The fourth quarter is “the most wonderful time of the year” for many retailers, and managing paid search programs well through this period requires particular attention to detail.

Let’s take a look at some of the common trends we see in Q4 data and discuss how to, and how not to address these effects through anticipatory bid management.

We’ll use actual data from one of our client’s first holiday season with RKG to highlight some of the common phenomena. The first holiday season is always challenging, because we don’t have data to use to help us anticipate the timing. As such, this data reveals some missed opportunities that we can model the next time around.

On each chart 100% means the average week between mid-September and mid-October. Lifts above or below reflect the week’s changes from that norm. We’re using “sales dollars” as a proxy for value because it is the most widely used. A better metric is margin dollars with frauds and cancels knocked out.

Newsflash: Sales go up at holiday!

This happens for two reasons:

• Traffic volume increases. More people search, more people buy => higher PPC sales and advertising costs.
• Shopping behavior changes. Not only are there more consumers in the market, their propensity to buy and how much they buy fluctuates throughout the period.

It is this second factor that allows us to bid differently, and makes anticipatory bidding imperative. An increase in the volume of traffic by itself does not change the value of the traffic or what we can afford to pay for each click; it is the changes in the traffic’s value that allow for bidding adjustments.

So, yes traffic volumes spike:

But it’s the change in conversion rates and average order sizes that is interesting and actionable:

(There is undoubtedly social commentary to be made over the fact that AOV dips at gift giving season—we spend more on ourselves than on our loved ones?—but it’s also important to recognize that discounting leads to thinner margins which may in turn demand different efficiency targets.)

When we roll these phenomena together we get the true picture of changes to the traffic value over time.

If the goal is to maximize sales during the period at the same efficiency as normal then we should push/pull bids proportionally to the changes in traffic value. That way we maintain constant efficiency, and generate the most sales possible. (Folks who budget search instead look to spend the budget at the lowest cost to sales ratio possible. This is a totally different problem, and, we think, the wrong approach to paid search.)

So, how did we do?

Not bad, given that we didn’t have good data to use to help us anticipate the changes. From this perspective it looks like we may have overspent just a touch the first two weeks in November, and may have underspent a bit the week of the Christmas holiday and the week after.

But hold on! This view ties the clicks and costs on a given day to the sales that happen that day. We know because of order latency that many of the orders placed today came from clicks on ads long before. This suggests we might have actually underspent during the ramp up and left opportunity on the table. So, let’s instead tie the orders to the time of the clicks that drove them and see what that reveals.

Well, it’s not an entirely different picture, but it is slightly different. By this view we were pretty much on target for those first two weeks in November, and it was the two weeks after that where we may have left a bit of opportunity on the table. Good to know! We’ll use these insights this year to do our jobs that much better.

Some folks out there will suggest that you should push the gas much harder early on to catch consumers in the “early phase of the consideration cycle.” I’ve heard folks say they’ll increase their cost to sales threshold by 50% or even 100% prior to the real increase in conversion rates, arguing that those “inefficient” weeks will appear very efficient when viewed by the orders those clicks seeded. I say: if the data from previous year supports that, go for it! But we haven’t seen shifts that dramatic, or anywhere close.

One other pitfall to avoid: Black Box bidding systems that don’t allow for anticipatory bidding. We see data that looks like what’s below every year from agencies that allow the algorithms to do it all, and every year the algorithm reacts too late during the ramp up and overspends greatly after the holiday ends.

These folks didn’t fish enough when the fish were biting, and fished too much when they weren’t. Overall, they may hit their efficiency targets, but they won’t end up with as many fish as they would have had with a smart analyst at the controls.

No two retailers will show the same trends, so let your data be your guide.

Here’s to a profitable Q4!

Many times we’re asked to present a plan for the next 30 – 60 – 90 days of a paid search program. This makes a great deal of sense when we bring a new client on board and have to fix or re-build their program. It takes time to build out a program comprehensively and laying out the priorities helps manage expectations. After that point that request for a “vision” or “plan” reveals a fundamental misunderstanding about the nature of paid search marketing.

Paid search is an amazingly powerful channel for capturing existing demand. It is not a demand generating channel. Like the yellow pages people only see the ads when they’re in the market for the advertiser’s products or services. Yellow page ads don’t make people interested in taking yoga; they convince people who are interested in yoga to try the advertiser’s studio. Paid search plays the same role, but it’s easy for marketers steeped in traditional channels to miss the implications of this fact.

For demand generating channels planning is of paramount importance. Television ads seek to 1) create awareness of a brand; 2) infuse that brand awareness with positive connotations; and 3) catapult that positive awareness into action. Marketers must plan the proper messaging and imagery for each phase, and plan the timing and budgets. Even direct marketers like catalogers must plan contact strategies, page counts and formats, circulation counts and frequency of touches. Planning is required because the marketer must generate interest and action, and the time lag between marketing investment and eventual payoff is both long and not guaranteed.

Paid search is different in every respect. Budgets don’t make sense for advertisers with shopping carts on their sites because the sales happen before they pay for the ads. We don’t have to plan what products to push or when; the ads will be served to the right audience at the right time and in the right volume to respond to consumer demand.

This is not to suggest that paid search requires no work. Paid search demands well honed processes for dealing with product turnover, landing page changes, promotional offers, seasonal shifts and more. Skilled analysts study data six-ways-to-Sunday to find nuggets of gold in search logs, in click-to-order intervals, and in new classification schemes that feed statistically meaningful differentiators to our algorithms. Reacting to shifts in consumer behavior and anticipating them whenever possible takes knowledge, skill and mountains of well-organized data.

However, it doesn’t really require a plan. Indeed, predicting what projects will be most valuable 30 days from now, or 60 is both impossible and unwise. The priorities change depending on all kinds of factors outside of the analyst’s control. Locking an analyst into an arbitrary schedule prevents them from adjusting to the priorities of the moment and thereby squancders opportunities.

Asking an analyst for a long range plan for a paid search campaign is like asking your stock broker what stocks you should buy and which you should sell 6 months from now. They can make those recommendations if you really want them, but usually you’d be wise not follow them.

Absence of “a plan” is unsettling, particularly when you’re working with an agency. You can’t see them hard at work at their desks every day, so it’s natural to wonder whether they’re asleep at the switch. This fear is compounded by the reality of PPC agencies that all too often are asleep at the switch.

Trust comes from results, and in PPC results come from smart people with great tools and great processes. Anticipating and reacting to changes in the landscape requires hard work, knowledge, skills and the flexibility to follow one’s nose.

Interpreting test results for paid search campaigns can be surprisingly difficult. One reason for this is order latency. The fact today’s clicks don’t all generate orders today, but instead sales trickle in over time means that analyzing new launches and tests can be tricky. Two ways to address this complication are described below.

Problem: Successful tests can look bad initially because of order latency

For example, let’s say the order latency for a particular advertiser with a 14 day cookie window looks like this:

56% of the orders come in within 24 hours of the time of the click, 10% come during the next 24 hour period, etc.

So, on that first day you see 100% of the clicks on your ads, but not nearly all of the orders those clicks will drive—actually, quite a bit less than 56%, as clicks late in the day have less time to “mature.” For sake of simplicity, let’s ignore that bit. Doing so allows us to map out what a tremendously successful test might look like.

Let’s say an advertiser launches a new product category and new keyword ads are developed. Let’s say the advertiser’s efficiency target is a 25% cost to sales ratio, and let’s say their brilliant PPC firm nails the bids right out of the gate. The clicks generated on day 1 cost $1,000 and will eventually drive $4,000 in sales, but here’s what the results look like as they unfold spending $1,000 every day at perfect efficiency:

Yielding a day-to-day apparent cost to sales (A/S) ratio that looks like this:

The first few days of the test it appears that the efficiency is way above the advertiser’s comfort threshold. It takes the full duration of the cookie window, for the observed efficiency to match the actual efficiency of the advertising. Advertisers who don’t recognize this effect may cancel tests, or pull back on the rudder too quickly.

Indeed, what this suggests is that every launch and every extra bid push will appear to be less helpful to the top line and more harmful to the bottom line than reality. On the flip side, every pull back on bids will appear to be more helpful to the bottom line and less harmful to the top line than it really is because of the lagging orders from the higher click volumes that preceded the test.

The greater the order latency, the bigger the impact. We typically find that more considered purchases, and higher average order value advertisers have greater latency than average which impacts the proper length for the cookie window.

However, no one wants to wait 14 or 30 or 90 days to read the results of a test. In the example above, the PPC agency hit the bids right on the head from day 1. When that doesn’t happen, it’s good to find out sooner rather than later that you’re undershooting or overshooting.

Two methods for evaluating tests

Shorten the sales window. Instead of evaluating the test based on the full cookie window, study the data based on a same session or one-hour sales interval. In the example above, if 35% of the eventual orders normally come within the first hour of the click, extrapolate the results from the first few days based on that number. If the ratio of cost to (observed 1-hr sales/0.35) is on target, the test is probably on target.

If an advertiser is attempting to learn the top-line vs bottom-line trade off of bidding to a 30% A/S target rather than a 25% A/S target, compare the % increase in 1-hour sales to the % increase in cost. That should be a pretty good proxy for the A/S ratio on the incremental sales.

Tie orders to the time of the click. Most reports show the PPC costs for the day, and the PPC sales taken that day. It’s entirely likely that half of the sales taken that day came from earlier clicks. By running reports tying the sales to the time of the click, rather than the time of the order, you get a much clearer picture of what your actions on that day did for you over the long haul. This is particularly useful for studying past tests and anticipatory bidding at the holidays to see whether you anticipated the improvement in traffic quality appropriately.

The problem with the first method is that it assumes the latency for the new product category, or incremental traffic, will be the same as it’s been in the past. Not a bad guess, but potentially misleading. The problem with the second method is that you can’t use it fully until the cookie windows have elapsed.

By using method 1 during the early phases of the test and method 2 after the test is “complete,” a good analyst can avoid missing opportunities and overspending during the test, and get a dead-eye accurate read on the results after the fact.

The direct mail industry is enormously sophisticated. They’ve been on the leading edge of data modeling since the 1970s, and smart PPC advertisers and agencies would do well to study them.

RKG is in the midst of a research collaboration with Digital Element and Kevin Hillstrom of MineThatData to determine if some well-known truths from the catalog industry also apply to the world of paid search, namely that geography matters.

Catalogers have known for 50 years or more that people in rural areas respond to offers at a significantly higher rate than those in urban areas. Indeed, postal zones C & D, corresponding to semi-rural and rural areas, have always outperformed zones A & B. Is the same true in Paid Search?

The early answer appears to be: “Absolutely!” Just looking at low population density states like Wyoming, Montana, Alaska, etc, the quality of the traffic appears to be more than 60% higher than that of more urban states. We’re going to take a look deeper along the lines of postal codes to see if this trend is as clear in PPC as it is in catalog mailings.

Another factor catalog mailers have always known: the presence of retail stores matters. Not surprisingly, if you send a catalog full of terrific products to someone who lives near a physical store selling similar products, you’ll drive a lot of sales to that store. If that store is part of your retail chain, great, if not…

Our study will take a look at the impact of having a retail chain store in the same zip code as the searcher. Indeed, this might allow us some insight into the elusive store spillover effect. By comparing the quality of traffic in similar zip codes with and without a physical store presence, we might conjecture that the difference is a pretty good proxy for the amount of spillover.

Kevin Hillstrom has done pioneering work in the field for catalogers. We hope to find out whether the same notions hold true for retail chains and online pure-plays that don’t mail books.

What’s the point? Measuring the phenomena doesn’t necessarily mean we can act on it. Who wants to set up complete campaigns for each zip code?!? No one, and indeed, slicing that thin would leave you with no data to model.

However, we hope that armed with data, we can convince the engines to give us two additional tools—er, beyond the one’s I already asked for—that would allow us to manage programs at the next level.

1. Population density settings. Maybe just 4 levels, corresponding to the postal zones. This would allow us to create at most 5 variants that would capture the benefits, and we might not need that many.
2. Zip Code list tagging. Let us set up a list of zip codes representing anything (our client’s stores, their competitor’s stores, whatever). That tagged group (”my stores”) could be applied to campaigns to either establish different efficiency targets—if I know 20% of the sales happen in my brick and mortar store rather than online I can target a different efficiency threshold for that campaign—or simply suppress ad service to avoid driving traffic to a competitor’s store.

Sophisticated marketing techniques allow retailers to generate more sales for their marketing dollars, and the more sophisticated the tools the more retailers can spend cost effectively. That’s good for the retailer, the engines, and the agencies that handle complex accounts well.

At the Rimm-Kaufman Group, we’ve spent a good bit of time over the last year studying the impact of multi-channel marketing on PPC advertising. By studying all traffic to our client’s sites we can determine how often multi-channel interactions happen, how the different channels behave, how they’re involved in multi-touch interactions and tendencies to be first rather than last. Armed with this data, we can then see how different allocation schemes impact perceptions of each marketing program.

Previously, we tried to put the “PPC Buying Cycle“—touches on multiple PPC ads—into its proper perspective. Turns out it’s a pretty small effect. Unfortunately, many agencies continue to hype the effect as they seem solely interested in having their clients spend more rather than spend wisely. Is the same true with the “Cross-Channel Buying Cycle?” Yes and no.

As we look at the data across a number of multi-channel retailers we’ve found that marked differences in the way consumers use each channel mean that cross-channel interactions have profound impact on some channels and not much on others.

Likelihood of multiple touches

Channels that have the greatest likelihood of multiple touches have the most potential to be impacted by changing allocation from last touch to something more advanced. Our data suggests that the channels most likely to involve multiple touches are affiliates, comparison shopping engines and email.

Consumers who buy after clicking a competitive (non brand) paid search ad are the least likely to have been to the site previously through a different channel. In our research, only 10 to 20% of buyers who touched a PPC ad last came through any other channel previously. Compare this to affiliate traffic, where 60 – 75% of buyers came through another channel first.

This means shifting from last touch to shared credit to first touch allocation only impacts 10 to 20% of PPC orders, while the same shift has a much larger impact on the perceived value of affiliates, comparison shopping engines and email.

Initiators versus followers

If channels were all equally likely to be first as last in multichannel interactions we might find that the net effect of changing allocation schemes is zero. That turns out not to be the case. Some channels are far more likely than others to be the first touch when more than one channel is used.

Competitive PPC is much more likely to be the first touch when there are multiple touches involved. This means that moving credit from last touch towards earlier touches does tend to “help” PPC. Natural search benefits from this same phenomena.

In contrast affiliates are almost always the last touch in multi-touch interactions, meaning shifts away from last touch credit have a decidedly negative impact on the perceived value of affiliate programs. Comparison shopping engines and email tend to suffer as well.

The net effect

What we’ve found is that these two factors together mean that yes, in fact, the perception of PPC benefits from crediting earlier touches in the cycle. However, because fewer PPC orders are in play than other channels—that first effect—the change is smaller than many folks seem to think. Indeed, in our research moving credit from 100% to the last touch to 100% to the first touch, competitive PPC only picks up 5 to 10%. Less dramatic allocation shifts take those numbers down even further.

Shop.org is organizing a group to define standards for credit allocation. I’m going to throw my name in the hat to join said group, but I fear that some of the folks in the group may be more interested in trumpeting the effect than measuring it. We shall see.

In the meantime, I’ll post more of our findings over on RKGBlog.

Does your PPC program drive more sales than it’s credited for? The answer is certainly “yes,” but the sources and volume of under-reporting might surprise you.

If you drive your program through your web analytics software you may be missing 10 – 30% of the sales! This is not the fault of the software as much as it is the convenience of using javascript tags for tracking. In past years, part of the problem came from browsers not happy about running third party javascripts, but those problems have been fixed.

The problem now is not with javascript flakiness as much as when the script is run. Almost everyone wisely puts the javascript tracking in the footer of their web pages. The reason? You don’t want the customer to have to wait for the javascript to run before loading the rest of the page. If there is any problem, you want the page to load and the footer to hang so that the user can shop unimpeded.

Therein lies the rub. Because it’s in the footer the javascript can only cookie the browser after the whole page loads. For heavy, slow, image-laden pages, customers often move on to the next page before the footer loads. If the user sits on that subsequent page long enough to fire the javascript the problem will be that you’ve lost the url parameters that allow your tracking system to know the source of the traffic. That user is now flagged as “untracked” even though they came through a paid advertisement.

We know this happens and understand the scale of the problem because it shows up whenever we do data audits with clients, but also because we sometimes employ both our standard tracking and a javascript tag when we’re studying marketing channel allocation for our PPC clients. The problem isn’t that orders we see are tracked to other programs; it’s that the sales we know came through a PPC ad aren’t tracked to any marketing program.

Knowing this, you might say: “Well, if I have a sense that this happens 20% of the time, can’t I just adjust my advertising efficiency thresholds by 20% to compensate?” Yes, you can, but the problem is that some destination pages are more susceptible to this problem than others, either because they load more slowly, or because users are more likely to navigate off of them quickly. This will disproportionately penalize some types of keywords over others resulting in lost opportunity as those ads are mistakenly bid down the page.

A better way to track high-dollar marketing programs is through use of a fast redirect. The redirect is fast if, and only if the redirect server doesn’t have to do a database look-up. If the server has to look up the destination url the redirect will be slow and the server will bog down during traffic bursts. We pass the final destination url to our redirector as an encoded parameter so the redirect takes less than 0.1 seconds and the volume of redirects is almost irrelevant.

Using a redirector provides much more robust tracking, but can/should be cause for concern as well. With all of that valuable traffic passing through a third party box, it’s valid, indeed essential to ask: “what happens if the box goes down?” We stressed out about this, too. Our approach was to build in multiple redundancy by having multiple redirectors. To keep these independent, these servers are located across the country, and use different internet backbones. All the servers share the work, and are self-checking and self-correcting. If a data center becomes unavailable—say, due to a server failure (almost never), or due to DC connectivity problems (very rare), or due to routing hiccups somewhere on the web (not rare, but brief)—we use smart DNS to reroute traffic to the healthy machines within a minute. No system is 100% perfect, but this redundancy and automatic checking provides extremely high uptime.

Handling order allocation issues can be done on the fly, with the confirmation page tag sending marketing allocation along with the order details, or through a back-feed of order IDs and marketing channel credited. Any competent search agency can then base its bidding and reporting on only those sales your system has flagged as paid search or unknown.

We see all the costs associated with PPC advertising, but we don’t see all the sales generated. While this post may cause some eyes to glaze over, understand that this is not minutiae by any stretch of the imagination.

Better tracking technology plugs a big hole, but others remain. Users drop cookies, use multiple browsers, and sometimes search on one machine but place the order on another after shopping around. We can track spillover to the call center, but measuring foot traffic driven to the stores remains elusive. However, those who throw up their hands and conclude that direct marketing metrics shouldn’t be applied to search simply aren’t trying hard enough.

No PPC agency is right for every advertiser.

Andrew Goodman’s recent SEL Post dividing the world of PPC practitioners into “Muddy Ones” and “Quants” was quite entertaining and accurately characterizes most firms in the space. We at RKG like to think of ourselves as “Muddy Quants” in Andrew’s vernacular. Alan has the PhD in Stats, and… well… I’m muddy.

We believe that for large, competitive, ROI driven accounts you must have both pieces. You need an advanced system to predicatively model data and execute bid changes automatically, because hand-bidding off of spreadsheets simply doesn’t compete in 2009. At the same time, we know enough about numbers to recognize the limitations of the data and data modeling, and the need for smart marketing analysts to control the process. The predictive model has no way of knowing that there’s a sale next week on Gibson Guitars or anticipating how that might impact performance; it has no way of knowing that you’re out of Merrill boots sizes 9-11; or that a retailer has just gotten some co-op advertising dollars from Kohler; or that customers of certain types of products are much more likely to pick up the phone and call or visit the local store than the average spillover rate for that client would suggest. Marketing is muddy, no matter how skilled one is at analysis.

But not every advertiser is both ROI driven and competing in a large, complex marketplace.

How does your company think about advertising/marketing? Is the primary goal of PPC advertising to:

1. Create positive awareness of your brand? or 2. Generate maximum revenue within some acceptable ROI?

Many will answer “both”, but in our experience only one of these dominates an advertiser’s thought process.

Agencies come with the same biases: either they’re fundamentally direct marketers and live and die by ROI calculations, or, they’re advertising agencies that focus on share of voice, brand awareness and creating positive experiences with your brand.

If the principal objective is branding, you don’t need the mathematicians/engineers, and likely you don’t want them. When you ask “What sort of ‘outside the box’ ideas do you have for creating brand awareness?” the quants will scratch their heads and send you a spreadsheet pointing out why the changes you suggest would damage conversion rates, lowering revenue per click forcing bids down the page. Not what you wanted to hear, and, frankly you’re asking number crunchers to paint a portrait.

If, like our clients, you enjoy talking about hold-out tests, the incrementality of a marketing program, lifts necessary for offers to pay-off, and the cannibalization that occurs between different marketing programs, you will be miserable if you hire the advertising agency style of PPC firm. They will talk about metrics that don’t matter to you like impression share, they will talk about buying cycles without supporting data, and most importantly the results will stink. At a gathering of agency heads I heard one say “(egads) Our clients are starting to ask us to separate brand from non-brand performance data, and (horror of horrors) some are even asking for keyword level performance data!?!” I almost fainted!

Agencies often claim expertise in both, but take a look at their “about us” page to see their roots. The roots will determine the type of tree. Those who aren’t direct marketers by training do not build the right analytical systems, the data warehouses, or the algorithms and don’t train their staff to eat, breathe and sleep ROI.

Advanced data modeling doesn’t help every company that seeks ROI.

Data modeling requires data. If an advertiser’s niche is too narrow in scope or geography, the best mathematicians in the world won’t be able to materially outperform the advertiser’s own staff. An agency of muddy-ones will do every bit as well as the muddy-quants like RKG, and probably better than the algo-only firms.

Finding the right PPC agency for your firm requires knowing yourself and whether your organization sees PPC as primarily an advertising vehicle, or primarily a direct marketing vehicle, and finding an agency that shares the same view.

Effective bidding systems set bids based on expected revenue, and do so at the most atomic level possible. We set bids at the ad level when there are multiple versions of a keyword running—representing different geographies, match-types, syndication settings, ad copy, landing pages, whatever—if we have enough data to distinguish between them. When the system doesn’t have statistically significant data at the most granular level it must cluster ads together, aggregating data to predict the revenue generated from a click on any of those ads.

Since a typical campaign has few ads that generate statistically significant traffic within a reasonable window of time, the clustering mechanisms turn out to be enormously important in the overall performance of a PPC program.

The goal is to set bids for each cluster that best predict the value of traffic from each ad in the cluster, so the more closely related the ads are the better the model works.

So how do we know that keywords are related to each other? Furthermore, how do we know which relationships are actually predictive and which aren’t?

There are two ways to think about keyword relationships.

The first and most commonly used is the taxonomy approach. Keywords are related to certain product categories, sub-categories, manufacturer brands, etc. The most closely related keywords are in the same adgroup, the most closely related adgroups share a campaign, and some go the extra mile to have multiple engine accounts, with each account holding campaigns that relate to one another.There are a number of problems with this approach.

• First, by tying the analysis to the account structure you’re forced to pay a great deal of attention to which keywords go in which adgroups and how those adgroups are laid out (i.e., is an adgroup defined by a sub-category and manufacturer brand or sub-category, manufacturer brand and gender? etc).
• Creating new keywords becomes incredibly laborious because finding the right spot to put them takes time. It’s kind of like taking a random stack of books and having to put them in the right place on the library shelves (dating myself, here!).
• Finally, because at most there are only three levels, the clusters mechanisms only have two or three levels. This is kind of like having dresses that only come in three sizes; they won’t fit like a tailor-made dress with the result being overspending on poor performers and missing opportunity on winners.

There is a more fundamental problem with the taxonomy approach, however: it’s rigid. Even if you had infinitely many layers, the connections between clusters are set in stone by the hierarchy. Say, for example, you sell consumer electronics. You might have a separate campaign for each product category, a separate adgroup within the television campaign for each manufacturer brand, but what do you do with the different types of televisions (flat screen, lcd, plasma, projection, etc)? What about the different sizes?

Since the adgroups define the ad copy, the right approach is split up the adgroups into small clusters, like “Sony flat screen – big”, “Sony flat screen — medium” etc. But now you have a different problem when you start doing data aggregation. If the adgroups are tightly defined, the next level up will encompass too much (all TVs lumped together). You also lose the ability to cluster data by just manufacturer and sub-category. You don’t have the ability to track the performance of large, flat-screen TVs across all brands, which might all get a bump near the Superbowl, or just a particular manufacturer brand across all product categories.

The right way to classify terms is not with a taxonomy, but with flexible attributes. Keywords in reality can fit into an infinite number of groups, and while the ad copy clusters might need to be one way, the analytics might suggest that other clusters are more predictive for bidding purposes. For an apparel retailer attributes of keywords might reflect: the type of clothing, gender, material, color, designer/manufacturer, style, discount related (”cheap,” “discount,” “outlet”) etc. For an electronics retailer in addition to the obvious, tagging keywords as SKU-specific or not can prove enormously informative for bidding.

Splitting the account structure into micro adgroups doesn’t solve the problem. You need to be able to analyze performance across any dimension or combination of dimensions, and rigid hierarchies simply don’t permit this.Because the account structure is inadequate to capturing all attributes of keywords, that information must be databased separately by the PPC agency or your in-house team, in such a way that keywords can have any number of attributes, and attributes can be defined and created after the fact: “Oh, shoot, I’d like to tag seasonal offerings as a separate set of attributes so I can see at a glance how all my Valentine’s day related terms are performing, or how the subset of V-day discount terms are doing.”

With this detailed information about keywords an advanced bidding system can essentially learn which combination of attributes best define a close relationship, and how to set bids correctly on the middle-to-low traffic ads. The structure also allows smart analysts to easily tune bids up or down in anticipation of performance changes that the algorithm doesn’t know are coming: promotions, stock positions, co-op advertising dollars on certain brands, seasonality (birthstone months across different types of jewelry), etc.

Without attribute tagging, analytic power is limited, bidding algorithms have less information to use for clustering, and analysts have problems fine-tuning to anticipate conversion rate changes. The details matter. Make sure your team has the right data structures in place to maximize campaign performance.

Is it time to fire your paid search managers?

Many online advertisers are seeing their first-ever year-over-year declines in PPC performance. Because of this, and the general need to find something, anything that will generate more sales cost effectively, PPC programs are coming under serious scrutiny from corner offices. We welcome this scrutiny, and hope to provide some useful guidance on how to measure success in PPC in a tough economy.

Last week on RKGBlog, I argued that the wrong way to evaluate success is by the amount of management “activity.” Tail-chasing activities don’t help and in fact usually hurt performance. Productive work often looks like staring at a blinking cursor scratching one’s chin. So, if you can’t tell from the external appearances the difference between excellent work and day-dreaming, how can you tell whether you’re getting the most you can from your search program?

The answer is “in the details.” Here are 4 critical data elements to put under the magnifying glass:

Keyword-level performance. The ultimate measure of execution is in the keyword-level costs and sales. We’ve written extensively on how to look at this data fairly, and how to identify the signatures of poor performance. Follow the methodology outlined in those posts.

There are no valid reasons for the keyword-level performance to look bad. Within reason, the keyword level performance on the high traffic terms should be pretty close to your targets for those terms. Performance of the tail should make sense as well when clustered appropriately.

The classic non-explanation for wide discrepancies in keyword performance is: “Oh, those keywords that appear to be overspending are actually just early in the conversion funnel. Tail terms get credited for sales, but these terms play an important role in the consideration cycle.” Horse-feathers!

We’ve been trying to debunk this myth for years, but it has strong legs.

Try this approach: “Oh, I see, so please show me the keyword level performance data giving credit to the first click rather than the last. That should look nice and clean, right?!?” That should produce a great deal of coughing and sputtering. I suspect it won’t produce any data. When they say “we can’t pull that data,” ask: “This consideration cycle seems to be a big deal. Why wouldn’t you keep track of all the clicks given that it seems to be critical to managing the program effectively?” Expect more coughing, the words “brand-building,” “best practice…” and reference to “other factors…” If they actually do produce the data, expect to find that the keywords that were overbid based on last-click credit were also overbid based on first-click credit.

Keyword coverage. The length of the list isn’t the only issue. It’s trivially easy to generate 100K+ keyword lists using automated tools, but these lists are chock full of holes when put to human scrutiny. Take a look at the keyword list, or just one product category as a sample. Look for obvious two and three word combinations that aren’t in the list. Oftentimes the bulk of these auto-generated lists are four, five and six word key phrases that simply don’t matter, and obvious synonyms and industry specific jargon that the machines miss.

Landing pages. After the click, do you land on a page that shows the widest possible selection that responds to the search? Taking the users too deep is just as bad as landing them all on the homepage.

Ad copy. When you search for your products and categories, does the ad copy seem compelling and appropriately targeted? Remember that the goal of copy is to sell your store, not the product. Copy should answer the questions: “Why should I shop for widgets at your store?” You don’t need to sell them on widgets, they’re already looking for them. Copy is not a “game-changer” and many fall into the trap of focusing exclusively on it, but having targeted compelling copy does help.

If under this scrutiny the program looks good, does that mean you’re doing the best you can? Not necessarily. Ask your PPC managers:

• Are we bidding based on margin-level information or just top line sales? The tighter your performance objectives are tied to the actual value of the traffic the better.
• Do our efficiency targets make sense given the current climate? This is a big one! If the team isn’t hitting the current targets changing the targets won’t help, but assuming the program is hitting on all cylinders, what should we try to accomplish in search? Is it short-term profit maximization? Longer-term profit maximization (placing more value on new customers)? Top-line maximization at a neutral bottom line? These target decisions are the MOST important pieces of the puzzle if you have a PPC management team that can hit the targets.
• Are we factoring in spillover to the phones intelligently? This usually impacts higher ticket purchases more than impulse purchases.
• Are we dayparting, to take advantage of the fact that traffic value changes by time of day and day of week?
• Have we tried out content advertising recently? Google Adsense can generate sales cost effectively these days for many advertisers. We haven’t seen it get to be more than 3% or 4% of Adwords, but who’s turning down 3% right now?
• Have we tested separating exact match and broad match versions with higher bids on exact match? Same for Google.com only vs Network partners?

The economy is in the tank and consumer confidence is pretty close to zero. The time to look under the hood to find opportunities is now. If the current management team isn’t making the grade it’s a good time to move, but make sure you grade based on performance—not bluster.

Investment banks call me just about every week these days, asking whether our clients’ search budgets are growing or shrinking given the state of the economy.  Their interest is in valuing Google and Yahoo stock, but I’m not sure my responses are terribly helpful to them.

Very few of our clients have set budgets for search.  The vast majority of our clients are retailers, and almost all of those view search as a direct marketing channel with well understood ROI expectations.  How much they spend is therefore a function of market opportunity, not a board room decision.

Granted, many of our clients have been forced to think about top line/bottom line trade-offs, but they’re thinking about those trade-offs as direct marketers do:  what is the incremental ROI on the next X dollars spent in this channel, and how does that compare to other channels?  This is very different than deciding: “We’re going to spend Y dollars on search this quarter”.

I get the sense that our clients are in the minority on this, but I don’t really understand why.

Why would a company keep spending money when the ROI turns south?  “By golly, we’re going to spend this last $10K even if it doesn’t generate a sale!”

Why would a company stop spending money when the ROI is good?  “I know that every time I give you $10 you’ll hand me back $11, but I can only spend so much…”

Neither of these statements make any sense to us.

Targeting ROI with undefined budgets is a fundamentally different approach, and it informs search practices.  Algorithmically, our bid management system was designed to generate the maximum sales/margin or leads without exceeding the client’s efficiency target(s).  Other systems are designed to get the most sales/margin/leads they can from X dollars in advertising.  That’s a different algorithm, and we think the wrong one.

The problem with budgets generally is best illuminated by the specific example of campaign budgets.  One obvious question is: why do I only want to spend X on this collection of ads?  Let’s assume there is a reasonable explanation for that, there is still another question:  if I’m actually hitting these caps am I managing the campaign properly?  The answer to this is pretty clearly “no.” 

Suppose the campaign can only spend $1,000 in a day.  At $1 per click you hit the campaign budget after 1,000 clicks.  Wouldn’t you rather have 2,000 clicks for $0.50 each?  In all likelihood, you’d double your sales on the same spend.  Granted, maybe the ads are in position 7 all day, rather than position 3 for a few hours, but who cares?  Is there any reason to generate less traffic and sales on the same spend?!?  Any time caps are hit, by definition you’ve lost opportunity.  Lower bids to get maximum traffic for the dollars spent.

With that tuning however, you run smack into the first question again:  now that the campaign is more cost effective, why shouldn’t I spend more on it?

Those who can’t shake the budgeting habit often try to pick a budget that will generate the ROI they need, but that’s still backwards.  Why guess at the market opportunity in advance, when you could instead seize the day with flexibility?

A number of our clients have some component of branding in their program, and budgets there certainly make sense.  The more divorced the goals are from profits the more budgets make sense. 

However, for those of you looking to make the most of search in these trying times, consider eliminating fixed budgets as a first step.