Attributing Unattributable Conversions

"I have too many conversions being attributed to direct or organic social -- what do I do?" We hear some version of this issue from almost every marketer we speak to, whether they're buying just FB+Google or have expanded to "Harder to Track" channels like Linear TV, Direct Mail or OTT.

They're frustrated because although they believe their marketing is working, it's hard to prove when upwards of 20-35% of all conversions aren't being attributed back to marketing. This makes it harder for marketers to know the true ROI of any of their channels, to figure out where to invest / divest and to maintain sanity.

How big a problem is this?

We wanted to get a sense for how big a problem this actually was. The quickest way to get some sense of the size of the problem was to look across our customers and see what percentage of their conversions are attributed to Direct (just so we're on the same page, this means there were no marketing touchpoints -- the customer magically just went directly to the website and converted). Note: Analyzing this to include organic traffic is a logical next step but gets a bit more complicated as a lot of marketers don't properly have organic search mapped in Google Analytics

Across our customers roughly 21% of all conversions are being attributed to Direct. That means your CPAs / ROAS can also be off by ~21% across all of your channels if you assume all of those conversions really were a result of marketing.

How much does Rockerbox help?

In a second I'll dive into the tactics that we use to help our clients here. But so as to not bury the lead, here are some quick results:

That's an ~43% decrease in conversions that are attributed to Direct traffic. While still too high in aggregate, it's a huge step in the right direction.

Background:

Before we dive into what's causing the problem, lets get some background. To begin the issue has many different names / manifestations:

- Direct % too high

- Organic traffic receiving too much credit

- Too many unattributable conversions

- Missing marketing touchpoints

All of the above are names for the same problem. For some reason a given conversion isn't being properly attributed to the marketing that drove it.

How can you lower your number of unnatributable conversions?

1) Synthetic Events

2) Including all your marketing data

3) Post purchase surveys

4) Promo code mapping

5) Identity resolution

1) Synthetic Events

As much as we wish we could, it's impossible to get every piece of deterministic data that we'd like for all of your marketing. There are clear cases as in Linear TV where impressions and clicks just don't exist all the way to more nuanced versions like Facebook where clicks are available deterministically but views aren't.

Synthetic Events enable Rockebox to determine the contribution of a channel even if we can't track it 100% deterministically. This is done by building techniques on a per channel basis to model against the data sets that are available for that channel, however incomplete they are. Lets take a look at some examples:

Both of these are cases where without Synthetic Events, you'd be stuck attributing conversions that are actually tied to marketing to either Direct or Organic.

How big of an impact can you get by implementing Synthetic Events? This obviously depends on your marketing mix but we've seen upwards of 24% reductions in unattributed conversions by Rockerbox clients that leverage Synthetic Events.

2) Including all your marketing data

There's a lot here but to keep it short, if you're relying just on click based data like in Google Analytics you're going to be misattributing traffic. The simplest example here are views -- every display channel is going to take credit for views but this data won't ever get into GA. I'll skip over the whole rabbit hole of whether display views deserve credit for now, but this is a good example of how conversion credit will be misattributed to Direct or Organic if marketing touchpoints are missing. What channels does this apply to?

- OTT

- Display / Programmatic / Video

- Native

- Linear TV

- Direct Mail

- Calls

3) Post Purchase Surveys

Also commonly known as "How Did You Hear About Us" surveys, this is a really simple yet effective technique. If you want to know how someone heard about you why not just ask? As an aside there are definite limitations here and best practices to help ensure this is done properly -- randomizing response order, including channels that you aren't actively running to build baselines. But all that aside, this is still a great way to uncover how someone that would have otherwise been attributed to Direct should be attributed.

How big of an impact can you get by implementing Post Purchase Surveys Events? We've seen upwards of a 10% reduction in unattributed conversions by leveraging Post Purchase Surveys.

4) Promo Code Mapping

This is another really simple yet effective technique -- map the promo codes that your customers are using back to marketing channels. For example, if someone comes to your site via Organic Search but uses a promo code associated with Influencers it's a decent idea to give credit to your Influencer efforts instead of Organic Search.

Given how obvious this is, why isn't it done more in practice? For one, it's been logistically difficult -- there's no easy way to override credit in GA based off Promo Codes. Another reason is that similarly to Post Purchase Surveys, the results can get a bit messy -- people share promo codes and products like Honey dirty up this data set. Nevertheless, we've seen customers reduce up to ~4% of their unattributed conversions by leveraging Promo Code mapping.

5) Identity Resolution

This is another huge topic that I won't dive into too deeply right now. But high level, we all know that your customers engage with your brand through numerous devices. They often have long lags between when they visit, resulting in cookies expiring. All of this results in user browsing paths getting disconnected and conversions being improperly attributed.

Rockerbox has built technology to help here. Of all the solutions listed above this one has the widest variance on a per client basis but being able to bridge customer paths together is a huge win for properly attributing your marketing.

Conclusion

So what does this mean for your marketing and your company? How big of an impact can these changes really have on you? Check out below for how big a swing we've seen for some of our customers in decreasing their Direct %:

On the one hand we have clients where the delta is pretty small -- decreasing only 4% for example (down from 25% direct to 21%). For other customers the impact can be huge -- a 72% decrease in Direct % being the largest (decreasing from 32% down to 9%).

What explains the wide range in differences of decreasing Direct %? For one not all of our customers have implemented all of the techniques outlined above. I'd certainly expect a larger decrease for a customer that implemented 4/5 of the techniques listed than just one. Besides that is the standard, yet not too reassuring answer, of all advertisers and companies are different.

Interested in decreasing the number of your conversions that are unattributable? Lets talk!

P.S.

Another major, related problem is the incorrect attribution of events due to inaccurate collection of marketing event data. In this post were focusing in on the impact to Direct traffic, but other major marketing channels like branded paid search are similarly impacted by including additional marketing touch points