Campaign management tools and analytics provide fundamentally different information, and while directionally similar in many cases, choosing the wrong data source can lead to inconsistent perception of success metrics between the executive level and those tasked with execution.

Understanding the difference between analytics and ad server data may be the only way to avoid making decisions that can potentially hurt your business in the long run.

Before getting too far into this, there are two unique types of reporting systems, analytics and campaign management tools, with a fundamental difference – conversion attribution.

An analytics system leverages day of conversion attribution; that is conversions / events that happen on a given day are associated with that day regardless of when the click occurred [which led to that conversion].

On the flip side, a campaign management system uses day of click attribution – conversions / events are associated with the day the click occurred which led to a conversion regardless of when the conversion occurred.

Note: some campaign management / analytics systems offer both types of attribution. These tools are versatile but cost money.

The Differences Among Reporting Systems

Different reporting systems each have their purpose. Campaign management systems (e.g. AdWords and AdCenter) offered by ad servers to manage campaigns may be dressed up with fancy graphs and charts to look like analytics systems; but if you look beyond the bells and whistles, they are still just management systems.

With a bevvy of reporting systems available, it’s no surprise that choosing which system to trust for internal reporting is a difficult task. Why not use the ad serving tools to determine performance?

In order to answer this question, we need to evaluate pros and cons of each attribution model. There is a place for both day of click and day of conversion attribution because they each provide unique, different insights in to the performance of your marketing efforts.

Day of click attribution is useful for managing the value of a click (bid management). In order to understand the value of a click, you need to know the number of conversions / events driven by common clicks (e.g. clicks driven by a given keyword, query, or demographic target).

Moreover, in order to understand hours of the day which drive more valuable clicks (for use in processes such as day parting) you need to be able to associate the time of the click with the conversion. What time a user converts is not relevant to the value of a click because we’re only concerned with whether or not a user executed a conversion after the click.

While marketers manage their campaigns, the finance team is measuring the financial flux of the business – in other words, the flow of money in and out of the company.

When reporting on what happened in a given time frame, a business operates on an accrual basis – money spent and revenue received in a given week is associated with that week – hence day of conversion attribution is useful for reporting. The CEO / CFO needs to know how much the company made and spent for a given date range. Without day of conversion attribution, this would not be possible.

The issue with day of click attribution is that when used for reporting purposes it can lead to well intended yet short-sighted decisions. Imagine a situation where you’re analyzing recent performance (let’s define ‘recent performance’ as clicks in a live cookie window; a widely accepted cookie window is 30 days).

If you aren’t looking back over the full cookie window, then you are most likely underreporting performance: clicks made a few weeks ago may have driven conversions that occurred yesterday. In the day of click attribution model, they would be associated with two weeks ago and would not be visible in your report for yesterday.

Compensating for the ongoing historical attribution requires a complete cookie cycle view of performance data sets as the accrual of conversions occurs. For this reason, we prepare daily rolling reports (a day over day snapshot of performance) for all performance monitoring to understand trends and expected performance on a single day’s events.

The window of data analyzed should account for the vast majority of conversions as observed in the average time for click to conversion, if not the entire observed cookie window up to a reasonable period of 14 – 30 days.

To understand relative performance trends, consider the case where we observe 66% of converting events on the same day as the click and the remaining 33% of spread out over the subsequent 29 days of the cookie window.

If your goal is a $5 CPA, a one day CPA of $7.50 is acceptable ($5 / 66%) and we would expect the CPA for that one day to make it’s way down to $5 over the course of the 30 day window (when using day of click conversion attribution).

If you forced yourself to achieve a $5 CPA on one-day performance then you will end up with a 30 day CPA of $3.33 ($5 * 0.66%) meaning that you left opportunity on the table.

Or breaking it down by day:

Another major advantage in analytics systems is de-duplication across channels. Online marketing is not a single channel ecosystem but we have to attribute conversions somewhere. Furthermore, the waters get even muddier when considering display attribution rather than search because advertisers will not only get click-based conversions but also view-based conversions via display.

Analytics will attribute a conversion to the last click-based referrer and as a result, a view based conversion will never be attributed to the display channel because, by definition, the user never clicks on the ad.

In order to accurately determine the viability of view based conversions, advertisers need to match up transactions between the ad server and analytics to see if the display channel was the only paid channel to reach the user or if there were other channels involved later in the funnel.

As analysts and marketers it’s our job to manage up, educating executives on the nuances of data latency and why one system is right for determining the value of a marketing channel while another is right for monitoring performance.

It’s rare that the C-suite wants to get into the weeds of campaign management; rather they want to provide accurate target metrics for their team(s) to execute on. Education about the difference in data within analytics versus management tools will only help the C-suite better understand and set those goals with more accuracy. Proper education should make it very clear why analytics is the right tool for reporting.

There are certain things which simply cannot be done on a mobile device – but the gap is quickly closing and tablets are at the center of attention as consumers are embracing mobile tablet PCs as their primary computing devices.

With “more normal” sized screens, tablets don’t have the same issues with site formatting that high-end mobile devices have long suffered from. As a result, we would expect users to have an easier time engaging with a website via a tablet.

In terms of market share, tablet PCs are being gobbled up by consumers (shipments are up 42% Q/Q and 210% Y/Y) so it’s about time you and your business take a serious look at how tablet users are engaging with your site and develop a strategy for reaching tablet users with your marketing program.

Fortunately, recent targeting modifications in AdWords allow marketers to isolate tablet traffic and even target ads by operating system. It’s only a matter of time for other marketing channels to step up their game and provide the same level of targeting. (Hello Facebook – maybe this is why they’re bringing ads to mobile operating systems…)

With the ability to parse data comes an opportunity for a test. Tagging URLs by device (I usually append the device to the campaign name and populate it into the utm_campaign parameter in GA) makes for a clean data set to determine the differences in user behaviors by device.

Comparing Desktop, Mobile & Tablet Performance

Below, I’ve normalized some key performance indicators for mobile and tablets relative to desktop. The terms being analyzed are generic category terms for an online-only service provider.

It’s worth noting that the company has a mobile and desktop version of the site but does not have a tablet specific site or app. Sniffer scripts are used to automatically send mobile users to the mobile site. Tablet users are sent to the desktop version of the site.

Albeit simple, the data set is quite revealing.

The most significant variance by device comes from impression volume. There just isn’t as much volume on tablets relative to mobile searches and more significantly desktop searches.

Even though relative click-through rates for mobile are higher than tablet users, tablet users show more intent to engage at a deeper level convert ata significantly higher rate than both mobile and desktop. Even more interesting is that the average order value for tablet users more closely resembles desktop than mobile search.

This is probably the most insightful metric on the table as it identifies the inherently unique, and impulsive, nature of mobile search behavior relative to desktop and tablet. Maybe mobile behaviors are different due to browser security concerns and the fear of credit card security (maybe forthcoming technologies like Google Wallet will fundamentally change conversion rates on mobile devices).

With such a massive discrepancy in average order value, it’s worth considering making a unique, cheaper, offering for mobile visitors to your site in an attempt to get them to stick around, engage, and convert.

Look At User Engagement By Device In Your Analytics

Diving deeper into the post-click analytics, we can better observe what happens between the click and conversion via on site metrics such as pageviews per visit and bounce rate:

With respect to PPV and Bounce Rate, once again, mobile shows poor performance relative to desktops with tablets in the middle, skewing closer to desktop behavior. Mobile users are less likely to stick on a site, moving faster and bouncing immediately more often than not, whereas users are more willing to engage with and seek out interesting content on a larger format device.

However, Time on Site shows some very interesting behavior – when a mobile or tablet user is engaged with a site, they spend far more time going through it.

There are many possible explanations for the dramatic variance in time on site, but I believe it comes down to when users are engaging on a given device. Tablets and mobile devices are more commonly associated with recreation while a desktop / laptop is more associated with work.

Personally, I use my computer for work, tablet for play, and when I’m commuting I use my phone. During a commute or while I’m sitting at home on the couch, I’m far more likely to spend a few extra minutes fumbling through a site.

If I’m surfing during the day, I’m usually pretty efficient with my time and less likely to spend too much time on a given site (except Search Engine Land, of course); and while I’m not a massive sample size, I don’t think it’s a far cry from the norm.

Conclusion

The numbers indicate that tablets do bridge the gap between desktop and mobile behavior patterns. Similar to mobile, tablet users search for a purpose and are more likely to click through an ad.

But unlike mobile, tablet users are not necessarily mobile and are more likely to engage with a site after clicking, displaying stronger engagement metrics and overall conversion behavior closer to desktop users.

Depending on your site’s method of monetization, the value of a click on a tablet may outweigh the value of a click form any other device.

While I’m not a huge fan of sampling data when not necessary, larger data sets put a significant load on servers and sampling becomes a necessary evil when trying to deliver quick high on high volume data sets. As a result, I’m a fan of how the GA team has integrated data sampling into reporting.

On the custom reporting tab there is a new button resembling a checkerboard. Below the button is the sample size.

To adjust the sample size, click on the checkerboard button to populate a sliding scale going from “Faster Processing” on the left to “Higher Precision” on the right.

Faster processing uses a smaller sample size, delivering results more quickly. Higher precision uses a larger sample size for more accurate reporting.

As with any data sampling process, the smaller the sample size the greater the margin of error due to the assumption that the sub-set of data reflects the trends of the aggregate data set.

This is a significant change in how we can read into data sets as 1) it gives analysts a mechanism for more real time insights as the data aggregation takes several hours before being made available in the interface; and 2) the integrity of data can be put into question due to statistically insignificant sample sizes.

When Does Sampling Occur?

As noted in the “Learn More” link under the sliding scale, sampling automatically occurs “By default (until you use the slider to change your sampling preference)…when report data exceeds 250,000 visits. However, you can use the slider to increase this threshold to as high as 500,000 visits…Sampling in Multi-Channel Funnel reports automatically occurs when the data includes more than 1 million conversion paths, regardless of your sampling preference setting.”

How Will This Affect My Data?

In short, you will lose clarity and there is the potential for misleading insight if the sample size is too small. If you have an account and run reports with more than 500,000 visits, your data sets will be truncated and assumptions made be made.

Is this cause for alarm? Yes and no. So long as Google delivers statistically significant data samples, then there is no cause for alarm. However, this feature is new and it’s unclear how statistically significant the data samples are. If you require no sampling you’ll need to run reports with less than 500,000 visits.

As noted above, one exception to the 500,000 visit limit is in multi-channel funnel reporting. Multi-channel funnels will be sampled once the number of paths exceed 1,000,000.

To be frank, 1,000,000 conversion paths is a lot of conversions and there aren’t that many companies out there who are pulling in more than 1,000,000 conversions in a relevant time frame (given seasonality; if you are one of these companies, I suggest the paid version of Google Analytics or another solution to mitigate the issue).

As a result, I don’t expect the relevancy of multi-channel conversion funnel reporting to be impacted by data sampling.

If anyone has already analyzed the effects of data sampling on the relevance of their reporting please feel free to comment below. I will be following up once I have enough relevant data to share.

Generally speaking, we use conversion funnels to show the value of early consumer touch points such as display ads, Facebook ads, or generic paid search.

So, for example, even if clicks from Facebook don’t lead to many conversions, we can determine that there is some value in guaranteeing that a user reaching your site fits a particular profile ad guaranteed by Facebook’s incredible demographic targeting.

Early touch points and clicks are often called ‘upper funnel’ clicks. A click not leading directly to a conversion is considered upper funnel if it leads to conversions after the user re-engages with the site via another channel or keyword.

By definition, an analytics system will not register a click in a conversion funnel if the user does not eventually convert. And by all logical optimization best practices, keywords and ads driving clicks not included in the funnels or directly associated with a conversion get bid down and/or removed from the marketing mix due to inefficiency.

But what if this best practice is leaving a significant percentage of potential conversions on the table? By depending on traditional conversions to measure user qualification we’re putting blinders on – grossly limiting the number of potentially qualified users we can identify.

More often than not, there is an action a user commits on a site prior to converting – for example in the retail environment, a user places an item in a shopping cart prior to purchasing it – which is not tracked as a traditional conversion by most marketing teams as a conversion event.

This data (users committing actions on a site but not converting) is the foundation for remarketing – which has been widely embraced as a strong performing channel…so why don’t other channels with more precise bid management capabilities leverage the same upper funnel qualification to assume potential value?

It just takes one extra step to understand the value of such a click and incorporate it into bid management practices.

To make this happen:

1. Define a new conversion for the upper funnel event (ex: entering shopping cart)
2. Define a conversion funnel for the shopping cart

Conversion funnels identify drop-off rate between sequential screens.

While they do not provide insight into the sources, keywords, or users progressing through the conversion process, they do provide a normalized conversion rate between the start and finish of a goal or action such as maneuvering through the shopping cart and making a purchase.

Identifying the conversion rate of a shopping cart

With a conversion indicator for the upper funnel action and a CPA – for the purpose of this example let’s say the CPA is $10 – and a 30.31% conversion rate (as shown above) between shopping carts and purchases, the theoretical cost per order is: $10 / 30.31% or $32.99.

Take is one step further by calculating a theoretical Return on Ad Spend (ROAS) by associating a channel/category average order value (AOV) and then divided by the calculated theoretical CPA. So if the AOV for similar keywords is $100 then the theoretical ROAS is $100 / $32.99 or 3.03.

The process is straight forward – all it takes is a little setup and all of the sudden we are associating potential revenue with non-converting events, providing insight into revenue forecasted to arrive as users move downstream.

By establishing a conversion event in a tool like AdWords we can further leverage conversion funnels to identify even more upper funnel terms, or multi-channel conversion funnels to identify those early touch points which driving users to the shopping cart (for which we now can associate revenue).

Looking at one marketing channel at a time provides incomplete insight into performance yet, ironically, many businesses split management of channels across multiple teams and incentivize marketing mangers on performance leading to internal competition for the same conversions.

For example, let’s consider a Google paid search account over the course of Q4 2011. AdWords gives us the basics key performance indicators:

Because AdWords (or any other channel using a proprietary cookie/pixel based tracking system) takes credit for all conversions that were touched by a Google paid search click at some point in the conversion funnel, conversion values reported in AdWords always skew high.

If we compare the same date set within our analytics platform, you will see dramatically different results:

In this case, the conversion count in analytics is significantly lower (about 33%) – enough to make the channel appear unprofitable. This is due to a last entry attribution model that is used by default in most analytics platforms.

So which one is right? It depends on how we want to interpret the data…

What distinguishes an analytics platform from is that it monitors multiple referring channels whereas a channel specific platform (eg AdWords) operates in it’s own little world without any insight or concern for other marketing channels with respect to conversion counting.

This often leads to double counting of conversions if we were to add up conversion counts across all paid traffic channels (AdWords, AdCenter, Facebook ads, etc.).

As a result, most analytics system offer some form of a multi-channel attribution system or conversion funnel. Whether it’s a slick visual graph like in Google Analytics (below) or a simple linear attribution report as provided by Omniture, the idea is to identify the true value of a click or a visit by parsing the value of each conversion across all touch points leading up to the conversion.

In order to identify the true value of a click from a given marketing channel, we need to gather some channel specific data:

1. Clicks
2. Conversions
3. Average Order Value
4. Number of Channels per conversion

Note: I prefer to use an average order rather than revenue in order to normalize the potential effect of a large order skewing data.

Clicks and average order value can come directly from analytics or the channel tool. Conversion count comes from isolating the number of conversions attributed to the channel through the dividing the number of conversions by the number of unique channels associated with each conversion.

To get these numbers, we need to use the multi-channel conversion funnel report and a few different attribution models:

1. First Click
2. Last Click
3. Linear

By multiplying AOV and conversion count (for the two high and low attribution models with respect to conversion count) and then dividing by the number of clicks we can get a range for revenue per click.

Consider the following conversion count per attribution model:

• Last Click: 2,975
• First Click: 2,399
• Linear: 2,203

The high is last click (2,975 attributed conversions) and the low is Linear (2,203 attributed conversions).  Running the calculations:

(2,975 conversions x $25) / 27,348 clicks = $2.01 / click

(2,203 conversions x $25) / 27,348 clicks = $2.72 / click

This tells us that we can afford an average CPC between $2.01 – $2.72 for Google paid search to run at break-even profitability.

The value of a click is completely dependent on the average order value. If your business has significant seasonal fluctuation in terms of search volume, competition, and/or average order values, I strongly recommend running through this exercise multiple times per year in order to best understand the current max bids to operate a profitable marketing campaign.

The dilemma is simple conceptually but digging into it brings up some fundamental flaws in conversion modeling and highlights the underlying catch-22 that marketers face. It all starts with a simple question: Should you optimize a specific marketing channel based off analytics data or proprietary channel data?

To answer the dilemma, let’s set the stage of why the dilemma exists in the first place: it’s impossible to avoid using multiple conversion tracking systems.

Moreover, it’s borderline impossible to have different tracking systems show the same performance metrics for one channel (ex: AdWords and Google Analytics will show different daily conversion counts for the same campaign; or 3rd party bid management systems will show different conversion counts than AdWords for the same campaign). This is because each tracking system collects proprietary conversion data and uses cookies to determine if their ads drive conversions.

One of the core competencies of an analytics system is to resolve multiple cookie issues and distill a complex conversion path into a logical decision of which channel gets credit for a given conversion (attribution modeling; ex: first click, last click, linear, reverse decay, etc). But the flaw lies in the purpose of an analytics system (reporting) versus the purpose of a bid management system (optimization).

Here’s a classic example: Google Analytics utilizes a 180 day cookie window (actions are associated with a given channel up to 180 days following their visit via the marketing channel) assigning conversions to the day the conversion occurred. On the other hand, AdWords utilizes as 30 day cookie window assigning conversions to the day the click occurred.

Now we get to the conundrum: I fundamentally agree with the logic behind both AdWords honoring the day of the click and Google Analytics honoring the day of the conversion. They’re both right – in order to understand the value of a click, you need to link click costs with associated revenue and a true reporting system should display total revenue captured in a given day.

That said, by assigning revenue to the day of the click for optimization purposes makes it impossible to determine how much revenue a specific channel drives any given day.

Solving The Dilemma – The Best Of Both Worlds

Several third party bid management solutions offer the ability to integrate Google Analytics or other 3rd party analytics data into their software – allowing users to manage bids based off revenue captured in analytics. So long as the bid management solution is able to integrate at the click level (via unique IDs per visit), it becomes possible for the system to assign analytics captured revenue back to the click that drove the revenue.

Because AdWords (and every other proprietary marketing tool such as AdCenter, Facebook Ads, etc) uses a proprietary conversion tracking system, each system will take credit for a conversion regardless if the click was the first in a cross-channel funnel, last, or somewhere in the middle.

So if you total up conversion across all proprietary marketing tools, your total will be much higher than the numbers in Analytics. Using an analytics system is the only way to ensure conversions aren’t being double counted and optimization efforts reflect true conversion data/revenue.

If using a bid management tool isn’t in the cards for you or your company, there are ways to take the reporting flaws into account, minimizing the impact on optimizations using proprietary tools.

Start by calculating the average daily delta between your analytics system and the tool. I recommend using several weeks worth of data, ideally a full 30 days:

Calculate the average daily delta by channel. Now do this for several different 30 day periods and average those numbers to derive a normalized multiplier. This is the multiplier to determine how goal CPA or ROAS should be adjusted.

For example, in the diagram above, the normalized delta is 9%. So by adjusting conversion metrics (divide CPA by 0.91 or multiply ROAS by 0.91) you can more accurately optimize based off de-duplicated analytics data.

Concluding Thoughts

In an ideal world, there is a free tool with a universal cookie allowing users to report and optimize on true/de-duplicated conversion data. The reality is that no such free tool exists and third party [not free] tools are imperfect and/or expensive.

No matter how you choose to cope with de-duplication and optimization, it’s a concept which must be addressed in order to provide accurate insight into marketing channel performance.

Some, such as AdWords and AdCenter, take performance management a bit further and allow advertisers to return a revenue value per conversion to report Return On Ad Spend (ROAS). But these systems stop here; failing to allow advertisers to optimize based on true, billed, revenue rather than booked revenue.

Booked orders are the gross – total orders captured via a standard online purchase process. Billed orders are the net – orders which are actually shipped and as a result, are a more accurate reflection of the revenue recognized by an advertising campaign.

Booked revenue rarely equals billed revenue because of several potential situations including but not limited to:

1. Credit Card Rejection
2. Order Returns / Exchanges
3. Order changes via phone
4. Cancellations

There is no proprietary campaign management system that allows advertisers to adjust order values retroactively. As a result, if you want to look at billed revenue for optimization purposes, it takes crafty reporting, well-tagged URLs for clean analytics, and a very dedicated analyst capable to merge data across multiple systems.

There are third party campaign bid management and reporting systems offering solutions for advertisers to import analytics via FTP and are allow optimization and reporting using billed revenue rather than booked revenue. That said, you don’t need the extra software/expense to understand and optimize off of billed revenue metrics.

Any decent analytics system allows an order ID to be associated with a purchase event. If the analytics system is properly tagging inbound traffic, each order ID is linked with the appropriate channel, campaign, keyword or placement, etc. Through order ID level reporting and reconciliation against adjusted order values, an analyst can easily update all metrics to more accurately reflect billed revenue.

The only real downside to this process is time and reactivity. Depending on the ease of the reconciliation it can take up to a day to execute all the required reporting and put the output in an actionable format which can be imported into AdWords, AdCenter, Facebook, etc to adjust bids.

To ensure data is statistically significant and previous bid changes into account, bid adjustments in a system like AdWords or AdCenter should only be made once every few days and must be done manually. But if a completely manual reconciliation and bid management extract doesn’t get you excited and you would prefer to use automated bid strategies available in proprietary systems, read on.

More interesting than cleaning up reporting is identifying deeper opportunities within new data sets. Truing up revenue is a time intensive endeavor but the net result can be leveraged to quickly adjust more real-time metrics. The relationship between booked and billed revenue allows an analyst to adjust booked revenue goals, effectively predicting billed revenue based on historical trends.

With enough time, a normalized trend line forms; defining forecasted billed revenue as a percentage of booked revenue. While not perfect, you can use this predictive analysis to get more reactive and manage campaign/channels using adjusted revenue goals to more accurately optimize to profit margins.

The same analysis can be done for each channel to identify appropriate multipliers and improve reactivity and predictions of true return on ad spend.

calculating billed revenue as a percentage of booked revenue

Operating without knowledge of billed return on ad spend creates a risk of over-reporting on marketing channel performance.

If your only insight into channel performance is through basic analytics or information provided directly from proprietary tools, you could be assigning too much revenue to various revenue streams and mismanaging leading to non-profitable campaign as a result of prior performance assumptions.

Why operate under assumptions when you don’t have to?

It’s inevitable after seeing other businesses participate on Living Social or Google Deals to think about giving it a shot, but before joining the social ad phenomenon, I’d like to offer some insight and analysis behind this new movement.

Like any other business, the Groupons and Gilts of the world have a business model: profit sharing — a 50/50 split with the advertiser to monetize their services. So if you sell $10,000 of goods or services via Groupon, the cost of the ad placement is $5,000.

Given the mainstream acceptance and exposure to “daily deals” and flash sales, consumers expect a bargain and will not purchase a “deal” if it’s not a deal.

So you have to take a few things into account before you can understand the financial impact and potential benefit of a social ad run on your business: cost of placement, actual revenue, margin and opportunity cost.

Let’s looks at some itemized sample data to give the profits and losses some color. Consider a business offering a $100 product for $50 on a flash sale site. Let’s also assume that the cost of production for this product is $25 and the company sells 200 items during the ad run.

• Sales: 200
• Average Order Value: $50
• Revenue: $10,000
• Profit Share: 50% x $10,000 = $5,000

So the advertiser and social ad company each get $5,000 for selling 200 units of the product. But what happens when we include the cost of production on the advertiser side? Remember that the advertiser normally clears a 4:1 profit ($100 MSRP and $25 cost of production):

• Gross Revenue: $5,000
• Cost of Production: $200 x $25 = $5,000
• Net Revenue: $5,000 – $5,000 = $0

The advertiser ends up making no profit. So what’s the opportunity cost of this social ad run?

A pessimist looks at this situation and proclaims that the opportunity cost was 100% of the profit margin and the ad run was a failure; but an opportunist recognizes that there are 200 new customers who are trying out this new product and are likely to become repeat customers.

The Holy Grail of web analytics is consumer lifetime value (CLV) — the value of a new customer acquisition via specific marketing channels. For any business, paying nothing for 200 new customers is a heck of a deal, and with a well-managed post-purchase advertising and engagement program, this ad run could turn into a smashing success.

Gilt Gets Customer Lifetime Value

While sitting on a panel at SMX East a few weeks back, I presented a similar case on a men’s clothing company that has run regularly on Gilt for almost a year.

Each Gilt run they execute increases monthly unique new customers by 50%! Despite very low profit margins on the original sale, they set out to drive CLV via brand messaging, incentives and engagement to make this influx of new customers the most profitable and viable customer base in their marketing mix.

What they found is that following up social ad runs with aggressive Facebook advertising and content development, email, retargeting, a loyalty rewards campaign (invite friends to receive store credit) and paid search, they were able to build a streamlined process for new customers to keep the brand top of mind and come back for more.

Breaking down CLV by initial inbound marketing channel, consumers coming from Gilt now boast a 25% higher CLV than any other channel!

The Social Customer

The caveat for all this is quality of product or service. In order for a post-purchase re-engagement program to be successful, the consumers must be very happy with the quality of the product and be willing to engage with the brand in a social setting.

The company I am profiling has been extraordinary in their dedication to Facebook content management and allocation of budget to advertising channels that did not produce returns immediately. But with care and proper management, the program has blossomed into a fantastic success story.

So why do companies use social deals?

Unless there’s a huge markup on the cost of production, social deals yield very little profit. The secret lies behind what happens after the deal is over. If your business is prepared to dedicate marketing resources capable of engaging new customers to interact and repurchase products or services then social ads present fantastic access to new segments of consumers who normally may not engage with your business.

This is generally some combination of a search/display/Facebook technology provider and one or more analytics systems such as Google Analytics, Omniture, Web Trends, or Doubleclick.

Each system requires some combination of dynamic parameters or URL redirects to capture the necessary data about a click and/or conversion, so there’s a science behind getting all the systems involved to work harmoniously.

Multiple systems can function together if the following criteria are met:

1. Custom parameters can flow through one system either onto the landing page or reports can be generated including custom parameters and imported into another system.
2. URLs do not require redirects; or if a URL requires redirects, it does not have to be the first redirect.

Analytics & Redirects

For reference, below is an example URL using multiple redirects. The first redirect is in normal font, and the second redirect is italicized at the end. Note that the second redirect is URL encoded within the first:

http://akatracking.esearchvision.com/esi/redirect2.html?esvstue=13133462492&esvt=4-GOUS{ifsearch:E}{ifcontent:C}543&esvq={keyword}&esvadt=939929-1-10-1&esvcrea={creative}&esvplace={placement}&transferparams=1&esvaid=41662&useesvcid=1&esvcid_fwdsize=L&url=

http%3A%2F%2Fclickserve.us2.dartsearch.net%2Flink%2Fclick%3Flid%3D430000048238949

URL redirects present a very interesting issue. Information from the original source of traffic can only be associated with the first redirect in a string of redirects.

This is because once a URL has been redirected, the URL itself changes and the information from the referring domain (ex: google.com after a search has been executed – http://www.google.com/search?sourceid=chrome&ie=UTF-8&q=mens+shoes&utm_source=google&utm_medium=cpc) is no longer there.

If an advertiser is using multiple redirects (one redirect wraps another as in the example above) the second redirect cannot capture any information from the original traffic source. This greatly limits the capabilities of the second redirect.

Doubleclick Search 3

Doubleclick Search v3

Historically, analytics systems have not attempted to block an advertiser from using other systems by requiring their redirect to be first.

This works because an analytics system use URL parameters to capture their data and do not require a redirect.

However, Google, with the release of Doubleclick Search 3, is preparing to do just that. In a recent conversation with the Doubleclick team, they indicated that DS3 must be the first redirect.

This is because DS3 utilizes a bi-directional synchronization system which overwrites unrecognized URLs in the search engine platforms.

This is a very useful feature used to ensure tracking is correct for their system, however, by requiring it to be used, Doubleclick is forcibly limiting an advertisers ability to utilize other tools in conjunction with Doubleclick Search.

This puts an advertiser currently using Doubleclick and other technologies in a tough situation. In some cases, the advertiser will have to choose between using Doubleclick Search and the other technology providers if the other provider requires a URL redirect.

Doubleclick seems to believe, with the DS3 release, it can compete with the major search technology providers and strongarm them out of the space. It’s an aggressive move and will force more nimble startups to get crafty and utilize URL parameters to capture all the data previously only generated using a redirect.

At this point, it’s worth asking why advertisers use Doubleclick in the first place. Advertisers use Doubleclick because it is an ad server tied into an analytics package.

Running multiple advertising channels through one system allows an advertiser to mitigate double counting conversions: a common phenomenon when using multiple tracking systems (technology provider with proprietary tracking and an analytics system).

That said, from a functionality standpoint, Doubleclick Search and DART for Advertisers (DFA) are definitely not the most advanced systems available save for specific features, such as the universal floodlight tag.

If you are using DS3 we’re very interested to hear feedback on the new platform – is it all it’s being hyped up to be? Since it’s not an inexpensive system to use, it better be good.

The reality is that we, they, you, I, don’t know the right answer. One thing is for sure: we should pay attention to the funnel of clicks leading to an event because there are generic, upper funnel terms driving clicks on more product specific and branded terms which lead to sales.

But how many clicks are relevant and what key performance indicators should we use to keep the account in check?

To answer the first part of this question, we compiled a data set from 40 etailers across multiple verticals.

To keep things consistent, we used an industry standard cookie window of 30 days.

The result: 93% of purchases occur within 3 clicks when respecting the 30 day window. That means for any given purchase, there’s a 93% chance that that the conversion occurs in 3 clicks or less.

Typical Conversion Time Lag

The average number of clicks before a conversion was 1.8.

While not quite strong enough to write a scientific theory, this does give some convincing directional evidence for how we can distribute conversions across multiple converting terms.

The Trouble With Calculating KPIs

Most online retailers sell more than one product, each of which has a different cost of production and margin. As a result, return on ad spend (ROAS) is a more reliable performance indicator than cost per acquisition (CPA; cpa = cost / orders) for standard paid search bid optimization.

If we try and attribute revenue across multiple terms leading to a conversion, we’ll end up with more keywords associated with revenue, but not meeting goals.

Imagine the case where you have 2 keywords in an account with a ROAS goal of 3. One keyword gets all the last click conversions and one keyword is always in the upper funnel. Both generate the same number of clicks at the same CPC, so spend is identical (quite a perfect world, I know).

Aggregate ROAS is 4 and goal is 3. So the last click term is looking great while the upper funnel term looks terrible.

In a linear attribution model (equal distribution of revenue for all terms in the click stream), both terms would have a ROAS of 2 and bid logic would tell you to bid them both down. Boo.

This is why it’s crucial to look at a click stream holistically – considering all costs of upper and lower funnel terms and total revenue in order to understand the true performance of your account. For this reason, I am strongly against using revenue in attribution and believe that attribution models should only consider conversions and thus the most logical key performance indicator is CPA.

Analytics Without Attribution?

So how should you go about optimizing your account, considering the upper funnel, but not using an attribution model to dilute data and screw everything up?

Assuming that you or your company do not have access to advanced conversion path reporting offered by a quality 3^rd party SEM technology provider and are limited to what’s available for free in AdWords conversion funnels, here are a couple of relatively simple ideas which can be used to extrapolate the value of upper funnel clicks leading to a conversion on a lower funnel term, commonly referred to as an ‘assist conversion’.

• Start with a keyword level report within the search funnels section of AdWords and a Keyword report directly from the AdWords campaign user interface.

• Identify all terms not meeting your CPA or ROAS goal (whichever is your key performance indicator of choice).

• Isolate these terms and filter out terms without assist conversions. Use the formula below to determine if these upper funnel terms are providing value to your account:

CPA = Cost / (Conversions + (Assist Conversions / avg clicks per conversion))

This calculation uses the average number of clicks per conversion (1.8 from my analysis…your account data is available in search funnels) to calculate a ‘more true’ cost per acquisition for upper funnel terms.

By using CPA instead of ROAS, you are not directly associating revenue but are taking into account that this term is partially responsible for driving conversions. I divide the number of assist conversions by the average clicks-to-conversions in order to de-duplicate assist conversions within each click funnel.

A more extreme measure is to determine a true multiplier for the impact of upper funnel terms on lower funnel terms.

Follow these steps to create a clean environment for testing the latency effect of upper funnel terms on lower funnel terms:

• Turn off upper funnel terms for 1 cookie cycle (30 days in my test – this is the default cookie length in AdWords). Upper funnel terms can be defined as any terms without or with minimal last click conversions and observed assist conversions.

• Measure the negative effect on branded/product terms in the following 30 day cycle. If there is a negative impact on brand and product level terms that is greater than the conversions associated via last click to those upper funnel terms alone [during the previous reporting period]m then there is a latency effect for upper funnel terms on lower funnel terms.

• Calculate the multiplier and this becomes your attribution multiplier for upper funnel terms.

Note: Seasonality and other marketing campaigns can come in to play and skew the numbers in the testing environment. An ideal scenario would be to test at a time when there are no promotions running and there is no significant seasonal impact.

Before you turn off the upper funnel terms in your account, please consider this caveat: this will have a negative impact on overall revenue/conversions despite improving ROAS or CPA in the short term…which is probably why very few advertisers do it.

This is the cleanest way to determine a real attribution model.

In certain scenarios, pressure from management or a client may put you in a situation where you need to conduct such a test. Only resort to this test if you must prove the value of upper funnel terms in your account.