Four Alternatives to Last-Click Attribution

Four Alternatives to Last-Click Attribution

Attribution Modeling

Advertisers have become accustomed to the belief that the final click that leads directly to the conversion is the most important click – hence the affinity for last-click attribution. But it’s important that businesses transition away from last-click attribution. That’s because last-click attribution fails to account for the value of the entire conversion path.

Most marketers would agree that their brand campaigns drive a large number of conversions and have very low costs per action (CPAs). Of course the cost per clicks (CPCs) in brand campaigns tend to be very low, but those campaigns are also benefiting from last-click attribution models.

Let’s think about a customer journey for a moment. With the holiday shopping season upon us, many of us will start our search for the perfect gifts with some online searching. Here’s how one of my searches might look:

Top electronic gifts 2018 -> Fitness Trackers -> Top Rated Fitness Trackers ->Apple Watch

In the example above, the brand campaign housing the keyword “Apple Watch” would get 100-percent of the conversion credit if you use the last-click model. Clearly, I did not start my search on a branded keyword, yet the brand campaign gets full credit. When marketers use last-click attribution, they generally see that non-brand keywords achieve low conversation rates and high CPAs, and brand keywords achieve high conversion rates and low CPAs. But is this approach really a fair way to evaluate our campaign and keyword performance?

Marketers have all seen non-brand keywords fail to work well in a campaign. They may be costly to run, and rarely do we see strong conversions. I have paused my fair share of non-brand keywords as I can’t justify their worth to my clients. Not surprisingly, I see search volume decline; and although my CPA often times improves, my overall number of conversions also begins to decline. What we have been missing is the ability to see the value of the entire conversion path.

Alternative Models

One of the main focuses for Google this year has been transitioning clients from last-click attribution into a model that gives credit to each paid click in the user journey. Currently, there several different attribution models available in Google Ads.

Let’s take a look at some of the choices:

Data-Driven Attribution

The model Google recommends most is data-driven attribution, which uses Google’s machine learning technology to determine how much credit to assign each click in the paid search journey. This attribution model is all based on an advertiser’s own data and continues to “learn” over time.

Data-driven attribution takes both converting and non-converting paths into account, and it’s powered by dynamic algorithms that assign credit to touch points based on fractional credit. Google recommends choosing data-driven attribution when available. Unfortunately, this attribution model is not always an option as it requires 15,000 clicks on Google search and 600 conversions over a 30-day period.  Although smaller advertisers will not have access to this attribution model, there are still some good options available.

Linear Model

The linear model distributes the credit for the conversion equally across all clicks on the conversion path. If it takes four clicks for a searcher to convert, each click receives an equal part of the total conversion credit.

Time Decay Model

The Time Decay Model gives more credit to clicks that happen closer in time to the actual conversion. For example, if the path to conversion takes five clicks, the time decay model would assign an increasing proportion of credit with each subsequent click, with the final click that led to the conversion receiving the most credit.

Position-Based Model

The Position Based Model gives 40 percent of the conversion credit to the first click, 40 percent to the last click in the conversion path, and the remaining 20 percent across the other clicks on the path.

A Recommended Approach

As mentioned above, if the data-driven attribution model is an option for your campaigns, always choose that. But if you don’t have enough data available for that option, how do you go about choosing among the other options? Google offers a few suggestions:

  • Choose a time decay model if your client has a conservative growth strategy, is a market leader, and has little competition. In this scenario, the final clicks in the conversion path will get more credit.
  • If your client is growth oriented, new to the market, and is facing a lot of competition, choose a position-based model where the first and last clicks in the conversion path will get the most credit while the clicks in between will receive a smaller portion.
  • If your client falls somewhere in between, you may opt for a linear model, giving equal credit to all the clicks on the conversion path.

There is no absolute right or wrong choice, and any of the models you choose will give you better insight into the complete conversion path more than the last-click model can. Google also offers an attribution modeling tool in Google Ads that allows you to change attribution models and compare results among the different model types.

Outcomes of Different Models

No matter what attribution model you choose, you should anticipate a decline in brand conversions and an increase in non-brand conversions. The actual number of conversions will remain the same regardless of the model you choose. But you will see fractional conversions reported, indicating each campaign/ad group/keyword that played a role on the conversion path.

So let’s revisit my holiday shopping search from above:

Top electronic gifts 2018 -> Fitness Trackers -> Top Rated Fitness Trackers -> Apple Watch

If I used a position-based attribution model, here would be the new breakdown for conversion credit:

  • 40 percent of the credit would be given to “top electronic gifts 2018.”
  • 10 percent of the credit would be given to “fitness trackers.”
  • 10 percent of the credit would be given to “top rated fitness trackers.”
  • 40 percent of the credit would be given to “Apple Watch.”

Using last-click attribution, I would see keywords “top electronic gifts 2018,” “fitness trackers,” and “top rated fitness trackers” appear to be poor performers, as all of the conversion credit would have gone to “Apple Watch.” Conversely, if I were to use the position-based model, I would see that all of those keywords together played a role in the conversion path — and I would have a better understanding of the value of my non-brand keywords. This insight would allow me to make smarter decisions when optimizing.

Without question, we are able to make smarter decisions when we have a better understanding of the full conversion path. I suggest taking some time to experiment with the various attribution models using the attribution modeling tool in Google Ads. Based on your findings, select the attribution model that best suits your goals. I have found the additional conversion path insight to be valuable.

For more insight into how to improve the performance of your online advertising, contact True Interactive. We’re here to help.

Photo by rawpixel on Unsplash

Five Options for Attribution Modeling With Analytics Engines

Analytics Attribution Modeling Spotlights

Man-outstretched-arms2Making adjustments. That’s the key to success, regardless of whether we’re talking about half-time adjustments in a big game or changes you make in your marketing plan. You have to observe, learn and adjust to what’s happening in your field of play.

That is what you will learn from our article “Five Options for Attribution Modeling With Analytics Engines,” a version of which initially was published on MarketingProfs (December 11, 2014), though its principles remain relevant today.

The objective of attribution modeling is to evaluate (by using analytics) what led to every sale—so we can replicate success.

In that post, we described five methods that command the most attention or offer the most promise.
MarketingProfs

  1. Last Click
  2. First Interaction
  3. Position-based
  4. Time-delay
  5. Linear

While “Last Click” used to be the only game in town, the rise of powerful – and often-times free – analytics tools like Google Analytics have greatly expanded the playbook for marketers.

Busy marketers often stick with what they’ve previously used. We get it. Learning and implementing a new model can take time. But the payoff is often worth the effort. For example, here are “3 Reasons to Drop Your ‘Last-Click’ Crutch.” When you’re ready to change your game plan, our previous post “4 Alternatives to Last-Click Attribution Modeling” also gives you guidance on choosing the model that’s best for you.

But never fear. Despite our admonishment to drop the “last-click” crutch, we let you know in “3 Tips for Managing Attribution Modeling” that it’s okay to use that crutch to prop up your analytics.

Today’s digital commerce is complex. In turn, smart marketing dictates that you make adjustments and invest your budget dollars where they will do the most good. We believe our tips will help point you in the direction of success.

Google Close Variant – A Match Made in Heaven?

Analytics Close Variant Matching

Puzzle_Piece_KWFLast post, we discussed how Google’s matchmaking algorithm Close Variant Matching (CVM) brings search terms and paid ads together at auction as applied to the Phrase Match and Exact Match categories. Recent changes to Google’s ad-serving policy have the digital marketing world asking – Is obligatory CVM a match made in heaven? Or will it end in heartbreak for paid search managers?

Before dissecting pros and cons of Google’s new policies, let’s examine how to capitalize on the latest shift in paid search. Here are a few key takeaways:

  1. Rethink your campaign structure – Is your campaign poised to profit from Close Variant Matching? If not, it’s time for a change. Start by measuring your campaign results and optimizing for CVM’s inclusion of atypical spellings and abbreviations. An aside – Think Google Quality Score when building campaigns, too.
  2. Check your work – and often – Running Search Query Reports (SQRs) frequently can help you stay abreast of performance. The new norm of paid search management is closer monitoring.
  3. Boost your top performers – Find your best key terms and re-add them as “Exact Match” keywords. This will boost traffic on your hottest terms.
  4. Yank irrelevant traffic – Adding negative keywords will sift out irrelevant phrases, minimizing spend for ill-fitting keywords and stimulating your paid search program.

What’s the Verdict?
Now for the debate: Does Google’s mandatory matchmaking policy help or hurt paid search marketers? While not all AdWord auctions are the pairing marketers would hope, the verdict may differ between campaigns.

Most simply won’t notice the change. Why is that? Well, CVM has been around since 2012 and was always the default selection in AdWords. Most marketers had no reason to avoid Close Variant Matching, since it pulled in even more potential eyes to view their content.

The pros? Close Variant Matching could lift campaigns honed on Gen Y searchers, whose abbreviations and SMS acronyms already are the stuff of legend. They are less likely to complete a word or phrase in a search bar versus older citizens of the Web.

Another factor – the speed of communication and commonality of analytics-backed search suggestions are so commonplace, many don’t take the time (or *gasp* may not know how) to spell the names of specialized products correctly. Rather than leaving these queries in ad-serving oblivion, Google’s latest stance on bidding for keywords ensures compatible searches are served relevant ads.

There is a downside to the change as well. With less control comes a greater potential for error. Sometimes a match may ring true to Google, but not represent a qualified lead for your business. The only way to mitigate the impact of this possibility is to monitor closely your paid search campaigns for leaky keyword results.

To dig a little deeper into Close Variant Matching and its implications for your paid search program, be sure to read our last post and check out our in-depth article in Social Media Monthly.

3 Tips for Managing Attribution Modeling

Analytics Attribution Modeling

For months, our rallying cry has been “Drop your Last-Click Crutch!

Our argument is straightforward: How can marketers track the digital pathway buyers followed to reach them if the only clue is the buyers’ final step?

Sure, not long ago, we had little choice but to rely on the last click. The data connecting each touch point along a customer’s path to purchase wasn’t available or reliable. What’s more is the tools for analyzing this type of information were neither sophisticated nor affordable.

And yes, time was spending most of your marketing time and budget figuring out whether or not marketing was working just didn’t make a lot of sense. What did make sense was giving the last marketing tactic in a campaign that produced a sale 100% of the credit, because the data was fresh and unambiguous. We knew it was real — even though we also knew a mix of campaigns and as many as five touch points preceded that sale. In those days, we just couldn’t see those facts in front of us.

Well, today we can. Google’s recent research tells us as many as nine of every 10 marketers has access to some form of analytics tool, and the reasons for using those tools are more compelling than ever:

  • 90 percent of media interactions today are screen-based, a statistic that includes TVs, PCs, smartphones and tablets
  • 67 percent of buyers start shopping on one device and continue on another.

In other words, today’s digital commerce is complex. In turn, Smart marketing dictates that you invest your budget dollars where they will do the most good. And in my last post, we offered four ways you can approach this analysis – i.e., we’re pointing directions for stepping beyond the last click.

We understand that throwing down that last-click crutch and venturing into new analytical terrain can be intimidating, especially without a trusted means of support. So, here are a few pointers for approaching attribution modeling with new techniques:

  1. Know Before You Go. Earlier, we referred to the way today’s customers come to your company as a digital pathway. So, let’s stick to the analogy. If you want to discover a route to a new destination, you look at a map. In the case of attribution modeling, you should plug data into your analytics tool – just like you enter addresses into mapping software before hitting the road. Be sure full details – tactics, dates, dollars, etc. — about each digital campaign are entered into your modeling software.
  2. Walk First, Don’t Run. Explore alternative attribution models one at a time before trying combined or comprehensive analysis with any given technique. Before attempting an obstacle course, walking along the path and examining each challenge is a great way to prepare and minimize stumbling when the time comes to run the race.
  3. Carry Your “Last-Click Crutch.” Ok, we confess we were going for dramatic impact when we told you to “drop” your crutch. But that doesn’t mean we believe you should disregard it. Last-click analysis plays an important role in attribution modeling. Your picture of your customers’ journey would be incomplete without being able to see their final steps. So, you shouldn’t lean on that crutch, but you should carry the knowledge of how to use it along your way.

Expect some confusion at first but your competency will grow with time. And after a few trips, you’ll be finding efficient shortcuts and some more scenic routes than you’ve traveled before.