Our Marketing Attribution in Practice

To be useful, marketing attribution must be practical. That’s why we’ve built entire sample real-world case studies to showcase our service. What do they include? Complete raw datasets – customer site behavior, customer sales data, fiscal calendars, promotions, even sample marketing & media flight information. Over the course of the next few blogs, we will be walking through how Bonsai’s solution works for these real-world case studies.

If you’d prefer to watch a live walk-through, sign up for our Bonsai Attribution Webinar on December 10th at 10 AM CST. We will demo our technology, Pyxis.

“Matt, attribution is a dirty word around here.”

An executive client of ours told me this recently, and I haven’t been able to get it out of my head ever since. It resonates because when we first got into business in February of 2020, we were giddy to help businesses better measure marketing, allocate investment, and do it all with data on hand. I was not crazy about what we would call our service:

Attribution? Even hearing that word made my heart sink.

I couldn’t agree more with the client’s sentiment: I wrote a whole blog about it way back in March. Lets see how ACME loses it’s appetite for advanced attribution.

We start with this typical report. ACME has spent $63K in Marketing across 10 channels in 2020 YTD. This online business has earned $154K in revenue, earning a $2.42 return-on-ad-spend (ROAS) according to the total business profit & loss statement.

The finance team’s evaluation of Paid Media relies on last-touch website attribution. According to their numbers, Paid Media is not contributing nearly enough revenue to the business to justify the expenses. The $1.14 ROAS attributable is not profitable, and the CFO’s recommendation is clear: cut marketing budgets, except Paid Search and Affiliate marketing.

The CMO begs the CEO to reconsider. “Finance’s measurement isn’t capturing the full value of our efforts. They list no impact to our upper funnel awareness channels like Broadcast, Audio and Magazine campaigns. They ignore our strategic efforts with Direct Mail and Programmatic Digital Video.”

The CEO puts the onus back on the CMO. “If you think these data are wrong, prove it.”

So the CMO invests in Marketing Analytics & Data Science and rushes to dive deep into customer, site, and marketing analytics datasets to prove Media does much more than what last-touch attribution shows. All Marketing teams agree — upper funnel media are the missing piece. These channels must initiate many of the conversions and revenues ACME sees on-site. Customer journeys are rapidly stitched together from various first and third-party datasets, and marketing analytics rushes to publish an operational measurement model that reports business revenue to a brand new, sophisticated marketing attribute: First-touch attribution.

At the next ACME C-level meeting, the CFO reads out the results:

  • Programmatic Display ROAS looks worse ($0.61) than before ($0.67)
  • Partners ROAS looks worse ($0.52) than before ($0.65)
  • Paid Search ROAS looks worse ($3.35) than before ($3.71)
  • Affiliate ROAS looks worse ($6.11) than before ($7.61)
  • Magazine, Video, Broadcast, Audio and Direct Mail still have no ROAS measured

Your finance team sees a lower overall Paid Media ROAS ($0.99) vs ($1.14), and thinks the ACME website generated $90K in revenue without Paid Media impact.

The CMO is under even more pressure. The CFO concludes they were right all along. The Marketing teams are upset – their hard work resulted in getting less credit than ever before. The media teams are furious.

It’s no wonder attribution is a dirty word.

Most anyone who has lived through this experience makes the same five decisions:

  1. Go back to the old last-click methodology.
  2. Deal with incomplete data and implausible analysis internally.
  3. Claw what little budgets you can from the CEO and CFO.
  4. Continue to operate on good old-fashioned gut instinct.
  5. Stop future attribution conversations in their tracks.

If you have lived this ACME example, you’ve either skipped this blog altogether, or you’re scrolling down to hear what comes next out of morbid fascination. For those brave enough to continue, let me pause here and start the demo with a question:

What if an advanced approach was actually better?

Link To Demo Video

A New Path: Pyxis Attribution From Bonsai

When you think of a database, what do you think of? Most of us know rows and columns, and what’s required to glean insight — millions of datapoints, calculations and statistics.

But what if your marketing datasets weren’t tabulated, but visual?

What if datapoints weren’t the focus, but instead the relationships between them?

Your marketing and customer analytics datasets all key on a simple concept – customers. Every marketing touchpoint occurred from a customer. Every media impression impacts a customer’s decision path. Every customer’s journey is a series of these interactions and conditions.

Above, we see how ACME customer and marketing datasets look to Pyxis, our proprietary attribution technology. Every node on the chart above is represented by the type of marketing driving the customer touchpoint. If a customer clicked on a Paid Search ad through to Acme.com, that interaction is captured in the “Paid Search” node. If that customer also connected to Acme.com through another marketing channel, an arrow from Paid Search to that source – say, “Direct-to-Site”, captures the customer’s journey. By following this simple approach, we can connect all customers’ journeys across Acme’s marketing efforts. And by representing success – a customer purchase – through green-shading arrows connecting nodes, we easily see which unique touchpoints are most involved in successful customer paths.

Immediately, our perspective on marketing impact changes. First, we see that customers interact across marketing efforts, and that they play a collaborative role in driving customer success. Conversions aren’t “this channel vs. that channel” — the question is how each channel impacts the whole marketing strategy.

You might notice that we don’t see Broadcast, Magazine, Direct Mail, Audio & Video nodes on the chart above. How can we understand which customer paths are impacted by display media?

The great advantage to understanding your customer outcomes through Pyxis is the ability to capture not just characteristics of a datapoint, but characteristics of the connections between the datapoints. Below, see the same dataset, filtered for only customer paths that occurred during and continue subsequent to ACME Magazine campaigns:

Example: Pyxis filtering for journeys occurring during and subsequent to the Magazine media flight. Every customer path above has potentially been influenced by the Magazine campaign.

Pyxis Customer Paths

Leveraging this methodology, we can immediately understand how many unique customer paths touched by every marketing channel in our mix, using a single analysis criteria: does the marketing channel contribute as a customer path touchpoint or condition?

Here we see contributed customer paths – measurable across all marketing channels – thanks to Pyxis.

Pyxis Path Conversion %

Once we measure the customer path, we want to know a simple criteria for success: did the customer path eventually convert? Converting customer paths divided by the total customer paths measured produces Path Conversion Rate. The revenue obtained from converting customer paths, when totaled, are Path Revenue.

The more green the line, the higher the path conversion rate.

Here, we see path conversion rate – the ratio of customer paths that ended in purchase, compared to all measured customer paths. Path revenue represents the revenue resulting from those distinct customer conversion paths.

Now we have customer-centric metrics: paths, path conversion rate, and revenue associated converting customer paths across every marketing channel.

Are we done? Not yet. We’ve successfully described what’s happened with customer journeys, but we haven’t teased out the bigger question: which marketing channels increase the rate of positive customer outcomes?

Conversion Path Lift % And $

Because we can easily visualize and quantify which customer paths a given channel impacts, we can understand what customer paths don’t include that channel, all other conditions equal. The Pyxis framework allows us to quickly calculate the difference in conversion path outcomes (Path Conversion Rate) and economic value of a path ($ Per-Path) when compared to the customer paths where said channel was absent either as a touchpoint, or a condition.

We call these respective metrics Path Lift % and Path Lift $:

Path Lift % & Lift $ measure the customer journey outcomes where a marketing channel played a role versus all customer journeys where that marketing channel doesn’t contribute.

Initiated Paths & Revenue: Insurance Against The Question “What If We’re Wrong?”

One thing no marketing attribution tool has ever adequately modeled is humility: what if our lift model is wrong? How can we account for this? Pyxis values initiated customer paths & initiated revenue, because the only thing you know with 100% certainty is none of your revenue occurs if customer paths hadn’t started in the first place.

To mitigate unknowns and to allow for long-term value modeling, Pyxis retains insight into every customer path back to touchpoints as far back in time as is measurable.

Understand where every successful customer journey started across every channel in your marketing mix, insuring against the risk of lost business from any future investment decision.

Optimize Across Every Customer Journey, Marketing Channel & Time Frame With Our Unified Framework

Combining these insights – customer paths, revenue initiation & lift alongside your total marketing costs gives you a single measurement framework to assess impact & reallocate investments towards channels that help you the most.

Marketing channel costs, customer revenue lift, and initiated revenue risk quantified in a single framework across your media mix. Most experienced practitioners don’t even need a further analysis – the example above is enough to make the next marketing mix decision.

Pyxis: Always Custom & Built To Suit Your Objectives

Pyxis flexes to your specific business objectives. Looking to simply improve profitability in-quarter as best as possible? Weight the model towards Profit. Looking to grow your business over the course of years? Weight your model towards Growth.

Sure, we might have a unified approach and measurement philosophy, but that doesn’t mean your attribution model is one-size-fits-all. Because our model acknowledges short term and long term business value impact uniquely, and accounts for the risk of unknown, your business’ Pyxis model will be tuned specifically to your goals and objectives. Don’t know how to tune your model’s weights? That’s fine — that’s our job.

Immediately assign short term & long-term objective bias & see a single measure of performance: impact ratio across every marketing channel in your mix.

Bonsai’s attribution solution remains a fully managed service for a reason: quality marketing measurement that improves business outcomes can’t come straight out of a box.

Balancing short term profits with long term returns, ACME’s Pyxis model returns an instant analysis across every marketing channel in Acme’s portfolio. We immediate know the channels where investment ratios detract from business value, and by how much ($5.5K). Because our measurement model covers all customer paths and remains consistent across channels, we can quickly identify the best places to re-invest those marketing dollars to increase returns, and for each channel, how much to increase (or decrease) future investments.

Make the changes: A single framework that summarizes customer revenue lift, future revenues generated, contribution, cost, and immediate reallocation recommendation to improve overall marketing impact and business ROI.

Attribution was a dirty word. We’re hoping to convince a few of you that it doesn’t have to be. This is a basic example of Pyxis’ power and our service in action. Believe it or not, everything else you can do with this technology is more powerful than what you’ve seen here.

None of what you saw requires you to place any new code on your website, or farm any more of your customers’ data out to third-parties. You can onboard your business to our framework in as little as a week. Our “pixel-less attribution” isn’t just beautiful – it’s fast and actionable.

For our next chapter, we will dive into something your merchants shout every time they hear about marketing and media impact:

“But what about our promotions? You didn’t factor those in!”

Pyxis does. See how in our next attribution blog.

To see Pyxis in action for your business, contact us to request a demo today.

About the Author

Matt Butler
Ex-Googler of 12 years, Matt was a founding member of Google’s analytical consulting team, developing analytics, statistical forecasting, auction modeling, and machine learning for companies such as Procter & Gamble, Coca-Cola, Unilever, Kohl’s, Best Buy, and many more. He went on to lead global technical partnerships before leaving to found Bonsai in February, 2020.