Solely relying on matchbacks and clicks to bricks studies to determine how to best attribute revenue no longer cuts it for many reasons.  Among them:  1)  Paid search media spends are too big to rely on loose rules;  2)  Non-brand paid search cpcs have risen so high that they are close to knocking a lot of marketers out of the category altogether unless the math can be looked at more deeply;  3)  A proliferation of more affordable and better performing display media now exists for direct response advertisers if we can just get the numbers to work;  4)  While analytics still aren’t where we’d like them to be, there have been some fantastic evolutions to facilitate more sophisticated attribution of revenue.  Some examples of this are ClearSaleing, Google’s Big Funnel Beta, and bid management systems, many of which can now accommodate different attribution rules and models.

With an answer still unattainable to the question “what is the optimum marketing mix of all marketing channels for my business?”, the good news is that strides – albeit small ones – are being made here and there to ultimately get the majority of online marketers to the answer.  Some of the areas marketers are taking a close look at to formulate their models are:

• Search funnels: In an effort to have visibility to which keywords in paid search are assisting sales that result from other keywords, PM Digital had for years been providing clients with clickstream data; compiling the information, however, was laborious and clunky and definitely not turnkey.  When Google introduced Search Funnels sometime last year, we no longer need to create these adhoc reports, which means more constant access and vigilance to this key piece of data.  With search funnels, we can see which keywords are assisting sales of other keywords down in the stream, and with that data, we can choose to attribute the revenue to last click, first click, last non-brand click – whatever makes most sense.  Determining the optimum way to set this attribution rule can have a significant impact on paid search revenue.

• Remarketing: Not all retargeting firms provide transparency on which sites sales are coming in from and which ads yield view through conversions, but you get all that with Google’s remarketing (and it’s cpc pricing, too!).   Some marketers measure the success of remarketing solely through click-based conversions, but others are giving a portion of the revenue from view through conversions back to the search campaign. The assumption is that viewing the remarketing display or text ad contributed to the sale.

• Catalog sales that match up to sales from paid search can be deduplicated from paid search reporting and bid management tools provided there is a backend integration of data.

• Cross Over Analyses show the overlap of sales amongst all the online sources.  Marketers who are tracking sales on a session cookie don’t have to deal with this very much, but since most are not and because we know that all vendor reports add up to a number in excess of 100% of sales, it’s important to have visibility to which of the online sources is actually yielding the highest amount of incremental and/or unique orders.  Those that do should be given more budget, and those that don’t may need adjustments to cookie lengths.

Continued innovations to support attribution modeling are very exciting.  For example, we are hopeful that Google’s Big Funnel will finally give visibility to how display ads assist sales in search (among other things it will also track).  The Big Funnels beta opened for only a short time a month or so ago before it closed back up again.  While there are other options to track this same type of data, they are way more expensive, lengthy and cumbersome.  It is this type of innovation (and hopefully others in progress) that will fuel new thought and ultimately more sophisticated attribution models.

Paid search has an 80/20 rule in which 80% of an advertiser’s paid search spend goes toward non-brand keywords while generating only 20% of the revenue.  For some advertisers it’s a little more, and others a little less, but 80/20 is on average.

Advertisers have long abhorred paying the search engines for their trademark terms but since anyone can bid on any company’s trademark, a competitor or affiliate can siphon off sales meant for the advertiser if the advertiser is not present on the trademark keyword.  This is especially so given that the paid search trademark ad is shaded and quite enticing at the top of the page, in that it contains four deep links to desirable promotions and parts of the site that the organic listing doesn’t.  Because of these factors, most agree that letting the natural search result be the gateway to sales from all searches from  trademark terms is too risky.

Over the years there have been numerous innovations rolled out by the search engines, technology companies and SEMs, all of which attempted to up the volume of sales coming from non-brand through better execution and optimization tactics (automated bidding, long tail strategies, portfolio bidding, match types and negatives, keyword generation tools, search funnels, view-through tracking, ad automater, keywordless ads and so forth).  Some have stuck, and some have not.  For example, long tail terms, once the pillar of a sound paid search strategy, now yield very few impressions (if any) these days in the Google algorithm.  Google Instant did away with it.

One particularly effective paid search innovation was rolled out to PM Digital clients in the 4th quarter of 2009 in the way of paid Sitelinks from Google.  PM Digital’s clients yielded 30%+ YOY increases in sales from paid search in 2010 vs. 2009 with increased contribution across the board.  While spend was also up YOY, sales were up more driven by higher clicks and AOV and better conversion rates.   We attribute much of this growth to the Sitelinks which were only available at the time for trademark terms.  This meant that even more sales for a retailer were now coming from branded terms, and the gap between brand vs. non-brand sales was widening.

Sitelinks were anniversaried in January 2011.  For many advertisers, that meant an end to the consistent double-digit YOY growth coming from paid search campaigns.  Google had thought PLAs (Product Listing Ads) would be the 2010 holiday’s version of Sitelinks but that didn’t pan out.  With growth slowing, advertisers are again looking for the growth to come from non-brand campaigns, but it’s more challenging today than it’s ever been.  On a straight non-blended last click attribution method of tracking ROI from non-brand keywords, achieving meaningful growth is not attainable these days for most.  Non-brand CPCs are rising (across the board).  We theorize that possibly the minimum bid in the blind part of the auction has gone up or that maybe competition has pushed up prices, and we are certain that new methods of attributing sales are playing a factor, enabling existing advertisers to bid more aggressively than they would in the past.

For many, growth has come by changing attribution methods.  While there is still somewhat widespread resistance to blend brand and non-brand revenue (i.e. portfolio approach), there is less resistance to attribute portions of revenue differently within a campaign whether that be from view-through orders or assist keywords.  Ultimately, the combinations of revenue from different places can create a much more acceptable ROI for non-branded sales than if they had to stand on their own.

Coming Up: We will explore these attribution options in Part 2.

Overall, the volume of sales generated from Facebook e-commerce tabs has underwhelmed us.  Conversely, we are seeing volume picking up from referral links in Facebook posts going to the main website, resulting in sales.  We are learning what moves the needle in driving sales from a content perspective.

I wasn’t surprised when I read that JCPenney put their entire catalog of merchandise online in a shopping tab on Facebook. This tactic provides another avenue to put product in front of customers.  In other words, it can’t hurt, and unlike smaller retailers whose acquisition investments are designed to yield acceptable profitability, JCPenney most likely has ample R&D marketing dollars enabling what is an experiment to gain learning and flourish in the long term rather than requiring the initial investment yield an acceptable ROI out of the gate.

When a high-profile retailer like JCPenney launches an effort like this, the questions abound about whether or not it’s a best practice and one that should be emulated by other retailers.  There are pros and cons, but ultimately, assuming JCPenney’s deeper resources, this is a test; I believe it would be best for most retailers to take a wait and see approach.

Considering what JCPenney’s objectives might be, there are surely a lot of benefits.  First, enabling shopping through their Facebook page won’t depress response from their other channels, so there is no risk in one pulling from another.  Like mobile websites, apps, and all the other marketing channels, enabling shopping on Facebook is another way to be where the customer is.

There is also a viral component of people “liking” the merchandise and having it spread throughout their newsfeeds facilitating a one-to-many shopping event.  JCPenney also targets a younger audience who is not worried about privacy.  (We speculate that an older audience may be wary of doing commerce on Facebook believing the sale is going through Facebook and not the retailer which causes PII concerns given Facebook’s negative publicity in the privacy area).

Younger shoppers, of course, spend a considerable amount of time on Facebook, so the targeting is appropriate.  To the latter point, younger people may be open to Facebook as a shopping channel since they haven’t been trained to shop on websites the way more experienced web shoppers have. Finally, the targeting options and the data Facebook can provide on the shoppers is definitely attractive.

For the more budget and/or resource strapped ROI-focused marketer who has already invested in a website with lots of enhancements added over the years, taking a wait-and-see approach on how big retailers do selling their wares on Facebook, as well as watching how they hone their strategies, is the better route for now.  While some may argue that it’s just an iFrame that enables the shopping in the Facebook tab to happen, that is not necessarily an easy integration for those with limited resources.

In addition to the large retailers with big budgets who can easily implement the e-commerce option on Facebook emulating all the same merchandise the regular website houses, there is another category who may benefit from selling on Facebook:  those who don’t yet have websites. For these retailers who are just dipping their foot in the e-commerce waters, starting out with Facebook may be an attractive option since they are not constrained by integration with another site.  We are already seeing new, small e-commerce retailers going this route, which may ultimately be the catalyst for a new reality of retailers skipping the website altogether and choosing to sell on Facebook only.  It is these smaller retailers who, en masse, may eventually make e-commerce on Facebook the new normal.

In addition to the patterns of when sales come in based on the natural rhythm of the customer base, securing high positions on the search engines is also important when emails drop, when catalogs are due to be in home, when sales are running, and peak holiday days. On these days,  results will be stronger than normal, so it’s necessary to inflate bids and override bid management algorithms, if necessary, to maximize the share of impressions.

A pattern of higher-than-normal online sales that PM Digital began tracking in 2010 are the days that coincide with when people typically get paid: the 1st and the 15th of each month.  Despite predictions that Cyber Monday (November 29), Green Monday and Green Tuesday (December 13 and 14h), and Free Shipping Day (December 17) would be the top days for online shopping in 2010, MasterCard SpendingPulse’s actual top three days for online sales didn’t correspond to any of these. MasterCard reported that #1 was November 30; #2 was December 1; and #3 was December 15. All three coincide perfectly with pay days.  (Note that both comScore and PM Digital tracked Cyber Monday as the actual top day for online sales).

The fact that online sales surge when people get paid is intuitive but was still missed by most of the key research sources when predicting the top sales days for this past holiday. For search engine marketers who manage eCommerce businesses, the implication of predicting incorrectly means potentially lost sales because of bids not pushed to the max, resulting in lost impression share. The MasterCard SpendingPulse findings for this year need to be considered when modeling for 2011– not just for Holiday but for all throughout the year.

Growth – Per MasterCard SpendingPulse, online sales surged 15.4% this year with some key pockets growing faster than others (apparel at +11.2% and jewelry at +7.2% were two notable categories).  Despite best efforts and lessons learned in prior years, some websites still couldn’t handle the sustained traffic generated on key holiday shopping days.  There needs to be an effective way to test for this earlier in the year with plenty of time to work out problems associated with overloading the site.

Online Advertising Budgets - The volume of online shoppers back in Holiday 2009 took a lot of marketers by surprise and budgets were expended before shipping cutoffs.  As a result, lots of advertisers had to shut down their online campaigns which resulted in less competition and lower CPCs during key shopping days.  Because of lack of budget, few advertisers were able to leverage those favorable factors.  This year, marketers budgeted appropriately and built-in flexibility to scale as needed so as not to have ad campaigns go dark on key shopping days.  This is a particularly important safeguard since conversion rates and average order values are higher during the holiday than at other times of the year, and marketers definitely do not want to leave extraordinarily great business on the table by going dark.

Average Order Values – Across PM Digital’s client base, average order values were up across the board an average of 6%.  What we don’t know is if the consumer actually spent more or if the spend simply shifted from brick and mortar to online.

Sales on Mobile Devices - No question, the volume of sales from mobile is growing, but it is big growth on a small base.  Much of this growth is from the iPad, which is closer to a laptop or desktop than it is to a mobile phone.  There was definitely a push from several retailers to have mobile versions of their websites ready for holiday and some, like Walmart, pushed traffic from search to their mobile site from the iPad rather than to their regular site  This was a mistake in my opinion as Walmart’s mobile site is spare compared to the richer shopping experience found on their regular site.  If the iPad continues to be counted as a mobile device, it will be difficult, in aggregate, to discern the real growth in ecommerce from mobile devices.   Once people in large numbers start buying merchandise on their iPhone, Android and Blackberry devices, the needle will really start to move.  In order for that to happen, data plan charges will need to come down and connection speeds will need to improve.  Even with the onset of mobile websites (which is a tactic all marketers should explore if they haven’t already), the buying process on a phone is similar to circa 2004 on a slow connection PC.

Wholesale Brands Selling Direct to Consumer – This has proven to be a very successful tactic.  Margins are better for the wholesaler, and the brand has more control over merchandising.  It is hard to believe that a consumer would be willing to pay more from a brand selling direct than from a retailer who may be running a sitewide sale, but the consumer is opting to buy directly from the brand in big numbers.  This is a huge win for the brands, and we expect to see more brands going direct to consumer next year.

Growth of Men’s Apparel Online – We have seen throughout the entire year the growth of men’s merchandise outpacing the growth of women’s, in some cases by twice as much.  Men’s merchandise is much more commoditized than women’s which is often more fashion-oriented and, as such, trickier to buy online.  A good portion of men’s apparel purchases is about replenishing core items, and with known brands and known sizing, buying online is an easy way to stock up.

In the next few days, we will discuss how promotions and social media were used in 2010, what we learned from these tactics, and how they will affect marketing in 2011.

Before tearing apart all aspects of a campaign trying to identify the problem, look first at web analytics to see if the percentages of sales from each of your marketing channels are consistent with what you typically experience.  For example, if you normally see that 25% of your online sales come from paid search, 15% from natural search, 25% from direct load, 25% from email and 10% from affiliates,  check to see if the percentages are the same in the week that sales are low as they are in a “normal” week.  In Google Analytics, this information can be found in the Traffic Sources Overview.  In Coremetrics, you can get at this data in the Marketing Channels report.  In Omniture SiteCatalytst, this same information is found in the Traffic Sources Referrer Type Report.

At any time of the year, looking at this data is a great way to benchmark your sales by channel.  By doing this, you are able to easily troubleshoot issues when sales dip for whatever reason.  If the percentage of sales from a particular channel is much lower than normal, look first at that channel to isolate the problem.  If the sales percentages for all channels are about the same as normal but sales are still low, it indicates that business is slow overall. A particularly strong email promotion can throw these percentages out of whack, but normally it is well known that the email was a strong producer, and fairly obvious then that a higher percentage in that channel will lower the percentages in all the others.

Try setting up auto-generated site analytic reporting to regularly and easily keep on top of sales trends.  With extremely high temperatures in the Northeast this holiday week, hopefully consumers are inside sitting at their computers shopping.  Or, they could be sitting under an umbrella on the beach with a tall cold drink… in or day or so, we’ll get a chance to see how July 4^th week, 2010, has performed.

Paid search is fundamentally about presenting a relevant ad to someone who enters a keyword in a search engine. Every month, 60% of the searches on Google are brand new. With the keyword list being the pillar of the paid search campaign, keyword selection is essential. Technology now exists to scrape a page and cull a list, but the fundamental strategies for effective keyword selection remain the same now as they have been for years. Here are some of the basic keyword selection tactics that apply to the retail vertical.

Top Sellers: Site analytics can determine the top selling products through direct load and natural search. These words should be part of the paid search campaign.

Top Searched Products on the Site: Site analytics can inform what people are searching for on the site, and these words should be included in the paid search campaign. Products being searched for but not sold by the merchant should be given to the merchandising team to potentially expand the product line.

Research Tools: Hitwise, comScore Marketer, Compete and Keyword Spy are all great sources to identify not only the top keywords driving traffic to a competitor’s site, but also to identify the top keywords driving natural traffic to the company for whom the paid search keyword list is being developed. Keyword coverage in both paid and natural search ultimately adds up to more revenue for the advertiser as it creates additional real estate for the advertiser on the search engine.

Website: It’s a laborious process, but making sure all products on the website have relevant keywords is essential as is ensuring that category terms make the list, as well. It is the category terms that will drive the most volume and the product-specific terms that will yield the best conversion rate.

Catalog:  If the paid search advertiser also has a catalog, ensure that there is maximum keyword coverage on products featured in the catalog as well as ensuring that best sellers have maximum exposure.

Datafeed:  Most merchants have a datafeed that’s used for various purposes. Identifying new items in the feed is a great way to ensure that new products don’t slip through the cracks. Additionally, active keywords for products that are no longer in the feed should be paused.

Search Engine Tools: The search engines make lots of free resources available for recommending like keywords to the base word that was input.

Plurals:  For the life of me, I don’t know why a singular (or plural) version of a keyword that is at the top of the best performers list is also at the bottom for the opposite version of the word (singular or plural).  Anyone who has seen lots of paid search results will recognize this curious dynamic.  For this reason, every keyword needs both plural and singular versions of the word to be covered.

Search Query Reports:  Refer to the search query reports provided by the search engines to identify the actual words someone entered into the search engine in order for an ad to be displayed.  Refine the keyword list and match type strategy based on these findings.

Google Insights for Search: This free tool shows rising searches related to a particular keyword enabling a campaign to yield scale quickly.

Synonyms:  If you sell women’s clothing, ensure that you have all variations of women’s (ladies, girls, etc) and add these variations to the base of every keyword.

Once the keyword list is developed, give it a reality check.  Many merchants name their products with language that is not common to the public.  Cosmetics companies may refer to “fragrances,” whereas the majority of searchers will be looking for “perfume.”  Make sure your product can be found by emulating consumer behavior and by applying common sense.

Suzy Sandberg is President of PM Digital.

The way Google has chosen to price this product, however, sets it apart from the other remarketing providers.  Google’s structure is CPC, whereas the other companies charge CPM or CPA.  Cost-wise, the CPC model has a clear advantage.  Comparatively, the cost savings to the advertiser can be huge.

With CPC pricing, despite the fact thousands may see an ad, the advertiser is only charged for clicks.  This is the preferred pricing model for direct response advertisers all over the internet.  Long ago, Google’s paid search ads were originally priced at CPM, and when they switched to CPC, the source took off.   With banner advertising, only a small percentage of people who see an ad actually click on it.  That’s why display ads are more commonly used for branding efforts—they are guaranteeing impressions, not clicks.

Display advertising in a remarketing sense, though, has gained traction with direct response advertisers as the audience is more qualified.  Since the person who is presented with the ad has already been on the advertiser’s site, some level of interest in the brand already exists.  Despite this, the click-through rate of remarketing banners remains less than 20%, and often is even less than that.  As we saw long ago with Google paid search, there is a clear monetary advantage for the advertiser using CPC over CPM.

For anyone not familiar with the way remarketing pricing has traditionally been structured on CPA, one may wonder why anyone would opt for CPC over CPA.  CPA has always had an allure, but as the saying goes, there’s no free lunch.  The issue is view-through tracking,  a method of accounting for sales that some people have a real issue with.  Most CPA remarketing programs charge revenue share using a two-tier structure:  the advertiser pays more for sales that originate from people who click through the banner since it’s pretty obvious that the traffic got to the site as a result of the person seeing the ad.  They pay less for sales that resulted from someone who saw the banner but didn’t click on it (called view-through tracking).

All view-based sales are allocated to another channel, which is why many advertisers don’t like this program.  No one can ever be certain that the fact that the person saw the display ad is why they later purchased on the site.  For example, if someone was on About.com and saw the advertiser’s display ad, and some time in the future (either hours or days) went to Google and clicked on a paid search ad to go back to the advertiser’s site and then made a purchase, the advertiser has now paid twice for the order:  once for the click on Google and again to the remarketing company as a view-based order.  In the same scenario, if the person used the search bar to get to the advertiser’s site, it’s possible that they now have to pay revenue share on a sale they would have gotten for free.

Advertisers who have tried to determine how many view-through orders are incremental usually find that roughly 30% are new buyers (although in some cases, it’s much less).  For some, a 30% new buyer rate represents a winning program, while others object to paying revenue share on 70% that they already paid for through another source or would have gotten for free (as in the case of direct load).

So when you compare the Google pricing structure of just paying for the people who click on the remarketing banner to having to also pay for view-based orders, there is no comparison to the lower cost of remarketing with Google than going with a CPA program.   Google’s CPCs for remarketing are more expensive than the average paid search click, but eliminating the entire view-through portion of the sales represents a tremendous cost savings.  (Note:  view-through sales can be tracked with Google’s program, but the advertiser doesn’t have to pay for them.)

There are a few things we need to watch out for with Google remarketing.  One is scale.  CPC ads tend to get prioritized lower by publishers who may also have CPM ads in the queue, as they make more money running them.  Publishers may opt to run the CPM ad over the CPC ad. The other difference is that Google does not yet have the ability to serve dynamic banners which can be effective with remarketing offers.  The big brother/creep factor with dynamic banners makes some advertisers shy away from them, so this may not neccessarily be much of an issue for now.  I would bet that Google has this enhancement in the works anyway.

Suzy Sandberg is President of PM Digital.

Google is heavily diversifying these days, rolling out a smorgasbord of new initiatives.   Tangentially, most enhancements are related to where search is now and/or where search is going in the future.  These are welcome innovations in that they focus on how paid ads are displayed to searchers.  Up until last year, paid search display had been remarkably stagnant:  one to four shaded sponsored listings at the top of Page 1 and the rest running stacked along the right.   

Google Paid Search Display

The organic and the paid sides of search are best described as church and state.  I’m not sure which is which, but on the organic side, innovations in the way search listings are displayed have been happening more quickly and steadily.  In the past, organic search results primarily contained a list of websites relevant to the search term.  Now keyword searches return a much more varied list of results that include content from blogs, YouTube, Flickr, news sources, Facebook, websites, image ads on product search and, now, tweets in real time. 

The merging of all these different content sources on Google’s natural search results is called Universal Search, and it definitely keeps evolving, reacting to, and incorporating new types of online content.  While the way paid ads are displayed has remained static, organic display has enjoyed a considerable amount of experimentation and subsequent rollouts over the past few years. 

So now enhancements to the way paid search ads display are starting to accelerate.  Ad Sitelinks, product extensions, product ads, and video ads are a pretty big change to what we’re used to seeing for paid search.  Some of these new paid search changes are yielding real increases in traffic.  For others, it’s too soon to tell what the impact will be.   Ultimately, though, it’s great to see the paid search side of the business catching up to the innovations enjoyed on the organic side.  We hope there are many more enhancements to come.

As we saw with search, there is no doubt that Facebook’s features and tools will become more sophisticated and radically improve over time.  Facebook would surely like to monetize its 450 million users, and we know there are enhancements to the program already in the works.  With the attractive CPC pricing model, Facebook and would-be Facebook advertisers are lined up and waiting to sync up with APIs or at minimum, get easy access to reporting and some kind of bid management tool.

Looking into the future, could Facebook CPC ads ever become a force to be reckoned with in the media mix, matching or even exceeding paid search as a proportion of total online spend?

If you are a direct response advertiser, the answer is no.  The reason is because Facebook users are not there with an intent to purchase.  It is the element of intent to purchase that has made paid search effective for ecommerce. 

Direct response marketing relies on a combination of intent plus interest or demographic targeting  in order to be successful.  It’s easy to explain this dynamic as it applies to direct mail list selection.  A toy cataloger’s best lists are probably other toy catalogers.  The fact that names on a mailing list have purchased toys through the mail in the past is the implied intent to purchase.  Further, if the list of prior toy buyers makes recency available, it is the most recent buyers who will usually do the best (called “hotline buyers”).

What probably would not work, though, is a list of people who are known to have children within an age range that is appropriate for the toy cataloger’s merchandise.  Just because a person has a child does not mean they buy their toys through catalogs.  While this person may hit the exact target the toy cataloger is looking for, the intent to purchase is not there.  If I were a direct mail list planner, I would not select this list.

Lack of known purchasing intent is the reason Facebook CPC ads won’t outperform paid search cpc ads for direct response offers anytime soon.  Even if Facebook could target a person with a nine year old boy, serving a toy ad on Facebook would not be as effective as it would be on a search engine served to someone who keyed in “toys online for a nine year old boy.”  The latter is more likely to result in a sale.

In terms of what would work well with Facebook CPC ads today is using it to build up fan bases, brand advertising, and possibly direct response offers in very niche categories.

Google has made significant strides in improving their content offering over time, which, like Facebook CPC ads, may be perfectly targeted but are lacking intent to purchase.  Facebook should be carefully evaluating the tools Google now offers to leverage content and apply similar principles as they evolve their CPC product.  Comparatively, contextual search is probably the most similar model out there to Facebook CPC ads, and everyday contextual search performs a little better for direct response offers.

Suzy Sandberg is President of PM Digital.