February 16, 2010
Although the industry has made vast improvements in attribution, it’s hard to know the impact of every marketing dollar. Thankfully, database analysis techniques are giving some marketers more insight than ever.
We spoke with an expert in marketing attribution to learn how he uncovers the marketing impact of murky channels. Take a look to gain insight, but be prepared: This will not be easy.
Uncovering which marketing efforts deliver the best results is more complicated than it sounds. You can easily see how many conversions came through a channel. But what’s less clear is whether that channel generated the conversion, or merely accepted a conversion driven by another channel.
Say a consumer encounters television, radio and print ads for a product. Each ad introduces and explains the product. Then the consumer uses a search engine to find and purchase the product. In most marketing reports, search will get full credit for the sale. But what about the other channels? Didn’t they contribute?
This is the riddle of multi-exposure marketing attribution, something Anto Chittilappilly, CTO, Visual IQ, spends much of his time thinking about. Visual IQ is a marketing intelligence firm that helps companies more efficiently spend their marketing budgets.
"Attribution is like a mirage. It’s out there," he says. "Many people just aren’t sure how to do it."
Chittilappilly’s team recently studied data from the last three years for more than 500 of their clients. They found that companies employing multi-exposure marketing attribution were able to increase return on investment by 22% year over year. That is realized through:
o 28% increase in revenue year over year
o 17% increase in customer conversion year over year
o 23% increase in average order values
o 26% increase in average customer lifetime value
As nice as those figures look, doing multi-exposure attribution is difficult for a variety of reasons. Take a look below to see some of the steps and challenges you’ll encounter, and Chittilappilly’s advice for getting through them.
Step #1. Define your goals
The first question you should ask yourself is "Why do I want to do this?" You’ll gain a wealth of insight by better understanding what portion of your marketing budget is helping your business, but you must set performance goals. They could be:
o Higher conversion rates
o Higher average order values
o More efficient spending
o Lower cost per action
Step #2. Get executive support
You need support from the top to make this work, because you’re going to need a lot of people’s time and effort.
Also, you might be stepping on a few toes along the way -- especially if your marketing team has several departments. For example, an attribution analysis might show that 25% of conversions that come through paid search marketing are partially attributable to other media. This is likely to upset your in-house paid search team.
Step #3. Collect the data
This is arguably the hardest step in the process. You need to pull data from every marketing channel you use.
Collecting data will be somewhat easier for online channels. Major ad networks and search engines have systems to directly send data from their networks to your database.
Getting data from offline channels -- such as newspaper and radio advertising -- is more difficult. But you must get all the data that’s available.
"Sometimes it’s very poor, aggregated, very high-level data -- but that is important. It’s better than nothing," Chittilappilly says.
Marketing channels such as print, out of home, television and radio often only report aggregated data across all users who encountered your media. The datasets normally have aggregate attributes such as:
o TV station
o Gross rating point
This data is combined with your team’s customer data, and should be supplemented with third party demographic data.
Step #4. Analyze data on a user level
You need to analyze your data on a user-by-user basis and see the marketing that reaches high- and low-value consumers.
Some channels, such as search and mobile, offer more user-level data than others, such out-of-home ads. The data tends to fall into three categories:
- User-level conversions and impressions
Channels offering this data show exactly who saw your ads, and exactly who converted.
- User-level conversions only
This type gives limited insight into who saw the ads, but clearly shows who converted. For example, it is easy to track response from a billboard with a unique phone number and a call to action, but it is only possible to estimate its total impressions.
- No user-level data
In these cases Chittilappilly’s team uses panels, surveys and sampling to gauge conversions and impressions, he says.
-> Tip: Be prepared to analyze huge amounts of data
Companies with large marketing programs will need large-scale data management. The databases can be measured in terabytes (1 terabyte = 1000 gigabytes) and might record millions of records per month.
Step #5. Create cross-channel metrics
After the data is gathered and organized, you need to use the same measuring stick to gauge each channel. Since television, online advertising and email marketing all use different reporting metrics, this will require creating universal metrics to compare the channels.
Two examples that Chittilappilly’s team uses:
- Initial action rate
This metric roughly translates to "clickthrough rate" when applied to online channels, but it is also applicable to offline channels. Any advertisement that causes a consumer to act -- perhaps by searching online, contacting a company or clicking an ad -- is said to have caused an action.
- End action rate
This metric roughly translates to "conversion" rate, and represents the percentage of people who an ad reached and who fulfilled the ultimate goal of that advertisement.
Step #6. Identify trends
Once the data is prepared, it is time to crunch the numbers and look for trends. The task is to understand which channels are achieving marketing goals, and which are wasting money.
Analysts should identify which marketing touch points are driving people to high value conversions, and identify the attributes of those touch points. Attributes might include:
o Type of media, such as video, audio or online display
o Frequency -- how often someone saw the ad
o Time lag -- the time between exposure and conversion
o Time of day
o And many others
"There are so many attributes you can look at," Chittilappilly says.
- Weighing the attributes
You can look at attributes such as the gasoline price index, or the weather, but they are likely to not have much impact on purchase intent. This is why it is very important to gauge the importance, or weigh, the attributes.
By looking for trends in the data, your team should be able see which attributes have the most marketing impact, and be able to numerically compare that impact to other attributes you’ve analyzed.
A key area to analyze is the difference in the media shown to people who converted and the media shown to people who did not convert, Chittilappilly says.
- Identifying "true metrics"
Once the data is analyzed, Chittilappilly’s team creates a set of "true metrics" for each marketing channel to tell marketers how many conversions, impressions and actions each channel truly drove. For example, channels through which conversions are not completed might have a positive conversion rate if they’re statistically shown to improve conversion rates in other channels.
These metrics are used to compare channels and optimize a team’s marketing spend.
Step #7. Continually measure and tweak
Don’t give up yet. Continually collect new data, prepare it, add it to the database and continue to crunch it -- because the results will fluctuate over time.
The first few rounds of analysis should help your team make better marketing investment decisions. The most significant changes will likely occur early on, and you can tweak as the data naturally shifts.
Step #8. Wait for results
Although you can have this process up and running within three to four weeks, have patience. Chittilappilly suggests waiting at least three to four months before seriously considering the results.
"Within a year, you can expect a 20% to 30% -- our average is 28% -- increase in revenue with the same kind of spend. This is because you’re getting a higher number of conversions and your conversions should bring in more revenue," he says.
Useful links related to this article
Attributing Conversions to Assisting Keywords Lowers CPA: 5 Steps to Optimize Bids
Special Report: How to Use Predictive Modeling to Pick Your Best Prospects & Boost ROI Up to 172%