A common way of managing a brand in paid search results is to bid all brand terms to the first position. While this strategy may seem correct from a brand management standpoint, it misses out on many of the nuances of search engine marketing(SEM). For one, many brand campaigns contain over a 1,000 brand variations, of which over 500 may get clicked on in a given month. When a query is typed, these keywords compete among themselves for the impression.

Consider a brand called 123loans. A searcher may enter the brand query “123loanz.” Your campaign contains the following keywords:

Given the CTR of 123loans, its very likely that 123loanz is mapped by the search engine. The end result is an impression sucking effect from the lower bid keyword to the higher one and the net result is a higher cost per click (CPC).

My point is that brand keywords should also be optimized based on return on investment (ROI) considerations.

We looked at some of our clients for whom we managed a large number of brand keywords and the CPC and discovered the position distributions were quite telling. First, the number of variations of broad matches in each campaign was far more than exact or phrase matches. Next, more than 50% of broad matches were below average position of 1.5. In contrast, for exact matched keywords almost all were bid to the first position in search results for optimal performance.

Next, we looked at CPC distribution, using a special visualization tool called a volcano chart. The “mountains” represent the distribution of CPC by keyword, with each mountain representing a match type. Each red square represents a keyword and these have been plotted by CPC. For broad matches, the peak of the mountain is close to $0.2, indicating that the mean CPC is about 20 cents. But look at the variations. It peaks at $1 or more. One might argue that the positions for broad matches are all the way from position 1 to 7, so a variation in CPC is to be expected. However, if we look at the exact matched keywords, most of them are being bid to position 1. Although the mean is 10 cents, the range is from 1 cent to 40 cents.

From a statistical standpoint, these keywords will exhibit heterogeneous behavior, meaning they will not behave the same way, deliver the same performance or have identical ROI potential. If that’s the case, why should they all be treated the same way from a bidding standpoint?

The key takeaway from this is that you should look at the performance of your brand keywords just like non-branded ones, and even if you want them all to be at position 1, you should do so by bidding smartly. In the absence of sophisticated bidding technology, a good rule of thumb would be to leave the exact matches at position 1 but to also treat broad matched brand terms just like a regular non-branded keywords. Bidding all brand keywords to position 1 is at best inefficient and at worst foolhardy.

A question often posed by marketers is, “What is the relative importance of different factors Google uses to determine quality score (QS)?”. Some of the factors mentioned on the AdWords blog are:

• Click through rate of the keyword and the matched ad
• Account history
• Landing page quality
• Keyword/Ad Group relevance

The question for an advertiser then becomes, “What factor should be the primary focus when trying to improve my quality score?.” The answer is overwhelmingly the click through rate (CTR).

For the analysis, we looked at several Google accounts and plotted quality score vs. CTR. A typical plot for a large account with 500,000+ keywords looked like this:

Several interesting patterns show up:

• Quality scores from 1 to 8 are very well explained by CTR. Notice how the linear regression (line fit) aligns so well with the observed pattern. The R squared of 72% of the trend is explained by CTR alone.
• There is a sudden jump at a quality score of 8.
• While CTR does not explain the jump between 9 and 10, there is a huge jump in CTR between 8 and the higher quality scores.

But there is more to it. Some quality scores appear more frequently than the others. We found in our analysis of millions of keywords that quality scores of 8 and 9 are very rare. Here is a typical example for a Google account. It appears that the quality scores of 8 and 9 are “transition” regions. While a linear trend explains the QS-CTR connection until the score of 7, keywords with QS of 9 and 10 require a very high quality score compare to the rest.

We found this pattern across the board, in all verticals.

The takeaway is that when looking to improve quality score, first seek to improve your CTRs. This will have the biggest impact, by far. Do not worry if you see very few keywords at a quality score of 8 or 9. These scores are rare and do not appear to have a direct connection with CTR related factors. There is not much you can do to get these scores. In fact, it likely has something to do with the way Google’s Quality Score algorithm works. Once you have the CTR piece of puzzle solved, work on the other factors (such as landing page quality, ad copy relevance and campaign structure) to improve your quality score. And don’t waste your time fretting about getting the highest score.

The beauty of online marketing is that every step in the click path is visible. A searcher clicks on an ad, lands on a merchant’s site and navigates to purchase—all of which can be viewed and analyzed.

This process is often referred to as multiple metric optimization. However, this term is often misused so it is worth spending some time on explaining the process and how marketers can use it to their advantage.

Simply put, multiple metric optimization refers to the class of methods that seek to maximize a marketing goal (for example, ROI/Revenue) using more than one event in the click path. Examples of these metrics are clicks, leads, revenue, purchase, visit time, times of day etc. It is important to note the word “optimization” in the context of bid management necessitates the need for statistical algorithms. Applying arbitrary rules or filters to five or so metrics does not mean you are optimizing to multiple metrics. Perhaps an example will make it clear.

Consider that you run the marketing campaign for an online magazine. You let visitors get a free one month subscription and after a month, many of them sign-up for a year. You realize revenue only when a free trial converts to a subscription. As a smart ROI conscious advertiser, you want to advertise to maximize subscriptions. Mathematically speaking, you want to maximize subscriptions subject to a budget/CPA goal by using a constraint such as not exceeding a certain amount. However, you then need only maximize to the revenue event i.e. subscriptions so there is no need for multi-metric optimization. So why bother?

The trouble is that between a trial and a subscription there is a 30-day delay. When you have enough data, you can predict the revenue generating potential of every keyword. However, this is seldom the case for long-tail search terms and sometimes even for mid-tier terms. For example: If 10% of all clicks convert to registrations and 10% of those registrations convert to subscriptions in a 30-day period, we are in effect saying that one in a thousand clicks becomes a subscription in a 30-day period. In other words, if a keyword on average got 30 clicks a day in a 30-day period, you could expect one subscription from it. By most definitions 30 clicks per day would be considered a mid-tier term. Hence, building good revenue predictions for most mid and tail terms using only subscription data would be difficult if not impossible. Enter multiple metric optimization into the equation. If a marketer uses another richer dataset to predict subscriptions, he or she might be able to build better revenue predictions.

A richer dataset can be created by combining the registration and subscription data into a new metric which acts as a proxy metric for subscriptions. The question is, how to create this metric. Here are the steps:

Step 1: Identify the relationship between the metrics

A simple way to identify the relationship between the metrics is to apply linear regression to the metrics. For 2 metrics, as in this case, it requires plotting registrations vs subscriptions.

In this chart, the registration line tells us that on average 6.3% of registrations become subscriptions. In other words, if you were to use registrations as a proxy on the tail terms, you should consider a registration worth only 0.063 a subscription. The example in the table should make this clear

Thus, we estimate keyword 2 will do better than KW1 even though KW 2 does not have a subscription as of now.

Step 2: Create the proxy metric

The example above should make it evident that, in this case, the proxy metric will look like:

Proxy Metric = Subscriptions + 0.063*Registratons

Step 3: Build models to the proxy metric and set bids after optimization

Once the proxy metric is created, every keyword should be modeled with the proxy metric as the revenue metric. After this, optimize the keyword set to maximize the proxy metric.

In the interest of brevity and to keep it from becoming too mathematical, I have made several simplifications in this analysis. However, several key points should become readily apparent. First, the metrics and the weights came from sophisticated statistical analysis. Second, even this simple case is relatively mathematically involved. We only looked at two proxy metrics. For more metrics multivariate regression methods would have to be used. Lastly, even though the relative weights were determined, we did not discussed the bids that would actually optimize to both. It requires optimization algorithms as any manual method would be a half-baked heuristic.

So the question is do you really need to optimize to multiple metrics? If you have a simple business model with a short sales cycle (less than a day for over 80% of your transactions) and where you are maximizing one metric (say revenue) then the answer is usually no. Remember that you usually have one metric and several predictor metrics. If the dataset is rich enough, then its better to optimize to the real thing isn’t it?

However, when the sales cycle is long and you are using long tail keywords, its an option worth considering. Also, remember that just because you are tracking 50 metrics you shouldn’t optimize to all. More is not always better. Metric selection for optimization should be done systematically via regression modeling and keeping only as metrics needed for the estimator metric to make your predictions accurate and robust.

In summary, multiple metric optimization is a powerful optimization technique that can benefit many advertisers. Among Its key benefits are predicting the conversion/transactional metrics when you do not have enough data or predicting performance when sales cycles are long. This is especially useful for the long tail where data is sparse.

However, the technique is not for everyone and is to be used with caution using sophisticated algorithms. Developing a sound modeling methodology coupled with a good optimization engine is key to make this strategy successful.

In a recent Search Engine Land opinion piece entitled Automated Keyword Bidding? More Like Automated Money Sink, the author espouses the view that automated keyword bidding is flawed because it is impossible for agencies to scale their systems to deliver positive results to clients.

While I agree with most of the points the author has raised, I believe his conclusion is wrong. Blaming automation for poor bid management is the equivalent of blaming a Ferrari for a car crash if a monkey was driving it.

The real debate is really optimization vs. automation. Many people confuse the two and perhaps it needs clarification. Rules-based bidding is sub-optimal and not a scalable solution for clients.

Rules in computer science are known as “heuristics” and are defined as rules of thumb that do not guarantee the best solution. However, a bid optimization algorithm does guarantee the best solution subject to the constraints applied – i.e. a spending target or a CPA/ROI goal.

There is also the issue that keyword performance based on 7 day or 30 day increments can lead to poor campaign management decisions. I agree with the author on this point. The flaw is that the marketplace is very complex and doesn’t account for factors such as news, seasonality or events making it impossible for a human to correctly figure out what bid would result in specific CPCs , ROI and position.

Simple curve fitting (i.e. what a smart high school student would do) will simply not work. Bid performance profiles for keywords can be vastly different. Keyword models themselves need to be sophisticated and require knowledge of techniques such as time series forecasting. Crunching this large amount of data needs a huge database and computational firepower. An excel spreadsheet will not do. However, the flaw here is not of automation but that of building poor keyword models.

Instead of avoiding firms that espouse “proprietary technology”, it’s better to ask the right questions to make sure they truly understand optimization. Some important questions to ask include:

1. Is your proprietary technology a rules-based or a portfolio approach? A lot of agencies say they do portfolio-based bid management, but what they really do is simply cluster keywords. That is not a portfolio approach. Quiz them about the portfolio approach and ask them to explain their method.
2. Is the technology you’re being sold a campaign automation tool or a bid optimization platform? The difference is huge. A campaign automation tool might have all the bells and whistles to speed up labor intensive tasks. However, it will not have the technology to help you maximize your performance. A bid optimization platform will maximize your performance and provide you with tools to speed up your day-to-day tasks. To take the car analogy further: an automation part is like the body of the car while the optimization is the engine. A sleek car body is nice to look at, but when you are performance driven, what you really need is a powerful engine.
3. What is the accuracy of your keyword models? If they are doing model-based portfolio optimization, they should be able to give you a simulation of revenue/CPA at every spend level. At Efficient Frontier, we do this for all our clients and have the capability to automatically generate these in real-time. Once you managed with the portfolio approach, you should find the simulations to be within +-10% accuracy.
4. What do you do for sparse data keywords? If data is sparse, you cannot use a simple rules based approach. Make sure you ask them their approach to solve this issue.

It is not automation that should be blamed for poor campaign management. It’s the improper usage of modeling and tools that leads to sub-optimal results. So remember, don’t blame the Ferrari for the monkey crashing the car. Ask the right questions of the owner of the monkey.

Optimal bidding for performance is a complex task for search marketers. When placing bids every day, you must be mindful of the volatility in the marketplace due to changing search traffic, competition, matching algorithms that vary across search engines and the changing needs of your business. For many of you, particularly if you work in the travel or retail sectors, seasonality—the cyclical patterns in the demand for various product offerings—adds an additional layer of difficulty that can make search marketing campaign management seem daunting.

It would be extremely time and cost prohibitive to manually sift through historical data on all keywords used in a large campaign to identify statistically significant seasonal patterns. But this information can be critical to ensuring the success of a campaign. In the retail sector, for example, summer clothing doesn’t sell in winter and vice-versa. Hence, to ensure optimal performance, seasonal keywords must be continuously identified and bid upon to appropriate positions.

The solution to this problem is to build predictive keyword models that correctly estimate expected revenue and spend patterns, while taking seasonality into consideration. Ideally, you should do this within an automated framework that, in addition to analyzing search traffic patterns, also helps to determine seasonal keywords that have a high probability of improving the overall portfolio ROI, and prioritizes the learning of these keywords while keeping the advertising budget in check. This automated identification of revenue-generating seasonal keywords can be particularly valuable to large and medium sized advertisers, as they must often create predictive models on hundreds of thousands of keywords a day.

There are three steps that cover the basics of predictive modeling to enhance campaign success.

1. If you have a seasonal business, try to combine seasonal keywords in a campaign. This will make monitoring easy. For instance, an apparel retailer can have a ‘winter sweaters’ campaign and a ‘summer shorts’ campaign.

2. Ensure keywords reflect the seasonal nature of the product to best optimize for results.

3. If you cannot combine seasonal keywords into a campaign, then make sure they can be tracked. One way to do this is to label the keywords in your database.

Once you have a seasonal campaign established, it’s important to analyze the data and then take action based on your discoveries.

Look at historical data for the season you are interested in and find keywords that have experienced a sharp spike in ROI. Then account for keyword position. In general, the lower a keyword position, the higher its ROI. So if you find an uptick in ROI at a lower position, the spike is not necessarily due to seasonality. The opposite – i.e. keywords with a spike in ROI despite an increase in position – present a strong indicator of seasonality. Remember, seasonality refers to both upward and downward spikes.

If the historical data points to a sharp increase in a keywords’ ROI, promote the keyword by a bid increment. A 20 percent bid increment/decrement is a good starting point. You’ll want to track the performance every two-to-three days and increment/decrement the bids as needed.

In a large campaign that contains many seasonal keywords, manual intervention is likely impossible. In this case, look to an automation tool with logic/optimization built in.

It is no surprise that the recession has affected online consumer spending habits. In lean times, people are more apt to curb spending and seek out the best bargains. From a search perspective, it is now more important than ever for marketers to adapt their campaigns to these changes in consumer behavior to better optimize performance.

As consumers modify their spending patterns, their search behavior—the keywords and terms they use to locate products and services—also change. We recently culled over half a million search terms containing a certain modifier. For instance, queries containing the word “cheap” would include “cheap flight,” “cheap car rental” and so on. Next we aggregated the keywords into groups based on the modifier and measured their year-over-year changes in impression volume and revenue per click (i.e. the per click monetization change).

The conclusions were significant. In the financial vertical, keywords such as “credit,” “lending” and “mortgage” have seen a huge jump in impression volume, while queries containing “loans” have seen a drop. Interestingly while mortgage and credit monetize better now than last year, loans and lending related keywords are not performing as well.

In our experience, changes in traffic patterns and behavior are closely related to changes in the economic environment. An increased number of consumers are searching for financial information but are not necessarily converting on the clicks. This is because they are not qualified as they once were (due to more stringent lending criteria) or are simply looking for information and are not interested in buying a financial product.

In the travel vertical, “cheap” and “discount” related keywords have seen big jumps in impression volume and are monetizing better. The same is not true for “cruise” and “hotel” related keywords where the revenue-per-click is lower than it was last year.

At a more granular level, the “hotel” set includes several thousand combinations of hotel with location names, hotel brand names, etc. This indicates a shift in consumer thought patterns. People are less brand focused, and more value conscious. Clearly frugality is in.

In order to effectively prompt the new frugal consumer to action, you must recalibrate your search marketing approach. Here are a few strategies to consider:

• Employ thrift-related messaging to your ad copy, stressing value over brand.
• If your business operates in the travel space, consider promoting cheap and discount-related combinations instead of focusing on brand and location-related keyword combinations.
• Monitor federal announcements related to loan and credit oriented policies, as they will have a direct and immediate impact on the quantity and quality of your traffic.

Maintaining focus on ROI while seeking greater efficiencies in keyword marketplaces will provide significant opportunities to increase your market share in key categories and obtain additional, valuable traffic at discounted prices.