Marketing attribution is about asking, out of the many marketing channels you are using and paying for, which marketing channels actually work? Which ones don’t? And are you paying the right amount for the ones that do work?

Measuring marketing attribution is one of the biggest challenges for high-scalability performance marketing, and I don’t believe there’s an easy solution without using human judgment.

Now, I know many vendors run around saying they know how to accurately measure attribution — if only you’ll spend six figures per year to implement their solution.

Their claims are partially true: tracking is essential to the attribution problem, and often the process of implementing these systems can help you more easily and thoroughly track and analyze what marketing channels are working.

But it’s not enough. No system or artificial intelligence can automate the problem away, because attribution has inherent, systemic biases that can only be overcome with good (and necessarily human) judgment. Let’s see why.

What is marketing attribution?

Marketing attribution is the process of determining what marketing channels contribute to the final success of a sale, and how much value each of them adds individually to that success.  And of course, marketing attribution should tell you which particular campaigns or individual “touches” within each these channels works best.

This is important for any marketer who uses more than a single channel to acquire leads or customers, because it tells you whether you are overspending or underspending on particular channels.

What are the common problems with marketing attribution?

There are two types of problems that interfere with accurate marketing attribution, some within channels, and some when channels interact with each other.

In-channel marketing attribution problems

Missing or uneven tracking. Any marketing attribution system will fail if the conversion tracking within that channel is absent or patchy — that is, some campaigns are tracked correctly and some are not. If tracking is wholly broken, cost per conversion will look bad. If you spend less as a result, the channel will enter a death cycle. If tracking is partially broken, then you’ll starve campaigns with broken tracking, even if they’re successful. Oops!

Badly optimized channels. A channel where best practices have not been introduced will underperform and lose resources, leading to a death spiral. In other words, you need to make sure that every channel is well-tracked and well-optimized before you trust cross-channel attribution metrics. (Here are some classic problems in search campaigns that tell you when you’re probably not optimal.)

Internal last-touch reporting. In the past, nearly all marketing channels attributed the success of a particular campaign to the last advertisement that touched it. These “last-touch” attribution models distorted decision making because they gave credit to the bottom-of-funnel advertising, like brand searches or remarketing ads, which added limited value to the customer. (This is because a lot of other marketing had done all the heavy lifting to make the customer aware and convince the customer well in advance). Not only that, last-touch reporting made marketers’ lives even more difficult because they rendered invisible ALL the other ads that might have touched a customer before they finally converted, greatly diminishing understanding of the customer’s multi-touch journey. Nowadays, many online channels have fixed this and let you see all touches. But sadly, many have not, and these same channels often make it tough to pull data about all the touches that reach an individual prospect.

Cross-channel marketing attribution problems

No unified tracking system. Many companies just don’t have a single system to track and store customer tracking across all their channels. Even when they do, they may forget to implement the necessary tracking codes or other disciplines to record the source of each touch. It’s crucial for this to happen somewhere, be it your CRM system, your marketing automation system, or a purpose-built attribution tracking system (like Bizible).

Easily undercounted channels. Most attribution systems can only track customers who arrived from tracked channels, and for the most part, the only channels which can be easily and reliably tracked are online. This means that all your offline acquisition channels will tend to be shortchanged by your attribution system, if they are even tracked at all.

Undercounted channels usually include:

• Direct mail
• Broadcast (radio/TV)
• Out-of-home advertising (street signs, billboards)
• Events (trade shows)
• Pure brand and awareness advertising, both online and offline
• Search prospecting campaigns (particularly if mixed with brand campaigns)
• …Or any channel with a call to action resulting in an inbound phone call (since call centers very typically “lose the scent” of prior activity)

Easily overcounted channels and campaigns. Some channels also play the attribution game all too well. Typically, they are online, get the last touch before a conversion, and have very high frequency. Unfortunately, they also tend to be of low marginal value, hard to scale, far down the funnel — and sometimes, really easy to abuse. These channels often “steal” conversions from other acquisition channels higher up in the funnel, making it harder to know where customers are first becoming aware of your company — and depriving you of crucial marketing intelligence. Improved attribution models are also easily dumbfounded by these channels because they often count for so many touches.

Remarketing is a classic example. I worked for one company where they were convinced that remarketing ads were the best and cheapest channel they could possibly use, because the remarketing ads successfully stole attribution from nearly every other acquisition channel. They actually bid 10 times normal for remarketing ads, used three different vendors for ads, and never removed anybody from their target lists, even if you bought!

The result was insane: If you touched the website for just a few seconds, for the next 6 months the entire Internet appeared to be wallpapered with ads from this company. The more they did this, the more conversions were attributed to remarketing, because there was no way to avoid the ads if you did anything on the Internet! This practice also made it very difficult to distinguish success in any other acquisition channel. Needless to say, we had to roll that back so we could even “see” results in any of our other channels. (Read more about the pitfalls of remarketing.)

Overcounted channels typically include:

• Brand keyword search campaigns
• Remarketing
• Digital marketing based on uploaded, already-acquired email lists (“custom audiences” and the like)
• Organic SEO on brand terms
• Email marketing (especially high-frequency, transaction-oriented newsletters or alerts)
• Online coupons and other easily-disseminated price promotions
• View-based display ads

How do we fix marketing attribution?

So how do you fix a problem like marketing attribution? Here would be my checklist:

1. Set up a unified tracking system. As stated above, your current CRM or marketing automation system may already be able to archive and report on all trackable customer touches. Read the instructions and start using it that way.
2. Track ALL touches you can. Try to source and track every single touch, even when it’s hard. Use UTM codes religiously, and don’t forget to record what all those codes mean. Put special effort into tracking offline channels: scan people who visit your trade show booth, use unique toll-free numbers for offline campaigns, ask people who call in where they just came from, etc.
3. Optimize your channels. Spend a month (or two, or six) making sure that your digital marketing channels are performing as best they can.
4. Use your judgment. Recognize that some channels (remarketing, brand search keywords) will mercilessly steal credit from others (offline, brand advertising). Allow some channels to have higher CPAs for the overall health of your marketing efforts — if you block the top half of your funnel, you’ll never get leads to the bottom half.
5. Measure each channel intelligently. Overcounted channels are difficult to attribute in isolation since their gain is another channel’s loss. To correctly measure their value, you have to turn the whole channel (or campaign) on and off, and measure the total difference in impact on your overall acquisition. (A pure “stealer” will have no impact on total conversion, while an added-value channel will steal some conversions but also add marginal sales.)

Unfortunately, none of the steps above can be automated away by a marketing attribution system. They all require people using critical reasoning to get right. Score one for the human beings over the robots!