• George Michie: Advanced Search Marketing Seminar, MarketingProfs, Feb 8th I'll be giving a Marketing Profs seminar on Feb 8th on advanced search marketing....
• Interview with Linda Bustos of Elastic Path Linda is one of the smartest people in online marketing....
• Interview: SEO Guru Stephan Spencer Stephan Spencer of NetConcepts graciously shared his thoughts on SEO in an online interview with us....

Compared to other data points, CTR is a relatively unreliable metric for judging the health of a paid search program over time because there are so many factors that need to be accounted for, some of which the advertiser has little to no control over.

If you’re worried about how your CTR looks month to month or year over year and are considering widespread copy changes, here are some factors to consider first:

Search Network Expansion:

For many, traffic from the engines’ search networks is growing faster than traffic on the engines themselves.  Click-through rates on this partner traffic also tend to be significantly lower,  by as much as 50-90%.  If you were to make no changes to how your ads were syndicated, you would see a lower CTR just from the expansion of the search network.  The upside is that the lower network CTR should not affect your quality score on the core domain.  As Google states: CTR on the Google Network only ever impacts Quality Score on the Google Network — not on Google.

So, if the network traffic is otherwise performing well, no changes are necessary, but its effect on overall CTR should be noted during trend analysis.  To get a more consistent view of your CTR over time, parse out the network traffic in the engines’ reporting tools.

Competitive Landscape Changes:

An ad’s position on the search results page has a major impact on its click-through rate for obvious reasons.  As Google’s Fred Vallaeys succinctly puts it: An ad in a higher position is predisposed to get a better CTR. If a high-bidding competitor swoops in and bumps you down a spot, you are virtually guaranteed to see your CTR fall.

If you are bidding based on efficiency and not position, as RKG recommends, you are left to accept the lower position or work to improve your quality score or conversion metric to try to overtake the competitor.  Copy or landing page changes could help, but if you’ve already optimized in those areas, you may be better off working to grow the program where there is a greater potential return on your sweat equity.  With any luck, the competitor will find they are overspending and you’ll see your position and CTR improve when they reduce their bids.

Note that Google’s Quality Score is normalized to account for the expected click-through rate differences in different positions, so all else being equal, falling a spot due to changes in the competitive landscape should not hurt QS.

Broad Match Expansion:

The engines, particularly Google, are regularly tweaking how their broad match algorithms function.  This requires them to balance ad relevance and revenue, but in recent years, the short-term revenue side has been winning out.  A broader and broader definition of broad match leads to keywords showing for more distantly related search queries and click-through rate suffers as a result.

The upside again is that Google is smart enough to take this into account when determining Quality Score, which is based only on the performance of the keyword when it matches the search query exactly.  (Consequently, changing a keyword’s match type from broad to exact will not improve its Quality Score.)  So, seeing a lower CTR due to increased broad matching is not a great reason to make copy or match type changes in and of itself.  At the same time though, this scenario may indicate that your account could use negative or tail keyword additions (with better targeted copy and landing pages) or increased parsing of exact and broad match traffic, assuming the broad match traffic converts at a lower rate.

Ideally, in our analysis we would account for the wider scope of broad match by looking only at CTR data when the search query matched the keyword exactly, regardless of the keyword’s match type.  Unfortunately, that’s not an easy task at adgroup or higher levels unless you are running mirror campaigns under different match types.  Even then, there may still be some bleed over.

Other Considerations:

• Portfolio effects: If you are frequently making keyword additions and bid changes, you may see your high level CTR stats improve or decline based on the type of keywords that receive more or less traffic after the changes.  In other words, click-through rate trends at an account, campaign or even adgroup level may not be telling you what you think they are.  For instance, growing incremental traffic while maintaining target efficiency is the goal for most advertisers, but it will lead to a lower account CTR as brand traffic with its higher click-through rates will make up a smaller proportion of the program.  Make sure to use as fine a level of detail as necessary to get an apples to apples comparison over time.
• Efficiency target changes: Aiming for a different ROI is likely to result in changes to your average ad positions and your CTR will change as a result.
• Content:  Impressions from contextual ads can fluctuate wildly, so content traffic should be parsed from search traffic by all means.

For an indication of what advertisers are up against, here’s a chart of RKG’s observed year over year change in Google Search impressions for a basket of well established clients versus the Y/Y change for Google Sites traffic under Comscore’s Core Search Report:

Even though we are taking regular steps to reduce poor quality broad match traffic, we have still seen our Adwords impressions balloon faster than clicks and much faster than traffic to Google and its partners as represented by the Comscore figures.  We’ve heard others recently ascribe lower CTRs to changes in consumer behavior during a weakened economic environment, but that would seem to be a minor effect relative to the changes coming from the engines.

Bottom line, if you are generating more traffic at the same efficiency despite lower click-through rates, chances are your program and copy are still in decent shape.  That’s not to say you should ignore the CTR metric, just be sure you know what the data is really saying before you change your optimization priorities.  Ideally,  PPC advertisers would have access to a more precise reading on quality score to steer by, but we’ll have to make do with what we have now.

• Yahoo vs. Google: Click-through-rate and Conversion The following graph shows Yahoo conversion rates (green) and Yahoo click through rates (blue) climbing in the waning months of...
• The Perils Of Computer-Generated PPC Keyword Lists George Michie wrote a solid SEL post on the perils of overly-automated keyword generating systems....
• PPC Bid Optimization Platforms: The Perils Of Inflexibility and Extremism As Joel experienced in the 58th St. Starbucks, going to extremes can lead to suboptimal outcomes....

Yahoo’s new system for managing bids on syndication partner sites is a big win for advertisers and an even bigger win for Yahoo.

We’re already seeing results. By segmenting traffic into Yahoo.com only campaigns on standard match and Yahoo.com + syndicates on Advanced Match, with appropriate bid differentials applied to the traffic from the partner network, we’re seeing materially improved results. By improvement we mean more sales at the same efficiency, meaning higher revenue for Yahoo, in fact much higher.

Let’s take a look at how Yahoo poisoned its own well:

{Okay, in truth, this is historically inaccurate. Yahoo bought Overture. Overture already served a wide network of sites including Yahoo.com, so there was never a time when they served ads only on Yahoo.com. I’m simplifying history to make a point: trying to grow revenue by expanding syndicate partnerships has actually reduced Yahoo’s ad revenue — at least from folks who actually pay attention to their bidding}

In the beginning there were ads on Yahoo.com. The quantity of traffic was good and the quality of traffic was high. But as the SERPs became saturated with ads the growth of advertising revenue slowed.

CASCADE FAILURE STEP 1: Contaminating the pool

Yahoo began to expand its network of syndication partners who would serve Yahoo’s ads for a price. The idea was that this greatly expanded exposure of Yahoo ads and would therefore increase advertising revenue.

For the sake of simplicity we’ll say the syndicate partners keep 80% of the ad spend on their site and Yahoo gets 20%.

This all makes sense, but there’s a problem: the quality of traffic on these syndicate partners was not as good. For smart advertisers who bid to the value of the traffic they must either advertise at lower returns on investment on higher advertising spends, or they must adjust bids to discount the poor quality traffic.

CASCADE FAILURE STEP 2: Inflexibility

Had they given advertisers the opportunity to bid less only on the low quality traffic, the network expansions would have been money makers, but until a few weeks ago that wasn’t an option. So, advertisers had to drop bids on all the traffic, and as we’ll see below, Yahoo ended up losing money as a result.

Let’s use the example of Acme. Initially, Acme was able to spend $5,000 a week on Yahoo and that money all went to Yahoo. Let’s say they got 10,000 clicks and could afford to spend an average of $0.50/click for that high quality traffic.

Now let’s see what happens when Yahoo adds syndication partners who drive an additional 10,000 clicks but whose traffic value is half that of the Yahoo.com traffic.

CASCADE FAILURE STEP 3: Smart Advertisers react

At first blush, in this simplest model, we see that because the advertiser is forced to bid an average of 37.5 cents across both Yahoo and the syndicates to maintain efficiency, Yahoo loses money. They lose money because while the advertiser ends up spending more money and generating more sales, Yahoo now has to split revenue with the syndicates proportionally to the volume of traffic, rather than the quality of traffic.

But wait! In reality, the effect is quite a bit worse than this because at an average bid of 37.5 cents, the advertiser won’t generate the 10K clicks on Yahoo.com it generated with a 50 cent bid average. If we say that because of the bidding, differentials (over paying on the syndicates generates 12K clicks, under spending on Yahoo gets 8K clicks) the numbers get even worse.

But wait! Getting more of the low quality traffic and less of the high quality traffic pushes down the average value, forcing bids down and shrinking revenue…

and, the lower bids shrink traffic across the board…

CASCADE FAILURE STEP 4: Add more syndicate partners

Seeing its revenue decline, Yahoo adds more partners to make up the difference => go to Step 1

WHY SMART PRICING FAILS TO SOLVE THE PROBLEM:

Through smart pricing, Yahoo applies some discount to the click fees incurred based on aggregated performance differences between the partners. This, it was hoped, would keep advertisers from dropping bids across the board. It didn’t work for two main reasons:

1. The discounts weren’t even close to what they needed to be to make up for the quality differentials; and
2. Good bidding systems don’t set bids based on cost data, they base bids on the value per click, hence after the fact discounts don’t make the systems think the traffic is of any higher quality.
   The discounts may encourage advertisers to aim a bit higher since they’re mysteriously ending up more efficient than they intended due to the rebates, but that overshooting doesn’t close the gap all the way.
3. Advertisers couldn’t see what the discounts were, so absent controls there was no reason to trust that everything would be okay at the end of the day.

ESCAPING THE LOOP: Providing controls

By giving the advertisers the ability to control the bid differential applied to the syndication partners, the advertisers can spend more money profitably and Yahoo gets to keep the windfall. No fun for the partner network, but tough turkey.

This isn’t just theory, we’re seeing it in practice. Early early returns suggest Yahoo may see a 15 - 30% lift in revenue from the advertisers and agencies attentive enough to take advantage of the tools. This growth will be a product of the huge percentage of traffic coming from the network partners and the huge differential in the average value of those cohorts.

Now, we’d love it if we could do it by syndication partner, rather than one discount across the board, but we’ll take this as a big step in the right direction.

Exciting stuff from an unexpected source!

• Yahoo Q4 ‘09 Financials and PPC Share Yahoo! announced their Q4 '09 earnings on Tuesday, making their best effort to portray a 4% year over year decline...
• eBay Adopting Syndicated Google Paid Search Ads, Dropping Yahoo Has eBay been phasing out syndicated Yahoo ads in favor of Google's? RKG records indicate a major change in the...

How can you support this claim?  Can you discredit the possibility that the apparent improvement is just noise? How can you apply that authoritative label of “significant”?

Here I’d like to walk you through a basic test of significance that you can use to de-mystify changes in your paid search data.

If you’d like to skip the math, click here.

Let’s start with a situational example… say you’ve added Google Site Links to your brand ads and you want to show that brand click-through rate (CTR) has improved as a result.

1. First, you need to know what value brand CTR is potentially improving from.  Let’s call this value mu (pronounced myoo), and you can choose it in a variety of ways: the average or median CTR over the past month, the average or median CTR from this time of year last year, etc. It should really be whatever value you believe CTR to truly center around.
2. Next, you need data points.That is, you need several days of CTR data since the Site Links have been running. How many days is up to you. Generally, more is better, but I’ll touch on that later. The number of days you have is n. Take the average of the CTRs from those days; this is called xbar. Lastly, take the standard deviation (excel function stdev) of these CTRs and call it s.
3. Now we can compute a t-score, and with it, the probability that the change in CTR you’re seeing is or isn’t attributable to chance. Set t = |xbar – mu| / (s/squareroot(n)). Then use the function tdist in excel, and for the arguments, plug in t, n-1, and 1. The number that this function returns is the probability that the change in CTR is simply due to chance, aka noise. If this probability is very small, then we say CTR has changed significantly.

Enough Math! Is The Change In My Data Significant?

I’ve prepared an excel spreadsheet that handles the arithmetic. In this model, change the gray shaded cells to reflect your data. Enter the data that you think has fundamentally changed in column C. Only include data points since the change began. Then, in cell G2, enter the value from which you believe the data to have changed. That is, the average value of the data before the change.

The value p, produced in cell G7, is the probability that the change you’re seeing is only due to chance, and thus meaningless. Typically, a p-level must be below 5% to be considered significant. (If you want to be super, super sure, you can use 1% or 0.1% instead.) In other words, if your p-value is 5% or less, you can confidently say that the change in your data is real, definite, and due to something other than statistical noise. It’s a pretty safe bet that whatever initiative you took – whether it was switching landing pages, altering ad copy, or refining your bidding – was the catalyst for the improvement instead.

Allow me to fill in the spreadsheet with an example. For an imaginary online retailer, brand CTR hovers around 4.4%, so I fill in cell G2 with the value 4.4. The retailer enables Google Site Links, and CTRs for the 3 days afterward are 4.3, 5.2, and 5. So I enter those three data points into column C. And voila… the p-level comes back as 12.66%. This says that there is a 12.66% chance that the rise in CTR was due only to noise.

Not significant. Sorry, click-through-rates haven’t really increased, or at least, we can’t be very confident that the observed change is anything more than random noise.

But… three days is not much data. As smart analysts, we are cautious when examining trends over only a few days, and this significance test incorporates such wisdom. As the number of data points (n) you use increases, p-levels fall. For example, if all the numbers in the above example were the same except that you used 7 days instead of 3 (so n=7), the corresponding probability drops to 2.6%. In this instance, it’s very unlikely (2.6% unlikely) that the increase in CTR was due to noise, so here you can rather confidently say, “Yes, CTR has increased, and it wasn’t due to chance. It was probably due to the site links.”

• When “Statistically Significant” Isn’t Direct marketing testing is both art and science. Here are three situations where "statistically significant" results might not have business...

The big number that jumps out is a 15% Y/Y decline in search advertising revenue.  From Yahoo:

Marketing services revenues from Owned and Operated sites were $971 million for the fourth quarter of 2009, a 9 percent decrease compared to $1,063 million for the same period of 2008.  The decrease was primarily driven by a 15 percent decline in search advertising revenue and a 1 percent decline in display advertising revenue.

Results from a basket of RKG clients are in line with that figure with a spend decline of 20% across Yahoo and its partners in Q4.  By comparison, RKG’s spending on Google rose 14% Y/Y.

While a general improvement in the economic environment helped all three engines’ Q4 numbers, Yahoo continued to bleed share to Google and Bing at a fairly steady rate according to our data:

Interestingly, Yahoo’s reported numbers were bolstered by a relatively impressive 6% Y/Y lift in revenue from affiliate sites.  While they don’t break out specifics, it stands to reason that a big portion of that revenue is from Yahoo! Partners on search.  It’s unclear if the 15% decline in search advertising revenue includes the partner results; If so, it’s an even more impressive gain, but one with not so obvious costs.

To Yahoo’s credit, they have offered advertisers the ability to block poorly performing partner sites for some time now; however, if most advertisers are not parsing their data this finely, the expansion of Yahoo! Partners with lower and lower quality traffic will continue to drive down ROI and ultimately spend on Yahoo O&O sites where the bulk of their revenue potential is.  Advertisers and agencies taking advantage of Yahoo’s partner exclusions (like RKG) benefit from this scenario through reduced competition in bid auctions and lower CPCs.  Good for our clients, bad for Yahoo.

Fortunately for Yahoo, this effect is likely to diminish over time as more advertisers wise up to just how much money they are throwing away to poorly performing partner traffic.  Yahoo has also made a smart move recently by improving advertisers’ ability to manage bids on their network.  In addition to the existing all-or-nothing exclusions, we can now apply a percentage bid adjustment to partner traffic.  This is a very welcome move and the benefits should flow both ways in the long run.

• June 2008 Google, Yahoo, Microsoft Paid Search Market Share Looking at our agency's client base in aggregate, last month Google received 79% of our clients' ad dollars. Yahoo received...
• January 2008 PPC Ad Spend Share: Google vs. Yahoo + Microsoft Here are our January 2008 PPC search engine share numbers....
• Paid Search Share July & Aug ‘07: Google Up, Yahoo Down, Microsoft Steady During July and August, across our client base, Google picked up four points of paid search market share, mostly at...

Today we describe a layered approach to evaluating a paid search program. The top level view might be the right view for the C-level execs who can’t get too deep in the weeds. The day-to-day management team should reflect on the high-level view, but also needs to roll up their sleeves and get dirty to make sure they’re getting the most out of the channel.

It’s ALL about the Competitive (”non-brand”) Search Keywords

The clear and actionable view of paid search at any level of detail comes from studying the performance of the competitive search program separated entirely from trademark/brand search. People searching for the advertiser by name aren’t driven by search, and paid search ads on those brand terms, already at the top of the page, can neither create more brand traffic, nor materially create incremental business from that cohort. “Brand” ads primarily cannibalize organic and affiliate links and managing those brand terms takes little effort. The real effort and the real incremental value of paid search is in the competitive (aka “non-brand”) side of the game, hence that is where the focus needs to be in the evaluation at every level.

The 10,000 Foot View

1. Is Competitive Paid Search hitting efficiency targets? If not, go no farther, the program is wasting money and heads should roll. Spending the budgeted amount for media regardless of the ROI is the wrong way to approach search.
2. Is the Competitive Paid Search program (herein after referred to as “the program”) growing in sales volume? If the efficiency targets are being hit and the program is getting bigger, that’s a positive sign. It isn’t a definitive sign one way or the other. The recession shrunk many well-managed programs; and programs can grow simply from channel shift without being particularly well handled.
3. Is the program growing as a percentage of total site sales? Much more telling than #2, growth as a percentage of the whole site shows that paid search is an increasingly important marketing channel which speaks well of its management. Again, shrinkage in this regard doesn’t necessarily mean the program is run poorly; as I discussed with Ken Cassar of Nielson: having a huge percentage of site sales coming through paid advertising isn’t necessarily healthy in the long run. Repeat customers are key to profitability, and they don’t typically come through competitive paid search ads.
4. Is the program growing relative to competitive natural search? One probably shouldn’t pit the SEO team against the paid search team as though they’re rivals, since great progress by one doesn’t necessarily imply the other isn’t doing well too, but it’s a metric to watch.

The Ten Foot View

Take a Keyword level performance report from Google for the last 2 or 3 months. (Whether this is first touch data, last touch data or some allocation doesn’t make much difference as we and others have shown time and time again.)

1. Do the high traffic terms each meet the efficiency objectives within reasonable tolerance for statistical noise?
2. In category/subcategory groupings do the lower traffic KW meet their efficiency targets in aggregate?
3. In other groupings meaningful to your business do the lower traffic KW meet the efficiency objectives?
4. Clustering by traffic volume, do the aggregates still make sense efficiency wise?

Bid management and analysis is half the battle, and the most complex half by far. This view will reveal the most damning evidence of problems if there are problems with the program. Do not accept “Buying cycle” arguments as an excuse for poor efficiency.

If the program looks good from each of the perspectives above then your paid search managers are handling at least the fundamentals of bid management well.

The Ten Inch View

1. How large is the list of active Keywords on Google? There exceptions to every rule, including this one, but 5 to 10 KW per product on the site is a pretty good benchmark. The Long Tail is valuable for almost every program, and broad match is no replacement for a comprehensive list.
2. What’s the quality of the KW coverage? Spot check some random sub-categories to see if all the obvious and not so obvious permutations are covered.
3. Are the Keywords machine generated gibberish? Machines do a lousy job of generating lists: leaving gaping holes, creating dangerous untargeted and inappropriate phrases and generating thousands of valueless phrases of more than 4 words. When evaluating the length of the list these should be ignored.
4. Are the landing pages appropriate? Placing visitors on a page equal in depth to their search improves conversion rates. Linking more general phrases to product pages leaves opportunity on the table. Only SKU specific Keywords should land on product pages. Ask for a KW - Ad Copy - Landing Page spreadsheet ranked by traffic volume descending. Check the landing pages for the highest traffic KW, and spot check others down the list.
5. Is the ad copy compelling and sufficiently specific? The goal is to sell the advertiser’s website, not the product. The user has already expressed interest in the product via their search, the goal is to convince them that your site is the best place to shop for it.

The Microscopic View

If the first three successively closer inspections reveals a healthy program, congratulations, you’re in the minority! Your program is healthy, but not necessarily in peak condition.

Peak fitness comes from applying the most sophisticated techniques in paid search, and executing them at the highest levels.

• Smart use of negative associations developed from user search strings. Proper databasing of referrer data and use of Google’s research tools should reveal opportunities to trim fat and raise bids on broad matched KW.
• Creating separate campaigns for exact matched and broad matched KW and bidding more on the exact match versions based on calculated performance differentials makes a material difference in performance.
• Placing the exact match version referenced above on Google.com only and pushing those bids more based on the syndication effects researched previously also produces positive returns on the effort. Using syndication partner exclusions on Yahoo while we wait with baited breath for their bidding differential option to go live will achieve the same effect.
• Anticipatory bidding based on seasonal shifts and shorter term promotions helps take the most advantage of those hot periods.
• Day parting done smartly (using the time of the click, not the time of the order) to exploit intra-day variance in performance values requires good analysis and superb tools.

These fine tuning mechanisms create the competitive advantages needed in a tough market place.

Still passed the test? Fantastic! Your paid search program is ripped!

But the fact that the program is managed beautifully doesn’t mean your business is getting all it can out of paid search.

• Are the managers aiming at the right targets?
• Are they factoring in returns and cancels?
• Are they measuring call center spillover?
• Are the targets set based on margin data? Handling co-op advertising dollars?
• Are lifetime value considerations factored in? By the type of first purchase?
• Is the tracking system catching all the sales?
• Is the cookie window sensible?
• Is attribution between channels handled correctly?

If you don’t know the answers it’s time to start asking why and when. Paid search remains the number 1 mechanism for generating incremental business online; the fact that it’s no longer new and sexy doesn’t mean it can be safely ignored.

• DMA Search Engine Marketing Certification Program: Audit Your Paid Search Campaigns If you're interested in auditing your paid search campaigns but haven't the time to do it yourself, give my colleague...
• 12 tips for winning at paid search in 2007 Here are 12 tips for winning at paid search in 2007. ...
• Evaluating PPC Performance in a Down Economy Is your PPC program hitting on all cylinders in this environment? The stakes are higher than ever....

Well, a lot has happened in the mobile space in the past year:  Apple sold another 20 million or so iPhones worldwide, Motorola and Verizon launched a $100 million marketing campaign for the Droid, AT&T claimed a nearly 7000% increase in data use over the last few years and Google finally debuted its self-branded Nexus One.  It’s a good time to revisit our numbers.

As we did previously, we are looking at click traffic from mobile devices with full internet browsers that trigger standard PPC ads.  In December of ‘08 we found just 0.4% of PPC clicks came from mobile users.  By October of ‘09 that figure had already jumped to nearly 1.3%.  (Although it’s dwarfed by Google, it was around this time that Yahoo began showing its full search ads on iPhone-like devices.)

In early January of this year, mobile’s share of PPC clicks is approaching 1.8%:

Traffic from devices running the Android OS has increased the most dramatically in recent months, with Android’s share of mobile PPC clicks rising from 6.3% in October to 10.5% this January.   Android’s quick rise corresponds with Motorola’s Droid launch as well as the launches of a slew of less-heralded Android devices late last year.

Share of Mobile PPC Traffic for Select Devices:

At the same time, iPhone’s mobile paid search traffic share (including iPod Touch) has been dominant, but somewhat flat, while Blackberry is down significantly.  It will be interesting to see if the Google Nexus One launch gives Android another big bump despite the reports of anemic first week sales.

So, we have this big increase in mobile traffic, it must be a boon to advertisers, right?  Unfortunately, it doesn’t look that way.  Conversion rates are very poor compared to traffic from desktop/laptop computers.   There are a couple of ways to think about the low conversion rates:

• Conversion Rates are Better than they Appear
  1. Cookie breakage as people browse on their mobile devices, but purchase on their home or work computers. 
  2. On some devices, there’s also more likely to be technical issues with tracking cookies or scripts that we don’t face on full PCs or Macs.
  3. People click on the ad, find the store nearest them and purchase in person.
• Conversion Rates Really are Awful
  1. Shopping on a tiny screen with two-thumb typing may be too much of a barrier for folks to overcome.
  2. People click on your ad and walk into someone else’s store to make the purchase.
  3. People are in someone else’s store want to compare prices online and click on your ad to show the clerk and get the benefit of the store’s price-matching policy.

If the advertiser doesn’t have brick and mortar stores, there’s a pretty good reason to be biased towards the negative assumptions. If you’re in that camp, it makes sense now for advertisers to parse out a duplicate campaign of top Google keywords to display on mobile devices while using the opt out option for existing campaigns.  The mobile campaign can be bid based on its own performance without diluting the results of the core program.

Mobile search growth is unlikely to slow anytime soon, while the lines between devices will continue to blur as more and more gadgets incorporate full web browsing functionality.  Google is clearly anticipating this as it pushes to get Android, with its close integration of Google services, on as many devices as possible.  Whether user behavior and our ability to track it begins to match what we see on traditional computers anytime soon is still an open question.

• Early Read on Google’s New Mobile Phone Network Setting Google's New Mobile Device Network Distribution Setting - Should We Care?...
• Paid Search Trends, 2006 vs. 2007 YTD Linking to a post today I wrote over at Search Engine Land on year-to-date trends in PPC....
• eBay Adopting Syndicated Google Paid Search Ads, Dropping Yahoo Has eBay been phasing out syndicated Yahoo ads in favor of Google's? RKG records indicate a major change in the...

How important is the tail? As with most pieces of the paid search game the answer is “it depends.” The importance varies tremendously by the type of business, the number of products/services offered, and how people search for what the advertiser offers.

DEFINITION:

There are at least two different ways to define the head/tail demarcation.

1. A fixed number of KW. We could make the case that the head is represented by the X highest traffic KWs, perhaps corresponding to the number of KW someone could manage by hand. We picked 250 as that number.
2. The number of KW with more than X amount of traffic. One could make the case that any KW that generates X amount of click traffic in Y days is a head term. For our purposes we picked 500 clicks in 90 days.

Because we’re making the point that the tail is important — albeit to varying degrees — we decided to use whichever definition above gave the advertiser the smallest tail. If the tail still looks important using the least generous definition then it’s hard to argue the point.

METHODOLOGY:

We studied 90 days worth of data from Q4 for ~20 Google accounts. We grabbed diverse advertisers with programs ranging in size from $15K in spend per month to more than $750K per month; ranging from advertisers carrying widely available commodities to those selling their own products; from 100s of thousands of products to less than 1 thousand.

For simplicity sake we aggregated the data by KW collapsing the distinctions between geo-targeted campaigns, different syndication and match-type settings.

We submit to our readers that all these programs have been “well-managed.” KW lists have been built exhaustively by our OCD analysts, the “head” KW have received heavy attention with smart use of combined match-types, syndication settings, thorough and appropriate negative lists, well-chosen and tested landing pages and copy.

FINDINGS:

1. The importance of the tail varies tremendously. Of the client accounts studied the least important tail generated 8% of the client’s competitive search sales during the period. The most important tail generated 83% of the sales — the top 250 KW by search volume netted only 17% of its total sales! The median tail for the group generated 31% of the competitive search sales.
2. People search differently for different types of services/products. The number of KW generating at least one impression over ninety days ranged from a mere 3,300 to almost 300,000. While this is somewhat related to the number of products on the site, that doesn’t explain all the variance. In one case an advertiser with only ~1,000 products had more than 90,000 KW fire at least one impression. At the other end of the spectrum, a retailer with more than 20,000 products had fewer than 9,000 KW generate an impression.
3. The importance of the tail does not depend on the size of the search spend. The advertisers with the most important and least important tails referenced in #1 each spend more than $250K per month on Google alone.
4. The tail is most important for SKU-based commodity retailers. Not surprisingly, those who re-sell goods in a competitive marketplace find the most value from a well-managed tail.
5. The tail is critically important for some advertisers. In the most extraordinary example, a client with 1,200 head terms (each generating more than 500 clicks during the evaluation period) nevertheless generated more than 71% of their revenue — over $3.2 million — from tail KWs.

From this analysis two facts are plainly obvious. First, with proper management, the tail is clearly significant to almost every advertiser. Even the company with a tail amounting to “only” 8% of its sales generated more than $600K in sales from that tail. Second the degree of importance depends widely on the types and quantities of products sold and how users search for those products. All of these accounts are managed by RKG analysts with the same core training, the same tools and the same approach to paid search management, yet the importance of the tail to these clients varies tremendously.

Some might object that the tail occupies too much of the management time and that by eliminating that burden greater focus on the head will yield better results. This may be the case for those lacking the proper tools. If 99% of the KW are “tail” and they require just as much time to manage as the other 1%, then clearly the cost of managing the tail may outweigh its benefit.

However, this is a false choice between two bad alternatives ignoring the existence of an obviously better alternative. A well-managed, extensive tail shouldn’t preclude a focus on the head. Smart algorithms and automation of repetitive tasks that can suck up management time should allow a good analyst to manage a monstrously large KW list in the millions of KW almost as easily as they manage much smaller programs. At RKG we devoted a good bit of IT time to building those tools for our analysts with the idea being that smart humans shouldn’t be wasted on robotic tasks.

Proper allocation of management resources should not require a “tail-ectomy”. There is too much value to be lost. Over-reliance on broad match to make-up for a missing tail results in less targeted bids, landing pages/ ad copy, less flexibility, and ultimately a smaller program. Folks who say otherwise either lack experience or the power tools necessary to manage a comprehensive program.

• Broad Match is No Substitute for the Tail Does broad match + negatives and lots of love and attention yield better results than a fully developed KW list...
• Darden’s Phil Pfeifer Turns The Long Tail On Its Head My beef is not so much about the Long Tail concept as it is with the name. ...
• Head Terms, Tail Terms, and SEM Agency Compensation Models Great post by Avinash Kaushik on the role of head vs. tail terms in paid search....

Experience has taught us to try to talk our clients off that intellectual ledge. They don’t always listen, and that’s cool. True, every business is different and maybe what hasn’t worked for the last 100 clients will work for this one…but I’d bet against it.

The person searching for “Gifts” has no flippin’ idea what they want. By definition, the chance that any category-specialist retailer will have it is really really small. Conversion rates will be dreadful. Combine that with the fact that so many other lemmings can’t resist the temptation that it might cost $2 per click to be on that first page and inefficiency is guaranteed. Worse, because so many people DO search for that it will not just be wasteful percentage-wise, but will be a LARGE loss as well.

Conversion rates reflect how well the advertiser’s wares match the average user’s intent when that user types in a particular phrase. The fact that some guys would LOVE electronic gadgets does not mean the electronic gadget store will do well on “gifts for guys.” Some guys are more likely to want sports equipment, or tickets to a big game, or power tools, or music, or…the list goes on. Unless the advertiser sells ALL of those things, they’re going to find “gifts for guys” on page one a really unfortunate place to be. I can hear our Google and Yahoo reps screaming as they read this :-)

We don’t always succeed in talking people out of these tests — sometimes the allure of those keywords-cum-bug lights is just too strong, and often it’s the corner office driving the decision — but we do feel an obligation to try. It seems to me that that’s the job of an agency — to use its experience to help their clients avoid train wrecks. We’d expect the doctor to advise us against hitting ourselves in the head with a hammer to stop a headache.

• Comments on The Lane’s Gifts v. Google Report by Alexander Tuzhilin After reading Tuzhilin's report, I have a better appreciation of Google's fraud detection methods. I come away feeling that fraud...
• Top 4 Paid Search Traps To Avoid in Early 2008 (plus final PPC stats from 2007 4th quarter) Did your Q4 PPC Sales meet expectations? How are those expectations set? Every retailer is different, and there can be...

Frustratingly for our loyal readers, I keep grabbing different baskets of client accounts. Part of this is just the devil in me acting up, but part is a function of the fact that we’re trying to illuminate consumer behavior through paid search. As such we want to look at clients that haven’t materially changed their efficiency objectives, their product offerings or pricing/promotion strategy. As clients included in previous studies change gears they have to be dropped and others brought in.

We also report on median values for the baskets not weighted averages. We believe medians paint a truer picture of what’s happening for “most” advertisers. The trouble with medians is they don’t “add up”. Looking at weighted averages the sales delta will be a product of the traffic volume change, the conversion rate change and the average order value change. While that’s true for each client studied, it isn’t true of the median values for each metric within the basket of clients.

Apologies finished, let’s look at the results!

RESULTS:

November sales through paid search were up from last year but not by as much as October’s were:

Indeed, the way the holidays fell, this monthly approach gives an artificially favorable view of November 2009 performance; this year “Cyber-Monday” was November 30th, last year it was December 1st. With that helpful tail wind, one would expect November to have been at least as strong as October if not stronger.

Also intriguing, while most of the performance declines this year have been a function of traffic volume more than conversion rates or order values, the pattern here is different. Clearly the volume of traffic is strong and getting stronger; it’s the conversion rates that have tailed off in November.

Pulling back the covers, let’s look at pre-Thanksgiving November vs the first big holiday weekend, from Black Friday through Cyber-Monday. {Does anyone have naming rights for the Saturday and Sunday? All in favor of “Michie-Saturday”?}

For this same cohort the sales growth was much stronger over the holiday weekend than prior to it:

• Pre-Thanksgiving November YoY (11/2/08 - 11/27/08 vs 11/1/09 - 11/26/09):
  • Sales up 9% on higher traffic volumes and stronger AOV, but
  • Conversion Rates DOWN 15%
• Holiday weekend YoY (Black Friday through Cyber Monday):
  • Sales up 27% on higher traffic and AOV, and
  • Conversion Rates are off only 5%

This got me thinking about conditioned behavior. Does this data point to consumers being trained to wait for the holiday sales? I think so.

If we break out the comparison by day over the holiday weekend the numbers are quite interesting:

Notice that the conversion rates are comparatively higher on Black Friday and Cyber-Monday than during the intervening days. Bear in mind, we’re comparing these days to the same days last year, so the above chart does not simply say “conversion rates are higher on BF and CM”, it says the consumer behavior on those two days are more extreme than they were in 2008.

Promotional calendars and the media hype surrounding the holidays seems to be conditioning shoppers to kick-tires earlier and come to buy when the sales are on.

Interesting stuff, love to hear if others are seeing the same types of things.

• Shop.org: 51% retailers surveyed report higher sales in November 2008 vs. November 2007 Some retailers are having a very very tough time. Other retailers are seeing growth. The situation is quite mixed....
• December Paid Search Benchmarks December growth rates were okay, but the downward trend from October is puzzling....
• Top 4 Paid Search Traps To Avoid in Early 2008 (plus final PPC stats from 2007 4th quarter) Did your Q4 PPC Sales meet expectations? How are those expectations set? Every retailer is different, and there can be...