In this post, I will detail the logic behind and process of incorporating margins into bidding at the CPA and CPC level, in case you prefer manual bid management to cost-per-acquisition based management.

Depending on available data sources, you might have access to internal analytics considering margins or you may be limited to ad server data. For the purposes of this article, I am going to assume that we’re starting with gross revenue aggregated to the keyword level (standard AdWords and Google Analytics reporting). As a result, this process is applicable if you have access to basic analytics or ad server data sets.

An ad serving reporting platform will only display revenue that is captured by the associated pixel or integrated analytics platform. Because pixels are placed on order confirmation pages in the shopping cart, standard practice is to display gross revenue in the reporting platform rather than net revenue. Marketers must manage advertising performance to a return on ad spend (ROAS) or cost per acquisition (CPA) backed out from profitability goals using gross metrics.

At a high level, it’s relatively easy to apply margins to gross revenue to extrapolate thresholds for profitability. If the average margin per sale is known, then the theoretical maximum an advertiser can pay for a transaction is equal to the margin (assuming a customer lifetime value of 1x retail price of their first purchase).

The idea behind average net profit per conversion is to identify the point in time when net revenue is equal to marketing expenses. Using a 30 day cookie, if we manage to a CPA equal to average net profit, then we are establishing a cash flow break-even point of 30 days. Advertisers with more robust analytics can do this same calculation to identify the point where marginal cost is equal to marginal revenue (0 economic profit) on a sliding timeline to determine how much they want to hedge short-term losses in favor of long-term profits.

The Calculation

Gross Margin = Revenue – Cost of Goods

Net Margin = Gross Margin – Operating Expenses

Consider the following situation:

• Average Order Value: $100
• Cost of Goods (cogs): $50
• Gross Margin: $50
• Operating Expenses: $25
• Net Margin: $25

In this situation, the net profit per order is $25, meaning that the maximum acceptable CPA would be $25 on a $100 order. If all orders have the same revenue, cogs, and operating expenses, then we could easily manage advertising expenses to a maximum $25 CPA or minimum 4:1 ROAS, at which point there is no profit to the business on a 30 day basis. If the business has a CLV greater than 1x, then we can manage to a CPA higher than $25 or a ROAS lower than 4:1 and still make a profit.

In most cases, businesses have multiple products with unique margins. As a result, we can make an accurate set of assumptions by using global average margin or category average margin to more accurately calculate average profit per transaction within an account. Simply create a spreadsheet of category margins and run a lookup against this spreadsheet to calculate average profit at the keyword level.

Because CPA = CPC / Conv Rate, we can easily add in profit per click:

So, depending on how much profit must be realized in the initial 30 days, Profit / Click defines the max CPC for a 30 day economic profit of 0. If profit in 30 days is required, simply multiply the profit/click or profit/conv by the required economic profit percentage to determine the ideal CPC or CPA.

Below is the same table with a 50% economic margin and updated iCPC / iCPA:

Conclusion

What makes each bidding system unique is the process by which observed CPCs and CPAs are optimized toward ideal CPCs and CPAs. Some just update to the new ideal number, while others slowly move in that direction so as to not shock the system. There is no right or wrong way, just options and iteration upon iteration. This is a good foundation to start with, but by no means the final product.

Please comment below, as I’d love to hear about the directions different members of the community have taken with their internal tools.

When asked if an advertiser should consider a 3^rd party technology to help manage ad spend, I would undoubtedly say “yes” because Google Analytics, AdWords, and Merchant Center were (and remain) imperfectly bound systems requiring a lot of proprietary work to merge and make informed decisions.

Much has changed in the last year. Google has made some significant changes (PLA, rules-based and improved algorithmic bidding, attribution modeling, enhanced campaigns…), making it increasingly difficult to leverage third party tools to gain an advantage over free tools like AdWords and Google Analytics. Having been on the inside at a technology house and heavily involved in marketing third party technologies, this is a difficult reality to embrace.

I’m going to go through the historical pros and present realities of 3^rd party paid search technology platforms versus their free ad server counterparts (AdWords, Bing Ads, and Google Analytics). For the record, I have no personal bias toward or against particular technologies and all statements below are generalized toward the entire industry.

Tracking

One of the most blatant shortcomings in the Google tracking ecosystem has always been conversion tracking — AdWords conversion events and Analytics goals/e-commerce tracking.

AdWords Tracking Setup

The inability to create composite tracking events, parse conversions event types for bidding, or modify conversion data after the fact can be frustrating if you depend on AdWords or Google Analytics as the only means of tracking conversions for your marketing program.

Google Analytics Goal Setup

Third-party providers have stepped in to provide comprehensive and flexible tracking systems with deep integration capabilities. But, while it sounds nice in a pitch, complex tracking integrations are only relevant if you have complex tracking demands. Some businesses require complex tracking solutions such as composite metrics or latent conversion attribution; however, the reality is that most advertisers have simple advertising tracking demands – either a CPA (cost-per-acquisition) or ROI (return-on-investment).

AdWords, Bing Ads and GA all do a very good job with CPA and ROI. And, with a little legwork and a half-decent internal database, most tracking needs can be distilled down to a CPA: the average net present value of a conversion.

Bidding

Building on the relevance of simple tracking demands, at face value, AdWords offers a fairly similar tool set compared to even the most flexible 3^rd party platforms:

• AdWords
  • Manual CPC Bidding
  • Rules based bidding
  • Algorithmic rank based bidding
  • Algorithmic CPA based bidding
• 3rd Party Software
  • Manual CPC Bidding
  • Rules based bidding
  • Algorithmic rank based bidding
  • Algorithmic CPA based bidding
  • Algorithmic ROI based bidding

Without any additional analysis, the only way to manage to a performance metric in AdWords that’s not CPA is via rules — an option that has been around for years and refined via 3^rd party tools, but only a recent addition for AdWords.

As previously mentioned, any revenue-based metric can be distilled down to a CPA, and as the ad server, Google has a massive advantage over any other algorithmic bid management provider.

Consider the number of variables a 3^rd party system can leverage for bidding in AdWords: in the enhanced campaign environment there are 2 bids per keyword (Computer/Tablet & Mobile). If the platform is robust, it may offer up to 8 day-parts, making 16 possible bids per day per keyword.

On the other hand, AdWords conversion optimizer can manipulate bids for all three devices, across both Google.com and search partner networks per unique impression. That’s several orders of magnitude of variance in possible bids per day!

The most common argument against conversion optimizer is that it’s a black box – but so are 3^rd party platform algorithms. The inherent advantage of the ad server is too much to ignore.

Large Accounts

Prior to Enhanced campaigns, marketers had a clear incentive to segment campaigns by as many values as possible to best understand unique click and conversion behaviors by device, network, etc. Take a comprehensive account with 50 campaigns and break it apart by three devices (computer, tablet, mobile) and three operating systems (windows, mac/iOS and others), and you’ve got a 450-campaign behemoth that’s borderline unmanageable due to size constraints.

Accounts that get too big become extraordinarily difficult to manage and manipulate as even the simplest tasks require mind-numbing repetition leading to human error.

As a result, 3^rd party platforms are built around automated processes and scripted bulk changes evolved to bear the burden of repetition.

For better or for worse, Enhanced Campaigns force marketers to reduce that load back down to, in my previous example, 50 campaigns – effectively negating the benefits provided by the 3^rd party tool sets designed to streamline bulk processes.

Bing

With all this talk about AdWords, we can’t forget about Bing Ads. A fleshed out Bing Ads account can make up 15-25% of overall volume in a paid search effort, so it’s nothing to scoff at. While Bing Ads lacks the same feature-rich design of AdWords, it does offer a very capable on- and offline editor. With a little elbow grease and offline analysis, AdWords bidding practices can be incorporated into Bing Ads.

With no proprietary bidding technology, we cannot rely on an algorithmic conversion optimizer to set bids for us in Bing. Fortunately, the very same tool we use to identify goal CPAs can be modified to calculate goal CPCs.

Consider the following known formulas:

• CPA = Cost / Conversion
• CPC = Cost / Click
• Conversion = Conv Rate x Clicks

Doing some quick substitution:

• CPA = Cost / (Clicks x Conv Rate)
• CPA x Clicks x Conv Rate = Cost
• Cost / Clicks = CPA x Conv Rate
• CPC = CPA x Conv Rate

So, we can take calculated goal CPA (net present value AOV) multiplied by conversion rate to identify a goal CPC at a keyword level.

Concluding Thoughts

Third-party software solutions exist to fill in the gaps where proprietary software falls short. For many years now, tools like Marin Software, Kenshoo, Adobe Omniture, and countless others have delivered enterprise-level tool sets designed to make up for shortcomings in the AdWords and Google Analytics tracking and management systems.

But, with the bevy of recent updates and changes, these platforms are struggling to maintain the edge that has made the search technology industry relevant. Keeping up with AdWords/GA has become the primary objective. Google is laying down the challenge to yet another industry: add value or move on.

That said, little has been divulged about how your analytics will be impacted by the enhanced rollout. So, I’d like to share a few thoughts on the subtle changes which, if not taken into account, may muddy the waters of reporting you’ve worked so hard to clear up.

The Dividing Line

The impact of enhanced campaigns on your analytics depends on which platform you use and the nature of the integration. This brings up a very clear distinction: you’re either using Google Analytics (GA) with a linked AdWords account or you’re not.

If you are using a linked GA account, you’re in luck – nothing changes in terms of data acquisition. Your analytics will work just fine without any major impact outside of campaign structural changes you’ll need to make to comply with the enhanced campaigns.

This is because linking GA and AdWords accounts requires using auto-tagged URLs. Auto-tagged URLs include a dynamic parameter, gclid, which is appended to all AdWords URLs and passes all available data between the two systems.

The Other Options

But, what if you’re not using GA with auto-tagged URLs? Now is a good time to either consider adopting GA and auto-tagging or start the process of updating campaigns and URL structure to adhere to the limitations of enhanced campaigns.

If you do fall into this category, you’re either using an unlinked GA account and UTM parameters for other internal tracking purposes or some other parameters specific to your analytics package.

First and foremost, if you have an unlinked GA account and are using UTM parameters for internal tracking purposes, it is possible to leverage auto-tagging and other custom parameters to leverage GA as well as another tracking solution.

The incremental data imported into GA with a linked account is invaluable:

• Match Type (Broad, Phrase, Exact)
• Ad Group (The Ad Group associated with the keyword/creative and click)
• Destination URL (AdWords Destination URL)
• Matched Search Query
• Ad Format (text, display, video)
• Ad Distribution Network (Google Search)
• Content-Targeting Option (Automatic Placements vs. Manual)
• Placement Domain (the domain on the content network where your ads were displayed)

Source

In order to understand what needs to be changed, we need to understand the nature of trackable changes in enhanced campaigns and further leverage ValueTrack parameters to ensure our analytics effectively capture all relevant data.

The only mandatory [trackable] change to campaign structure is the consolidation of devices into a single campaign. Breaking out campaigns by device is not an option, so conveniently labeled campaigns like ‘Campaign1_Tablet’ are a thing of the past.

To demonstrate the impact of the change, a standard tracking URL might look something like this:

www.company.com?source=google&medium=cpc&campaign=campaign1_tablet&ad={creative}&term={keyword}

Now, in order to capture the user’s device, include the {device} parameter so the URL looks something like:

www.company.com?source=google&medium=cpc&campaign=campaign1 &ad={creative}&device={device}&term={keyword}

The {device} parameter will populate values ‘m’, ‘t’, or ‘c’ for mobile, tablet, and computer respectively.

If your site has a noticeable variance in user behavior between device models, via the Google Display Network (GDN) you can leverage the {device type} parameter (ex: Apple+iPhone) to modify URLs and serve up a unique user experience per device.

While this parameter is not currently available for Google Search, I have to believe that it will become available via the search network sooner rather than later.

If you’re dead set on providing a unique user experience when the user is on a mobile device, you can leverage the {ifmobile:[value]} parameter. This conditional parameter only populates if the user is on a mobile device and gives the advertiser an opportunity to modify the URL.

For example, www.company.com?{ifmobile:site=mobile} will render as www.company.com? If the user is on a computer or tablet versus www.company.com?site=mobile if the user is on a mobile device.

It’s worth noting that the forced migration to enhanced campaigns is still months out and, as always, it’s possible that Google makes changes which alter the viability of the tracking statement above.

Concluding Thoughts

Taking a step back, outside of how you collect user device, enhanced campaigns should not change too much in your analytics and reporting. That said, knowing user device is critical and if the available ValueTrack parameters cannot work for your situation, there is a significant add-value to setting up Google Analytics and linking your AdWords account.

If you have been putting off migrating to or learning about enhanced campaigns, you should consider raising the priority level enough to stay ahead of the learning curve, especially if you run a highly integrated, non-GA based system.

The levers that we can push and pull exist within two unique systems — AdWords and Merchant Center — and, as a result, we need to do some internal legwork in order to effectively merge the inputs and outputs and come up with actionable analytics to optimize on.

At my agency, we’ve been working on some different ways to more effectively work with PLAs, given the available analytics and levers for optimization. This article is intended to share some ideas and processes that we’ve found helpful. Hopefully, they help you, too.

Placements Not Products

It all begins with how you think about PLAs. Back in the good ol’ days, PLAs were just products and descriptions. That is no longer the case – the modern incarnation of PLAs are contextually targeted placements. The issue is that the placements are defined by auto-target filters set in AdWords based off products defined in the Merchant Center being contextually matched to queries deemed product-specific by Google.

Historically, it was a best practice to submit all available products to your merchant center feed to get exposure and potentially free clicks. That said, now that clicks aren’t free, you most likely have some products that aren’t ROI positive, and conversely, products which are ROI positive. So, like any other paid channel, we need determine the winners and losers and optimize.

For example, if my CPA threshold is $150, then I’d want to either bid down or pause all auto-targets above $150:

Herein lies the problem with the analytics provided by AdWords / Google Anlaytics and associated management tools in AdWords. What if there’s only one product or, more likely, one query that is driving ups costs on an auto-target?

The logical action is to manage products like keywords: exclude those products from your shopping feed that are not profitable and bid more for highly-profitable products. However, AdWords does not show queries by auto-target, so it would appear impossible to tie the query to the trigger product to the actual product sold.

Fear not. It’s not impossible to make the linkage between query, trigger product, and product sold; but, it is a bit of a pain. Here’s how you can do it:

• Leverage the ValueTrack Parameter {adwords_producttargetid} to capture the trigger product ID in the URL
• Capture the query using the ‘q=’ parameter in the referrer

So, your product URL in Merchant Center would look something like this:

{http://www.company.com?blah=blahblah&parameter={adwords_producttargetid}

Note: I like to leverage the utm_term parameter to grab the product ID within Google Analytics.

And, on your landing page, you can use the HttpUtility.ParseQueryString method (or something similar; just get your tech team to do it) to grab the q= parameter from the referrer.

Assuming you’re not getting an overwhelming amount of (Not Provided) traffic, you should be able to build a table linking query/trigger product to the actual products being sold. Through analysis, you can either identify trends or go one step further and track at the visitor level to accurately attribute revenue to the right query / product.

Armed with this data, you or your marketing colleagues can more effectively build auto-targets isolating profitable products and enabling more precise bid management.

Consumer Behavior & Intent

While PLAs may be triggered like a placement, there is an inherently unique way users interact with PLAs because they are an entry point to comparison shopping. That said, if you’re able to link the query with trigger product you can gain some insight into user intent.

For example, if I search for [running shoes], I’m most likely in the research phase and am not likely to purchase a pair of shoes if I click a PLA on the SERP. That said, advertisers are unable to provide a category level landing page because PLAs are triggered by specific product listings:

Showing product-specific ads for research terms is fairly ridiculous, given the unlikelihood of a resulting conversion. However, by capturing query / trigger product you can identify which products are being triggered by research-driven queries and modify product pages to be more engaging and keep the visitor on your site.

Organize query / product by visits and append a custom field ‘Category’ to filter out specific types of queries such as Generic, Product, and Brand:

Identifying the top generic query traffic generating products, create product page experiences that more closely resemble a category or brand level page through intelligent design of a left nav, breadcrumbs, visible filters, or strategically bringing similar products above the fold.

Analytics Or Marketing?

The truth is that it’s a bit of both. The person in charge of PLAs must have access to marketing tools, analytics, and most likely, SQL to gather all required data. PLAs require a solid understanding of what data can be captured via URLs and referrers, as well how to present that data in a way that is actionable given the optimization tools at hand.

Most marketers are, at some level, analysts, and in most cases, the Web analytics team is heavily involved in the online marketing programs at their business. Getting the two teams to work together will bridge the gaps in intelligence otherwise missing from default versions of Analytics and AdWords.

Connecting Online & Offline Advertising

Drawing connections between online and offline advertising efforts is full of externalities and question marks. Seasonality, sales, budgets, timing, and the lack of tracking in offline mediums combine to form murky waters and inconclusive correlations between offline flights and online marketing results.

In response, we do our best to ensure common messaging to users at the same stage of the buying cycle. Online display or video should be inline with TV as both increase branded awareness reach, while email and direct mail solidify relationships through deals and search results, both paid and organic, should highlight key value propositions learned in the research phase.

Despite all the externalities working against online and offline correlations, at some point, if there’s a strong enough connection between offline advertising and online marketing success, we have to take notice.

How TV Ads Boost Traffic From Search & Other Channels

One such case is the impact of television ads on online marketing channels. An advertiser I’ve been fortunate enough to work with over the years, an online subscription service provider, has intermittently run TV flights over the past few years with great success. TV ads get run and various online channels observe a bump in volume during the flight as shown below.

The graph above shows the relative boost in traffic broken out by channel during a TV flight. During the 10 week reporting timeframe, visitors increased by 15-25% across organic search, paid advertising, and email relative to the rolling average (100%).

We would expect a significant brand awareness push to trickle down to branded paid and organic search, but the reality has extended beyond brand into core non-brand terms as well, as shown below.

There is a logical lag between launching the TV flight and search getting a lift. On the flip side there is also a lag time between the flight ending and the impact on search reverting to a more normal trend.

Most importantly, this is not a one-time occurrence. Over the course of more than a dozen TV flights, we continued to observe core non-brand search terms generate 25-30% incremental visitors while maintaining normal conversion rates, resulting in great performance.

How TV Flights Impact CTR In Paid Search

It’s important to note that in non-brand paid search visitors/clicks increase disproportionately to impressions, or in other words, CTR goes up during the flight, highlighting increased brand awareness during and immediately following TV flights as shown below.

With a naturally increased CTR, we can push for incremental volume through bidding and maintain normal CPCs as we benefit from a Quality Score boost.

So, does this mean you should go out and spend millions on TV spots? Not necessarily… this is one example where it works. Before investing in TV, maybe try display and develop a strategy to align messaging across channels, reaching users at common stages in the sales funnel.

As observed in the example above, an end-to-end brand and direct response strategy can have a cumulative impact on user engagement that’s greater than the sum of its parts.

Experimentation is good, and while tracking is difficult between online and offline advertising efforts, if you can create a simple enough environment by limiting or normalizing for externalities, then the analysis becomes easier to distill into why things happen rather than just what happened.

Before getting into some results and the impact on marketing efficiency, I received several comments in response to my last post surrounding how consumers behave within PLAs. Personally, I do not think of PLAs as ad placements like paid search because, unlike the ads in paid search, shopping results are not clearly labeled as ads.

On a normal Google search engine results page, ads are clearly labeled as ads.

But, within Shopping, there is no clear notification that results are paid placements.

Unless you click on the ‘Why these products?’ link in the upper right, in which case you will see:

If there has been no significant visible transition to users, we can assume that user behavior within Shopping / PLAs has not changed, either. After all, the nature of the product, comparison shopping, remains the same. However, looking at the distribution of traffic and transactions over time for a composite of several long-time PLA/Shopping e-tailers suggests there have been some interesting things going on:

Both tables provide insights, and I’ve highlighted the most significant metrics. Distribution of traffic tells us that the visitor inflection point occurred in July — meaning that August was the first month that Google pushed the majority of traffic into PLAs.

However, distribution of transactions shows that traffic driven by PLAs in Q3 was highly unqualified, resulting in very few relative transactions despite more volume being pushed via PLAs.

The disruptive switch could be due to a multitude of reasons, including extensive testing in the layout or the evolution of auction settings such as bid floors, both of which could have significant ramifications as the market settled into place.

Whatever the cause, it had a massively negative impact on PLA conversion rates relative to Google Shopping in Q3:

That said, by October, things appeared to settle down as PLAs steadily took over nearly 100% of traffic.

So, if we ignore the rocky Q3 and look at the aggregate shift from beginning to end of 2012, general user experience within Google Shopping has not changed too much. So what has changed?

To get some answers I reached back out to my good friend, Erick Barney, VP of Marketing at Motorcycle Superstore (MCSS) to get his thoughts and insights on the last few months of PLAs and their underlying market currents. The following is a summary of our conversation along with some quotes from Erick.

It’s worth noting that MCSS has been in the Google Shopping space for many years and had a highly optimized merchant center feed prior to the PLA switch. Since the completion of the paid click transition, they have observed a ROI via Shopping that exceeds paid search.

Despite all of the product-level details in the merchant center feed not being present in the new placement formats, Erick believes that as a category leader with an optimized feed, MCSS is benefitting from the new auction based model.

“PLAs represent Google’s effort to reduce the clutter in the Shopping experience. By charging for clicks, companies need to have a strategy surrounding product submission, feed optimization, and bid management,” says Barney.

Even with the new associated costs, PLAs have been good to MCSS. “There’s no point in getting hung up on the cost of PLAs. It is what it is. We should be grateful we got all those free clicks for so many years,” says Barney.

Category leaders and price competitive e-tailers stand to benefit the most from the paid shopping experience as consumers purchase either from trustworthy or cheap stores. In response, Google has also recently introduced the Google Trusted Store guarantee, which highlights category leaders and incentivizes users to purchase from trusted retailers.

All said and done, the biggest change in PLAs remains to be the absence of Amazon from the listings. If you search long and hard enough you will find some Amazon listings in the results but only at a fraction of the company’s former visibility.

Just a few months ago, Amazon dominated Google Shopping. They ranked #1 for virtually all results. Now, other brands can take up those top positions. But, is that a good thing? It depends.

“Amazon used to drive huge volume but you don’t own the customer from an Amazon transaction so lifetime value is relatively low. Removing Amazon from Shopping has freed up the top spot on most product searches but has also led to reduced overall conversion volume,” says Barney.

While category leaders have picked up traffic through increased exposure and increased lifetime value through ownership of the customer, they have lost net revenue. As a result, smaller retailers who were not previously subsidized by Amazon stand to benefit the most from increased exposure.

“Amazon faces a very interesting predicament. It was historically easy for them to submit their entire catalog to Google Merchant Center and rest on their laurels as the unrivaled e-commerce leader to generate massive amounts of traffic. Now they need to worry about bidding and margins like the rest of us, which appears to have taken them out of the shopping results,” says Barney.

Maybe it’s temporary, but the results have impacted businesses in a big way.

The Bottom Line

The process of shifting traffic from fee to paid clicks was by no means smooth, but it appears to have normalized leaving opportunity for price-competitive businesses to stake a claim to placements that have previously been reserved for Amazon.

That said, because there is now a cost of engagement, businesses need to be strategic in which products are submitted to the Google merchant center, continue to focus on feed optimization, and consider margins in determining an ROI goal for PLAs.

If you gave PLAs a shot back in Q3 2012, it’s worth going back to see if you may have been victimized by the rocky transition. Compare relative conversion rates and distribution of traffic to get a better picture of the impact of the transition and you might find it’s worthwhile to put together a test budget for Q1 2013.

If one channel looks better than the others you shift budgets or bids accordingly. It’s standard practice but is it the complete picture?

What if all channels aren’t equal? A click from one channel might be worth more than a click from another channel, especially over longer periods of time. At face value this seems logical and if you dig in, it becomes clear that time skews our ability to determine the true value of a click and expected performance over a static period of time.

As a result, we need to factor time into the equation in order to understand short term versus long term channel performance. The results are not only surprising, but may completely change the way you manage performance marketing channels.

For starters, let’s look at a Google Analytics 30 day time lag report for all channels (Available in multi-channel funnels reporting section):

Different business will observe different patterns. I’ve chosen a particular set of data which highlights the latency in conversion to make my point more clear. The graph shows that roughly 46% of conversions occur within the first 7 days. That means 54% of conversions in this analysis occur between day 8 and 30 following the users’ first entry.

Let’s go back to the channel performance report and look at one channel, let’s say PPC, over a 12 week period:

Ad servers (ex AdWords, AdCenter, DoubleClick, etc) generally attribute conversions to the day of click, so over time conversion numbers continue to look better and better. The time lag report makes sense of the rate at which latent conversions come in.

Google Analytics also offers a similar attribution process through multi-channel funnels. It’s worth noting that default Google Analytics reporting leverages day of conversion attribution – by that I mean that revenue is associated with the day that the purchase was made.

But via multi-channel reports, we can create a conversion segment to display revenue attributed to the first interaction, in this case, Paid Advertising.

Hopefully you see where I’m going with this by now. A one-week snapshot of performance by channel attributing to first interaction or day of click, while ideal for optimization purposes, will always make things look worse that they really are in the short term.

It’s easy to use day of conversion attribution reporting to make things look better but you’re most likely not taking into account the true effectiveness of marketing dollars spent due to latency of conversions.

For example, if you push for more volume this week and report with day of conversions attribution then you will most likely show relatively poor performance as conversions volume recorded this week is commensurate with marketing dollars spent in previous weeks.

Being patient and waiting out the cookie window might just show that what was initially perceived as poor performance was in fact a highly productive and efficient.

By running channel specific time lag analyses, we can build models to predict the latent efficiency on weekly data. Incorporating these models / multipliers into reporting we can effectively manage expectations of bosses or clients so they see the benefits of pushing for more volume rather than rushing to judgment during the sensitive period of time pushing for volume and reaping the rewards of additional conversions.

Hindsight is always 20/20. Forcing marketing data that operates with latency into a one week view will never tell the whole story. Leverage existing data to build a similar model and educate your team on your company’s normal conversion behavior / time lag to garner support to grow your marketing channel.

The byproduct of conservative projections is more difficult forecasting mid quarter as it becomes apparent that not all referring traffic sources are equal in traffic and value.

The inevitable question comes up: how much more revenue can be delivered from channel X with additional budget?

To accurately answer this question requires several channel specific insights including:

1. The relationship between visitors and cost-per-visitor
2. Incremental visitor opportunity
3. Conversion rates
4. Average order values
5. Latency in conversion

Forecasting With Google Analytics

Analysis requires multiple regression models to understand marginal cost and revenue for each channel. But before you get neck deep in spreadsheets, I want to highlight a few tools at your disposal, which can guide you in the direction of specific channels and subsets of those channels to pursue for the lowest hanging fruit.

Google Analytics has a fantastic tool set called Multi Channel Funnels. The purpose of the tool is to showcase the overlap of channels leading to conversions through data tables and visualizations:

By default, Multi Channel Funnels break traffic sources into basic channel groupings, as well as the more standard Source/Medium groupings, and identify where cross-channel interactions occur including the following analyses:

• Assist conversions – interactions which lead to additional interactions which convert
• First interactions – attribution to the channel driving the first interaction prior to a conversion
• Interaction paths – cross channel paths by which users reach a site repeatedly prior to converting
• Path Length – the number of interactions prior to conversion
• Time Lag  – the time line from first interaction until conversion

A quick look at a top conversion paths analysis will immediately highlight opportunities for incremental volume:

Numbers 8 and 10 showcase immediate opportunities – paid search and ‘other advertising’ lead to latent direct conversions. This means that pushing on paid ads in the short term will lead to direct conversions in the longer term.

In aggregate, the data is insightful, highlighting how people are getting to your site over time. However, the Conversion Segments feature takes basic Multi Channel Funnels from presentation material to action ready.

By creating a custom conversion segment, Paid Search to Direct, we can get a clear picture as to how many different ways users interact with Paid Search and then additional channels prior to purchasing, as well as the exact percentage of conversions and revenue associated with this path (I am not displaying conversion count and revenue, but they are available).

Switching gears over to the assisted conversions report (keeping the conversion segment Paid to Direct applied) we can switch the primary dimension over to use AdWords Matched Search Query to identify paid search keywords driving clicks which turn into conversions via direct.

Finally, moving to the time lag analysis, we can understand how long it will take for incremental paid search clicks to deliver results via direct clicks:

Taking a step back, this series of analyses reveal:

1. The exact set of keywords to push on in paid search to yield conversions via a direct last interaction
2. How long it’s going to take to observe those conversions

Armed with what keywords to analyze, go back to and execute the aforementioned regression analysis to understand the projected cost of the incremental paid search clicks. Merge it with the time lag analysis to showcase the expected time frame for incremental conversions. Now, that’s an accurate forecast.

The problem with Google Analytics, or any analytics package for that matter, is that even if my site is properly tagged and I’ve developed a systematic inbound link tagging notation, it’s almost certain that my Source/Medium/Referral URL reports will still include inconsistencies that need to be resolved to provide accurate reporting:

Software doesn’t know what isn’t logically defined. Google does a decent job of auto-tagging referring traffic and categorizing into Source/Medium buckets which are logical.

However, a bunch of stuff slips through the cracks into the catch-all “Referral” bucket. Over time, the bucket grows big enough to constitute a very significant percentage of overall traffic and needs to be addressed, cleaned up, and recategorized in order to provide accurate Source/Medium reporting.

A classic example in the table above is bingiton.com. It’s unclear to Google Analytics as to what the site is, so it’s tagged as a referral.

However, by visiting the site, we can easily see that it’s a site comparing Google and Bing organic results and as a result, the source can either be Bing or Google (although the goal of the site is to show Bing organic results are better than Google’s), and Medium should be Organic, as the results are clearly coming from a search engine, as shown below.

searching [analytics] on bingiton.com

The good news is that there is a solution; and furthermore, it’s not too difficult to put in place. At my company, we call it Clean Source Tagging. The idea is that there’s a raw source and a clean source. We take what GA gives us (the raw source), manually review it for accuracy, and compile an Excel database of all traffic sources and a reviewed and approved list of true clean sources.

If you don’t recognize a source, visit the site to determine what it actually is. We can even get a little fancy with the output and include primary source, secondary source, and medium to gather all relevant information:

We then take the standard GA Source/Medium output and run a simple vlookup against the clean source table to associate each referral with the correct source/medium.

If it sounds like a lot of work, it’s because it is. But it’s a one-time project with a small ongoing component to keep the database fresh and up to date. The bulk of referring URLs are static week-over-week.

As a result, the initial project may take a while; but if you keep the database handy, run the vlookup and address the new unknown referrals, it’s a quick job even an intern can handle in under an hour per week.

So, now that we’ve got a clean source table, what’s next? Dashboards. The clean source table gives us the ability to create the simple, intuitive and insightful dashboards executives crave:

With a little extra work up front, we can create a beautiful Excel template, drop new data in, update the clean source table, and deliver source performance reports in no time. What was previously an irrelevant and unusable report in GA, or a mountain of work to get right, becomes a near automated solution that is a staple of weekly reporting.

Now you can hit snooze one more time on Monday morning. You earned it.

Beyond the obvious financial impact on businesses caused by having to pay for something that was once free, are Product Level Ads (PLAs) really a bad thing? Is Google getting greedy? Or are they just taking another step in their eternal quest to improve user experience?

To get the low down, I called on an old friend, Erick Barney, VP of Marketing at Motorcycle Superstore, and asked him how PLAs were impacting his business. The following is a recap and analysis of our conversation.

The impact of PLAs and paid shopping clicks extends far beyond the simple explanation of paying for clicks. At a high level, there are four fundamental factors that are working together to produce a new Google Shopping environment for etailers:

1. Paid clicks from PLAs are increasing
2. Free clicks from Google Shopping are decreasing
3. Amazon sales are declining
4. PLAs are cannibalizing paid search

Each of these four factors must be considered in order to understand the overall impact of PLAs on the business.

Paying For Clicks You Previously Didn’t Have To Pay For Stinks

Businesses are now faced with a decision to allocate budgets to PLAs and Shopping – sources that have been historically free. Compound that with a difficult ad format to manage and it’s a recipe for frustrated marketers.

Even a well-optimized Google Merchant Center Feed provides little control over the product level messaging displayed on Google and bid management is limited to auto-targets per product filter in AdWords.

It’s noteworthy that Google appears to be testing pretty heavily during the rollout of paid PLAs and it shows in the analytics. There are no year over year comps so we’re limited to week over week and the volume is totally erratic – there is no week over week consistency other than an overall trend of volume increasing over time.

Verdict: Bad for advertisers.

Free Clicks Are Going Away

While paid clicks are on the rise, free clicks continue to decline in volume. Fewer free transactional clicks mean rising ad cost and even more frustrated marketers. Year over year comps show a dramatic drop-off in free Shopping traffic, not surprisingly roughly the same volume as the rise in paid traffic.

Verdict: Bad for advertisers.

The Amazon In The Room

The business most negatively impacted by PLAs is Amazon because Google has drastically reduced their presence in Shopping – a channel that must have been one of their leading referral sources because they were listed #1 for virtually every product level search.

If you’re not familiar with the Amazon affiliate model, it’s a percent revenue share (in the neighborhood of 10-20%) and Amazon gets to keep the customer. This last part is probably the most significant factor because the lifetime value multiplier for a customer you don’t own is one.

It’s worth noting that Erick and his team are savvy feed optimizers and were pushing hard on Amazon prior to the PLA transition. Ever since the transition year over year comps are falling off a cliff. However, with the absence of Amazon, there’s a significant opportunity to gain market share in Shopping…the downside being that you have to pay for it.

Yet, I still see the removal of Amazon from PLAs as a net positive change for the average etailer because the fallout gives the business an opportunity to manage the client relationship to long-term success and a higher lifetime value.

Verdict: Good for advertisers.

PLAs & Search

The relationship between PLAs and search is unique. I’ve always been a fan of product ads in search and I believe the format of PLAs is ideal for product level queries in search:

Because of the high relevance of PLAs for product searches, PLAs have begun to cannibalize search volume. This is neither a good or bad thing, so long as the product being displayed in PLAs are highly relevant.

A business with well optimized feeds will be rewarded by PLAs in search. This has been one of the better integrations of the technology and highlights everything right Google has done in this ad format.

That said, when the query gets more generic, Google shows a fundamental lack of understanding in user intent:

In this search the user shows no brand, vendor, or style preference yet Google has opted to show a specific tire rather than an informational ad offering insight and the opportunity to learn about motorcycle tires. Hopefully the Googlers will figure this one out and remove PLAs from non-product level searches.

Verdict: Null.

Adding Up The Damage

To be fair, the PLA transition is still in progress so jumping to conclusions is premature. That said, the net sum of PLAs is not positive on the business. Erick said it best: “PLAs present a highly disruptive change to a previously stable environment. We’re at Google’s mercy while things shake out and we’re not going to know the full impact until October.”

Check back at the end of October when I’ll be writing a follow up where we dig into the numbers at the conclusion of the transition.