nov 20, 2020
Multi-touch attribution (MTA) holds the ability to reveal the combined converting power of your marketing efforts. Yet, most organizations don’t do it well.
At McGaw.io, we implement solutions for MTA often. For some of our projects, we’ve collaborated with the teams at Looker and Segment to leverage their integration. Today I’d like to offer our time-tested strategy for breaking through two of the most significant attribution challenges: complex tech stacks and inaccurate attribution across marketing touchpoints. I’ll also share a step-by-step walkthrough of how to interpret a multi-touch attribution report in Looker.
Two common marketing attribution challenges
1: The complexity of tech stack silos
The complexity of the modern marketing technology stack is a significant hurdle in analyzing the impact of customer touchpoints. We see most marketers use a stack of a couple dozen tools. Issues then arise from the scattered, siloed nature of customer data, which makes collecting, processing, and interpreting such datasets much more difficult. This leaves marketers with data that is less accurate, less trusted, or both. Opportunities for revenue growth are underutilized.
2: Over-crediting channels closer to conversion; under-crediting assist conversion channels
Circuitous conversion paths are the norm, not the exception. An AdRoll analysis found that the average lead passes through more than 50 touchpoints1 before converting. Over-attributing to touchpoints close to conversion confuses ROI analyses, and undermines budget allocation decisions.
The three-part solution
1. Multi-touch attribution models
How exactly the conversion credit is assigned is based on the MTA model used. The most common models include:
Data-driven, Shapley, Bayesian
Determining which MTA models are right for you is an important first step. I recommend reading up on the most common ones and using a handful of them at a time. Set up four to six models for your touchpoints, then create a report that shows attributed revenue for each, and compare. If you see a trend across several different MTA models, you’re on to something.
2. Segment CDP for unified data collection
To get the most out of your MTA models, it’s crucial that the underlying data is of high quality. Weak or inaccurate data will harm the reliability of your multi-touch attribution analysis. Ill-informed analysis leads to bad decisions and lost revenue.
In other words, it pays to have clean and orderly data. This is where Segment shines. The customer data platform (CDP) can collect, clean, and organize data for later analysis, making it a perfect solution to the risks involved in working on low-quality data — and an amazing tool for implementing your multi-touch attribution models.
But unlike Looker, Segment is not a reporting solution. The way to take your quality Segment data to the next step is to use one of the Looker Blocks that have been built to connect Segment with Looker.
3. Looker Blocks for attribution reporting based on Segment data flows
Looker Blocks are shortcuts that help Looker users analyze specific data sources quickly by defining a starting point for their LookML model, with key KPIs and dashboards. Then custom modifications can be made to create the metrics and dashboards you need. Additional data sources can be joined in as well, such as ads sources, CRMs, and transactional databases.
Looker integrates with Segment by connecting to the database that Segment pushes data to. So once you’ve implemented Segment as your customer data platform, you can easily connect Looker to start getting value from your data. Analysts can leverage Looker’s modeling layer to define data definitions that can be leveraged across reporting.
Let’s dive into the type of analysis you can do with the Multi-Touch Attribution Block.
Reporting on and analyzing your multi-touch attribution models
How to analyze a multi-touch attribution report
We pulled this sample report from a prior marketing campaign to walk you through the process. If you’d like to see or use the multi-touch data we used for the report, here’s the public data sheet.
The Attribution Analysis
For context, it is important to notice in the filters that the numbers we’re looking at are for the time decay model. We decided based on previous experience with this company’s campaigns that this attribution model was the right fit for how the different touchpoints interplay.
The top row
Overall, the top row of the report shows a healthy and efficient marketing engine built primarily on organic channels. The 4.8% conversion rate yielded more than 22,200 converting customers, amounting to just over $3.5 million in value.
Zooming into search performance, we see that Organic Search converts significantly higher than all other channels, even though it makes up a relatively small portion of overall sessions. At 5.9%, it’s the highest conversion rate reported. It also generates the highest revenue per session. We can conclude that content and SEO are worthy investments. Further, this encourages us to analyze and optimize our other inbound efforts.
An ideal mix: organic search, email, and internal traffic
We also see that along with Organic Search, email and internal traffic are major sources of revenue. This tells us that regarding our current marketing efforts, the synergy between Organic Search, email, and internal traffic is the channel mix that generates most of our revenue. When we make budget allocation decisions, we’ll want to keep investing into those three. Taking one out would likely disrupt the flow of touchpoints and conversions.
Internal traffic represents the traffic moving between our various web properties. The company routinely optimizes this user flow, and based on this report we can say that the results validate the importance of that effort. It is the second biggest traffic source with a high conversion rate, and second-highest revenue total.
Direct traffic is a high performer as well. We understand Direct as the channel that closes deals for us once the users become familiar with our offerings elsewhere. We can’t directly control Direct, but we can use this data point as an indication that our other major sources of revenue — Organic Search, email, internal — contribute with even more value than what’s reported here.
Organic Social performs the worst. That holds for absolute numbers such as volume of traffic and conversions, as well as relative numbers such as conversion rates and value per session. If we invest into this channel, we’ll want to rethink our audience. We can also consider how important new traffic is for us, as that is the one metric in which Organic Social performs well.
Paid shows a high value per session, which shows an opportunity. It’s likely that we can turn the high value per session into a high revenue total if we scale the campaigns carefully.
We’ve drawn useful conclusions from this analysis:
Investing more in our Paid channel will likely pay dividends
Although Organic Social brings in the most new traffic, it performs poorly in every other metric — so we should only invest more here if new traffic is important to us
Direct traffic performs well, but since we can’t control it, it’s prudent to view it as an indicator of even more value coming from channels we do control (such as Organic Search, email, and internal traffic)
That’s a lot of good information — and we can verify (or discard) these conclusions by checking them in different models. The next step is to toggle to a few different multi-touch attribution models in our Looker dashboard, look through the report again, and analyze how assigning different weights to our touchpoints validates or negates the conclusions we made above. That’ll help us understand the full picture even better.
And with a full understanding of how our marketing efforts are performing, we can invest in channels we know will yield good returns — and truly become data-driven marketers.