Although we at OptiMine are focused on keyword bid optimization, there are other elements you need to keep an eye on if you want to succeed in arena of paid search. One that ranks near the top of the list is Quality Score.

Search Engine Journal has published an article, “What Is Quality Score And Why Does It Matter?”, that clearly explains the concept, how it impacts paid-search advertising and what factors are included in the calculation.

In the simplest of terms, the primary determinant of your quality score is the relevance from keyword to ad, and from ad to landing page. A higher degree of relevance leads to a higher score:

Search engines evaluate a variety of factors when determining quality score, such as your keyword clickthrough rate (CTR) history, the quality of your landing page, the relevance of your text ad, and the relevance of the keywords you are bidding on. 

With a nod to Monty Python, don’t go all Sir Robin and think this is easy, because there is a level of complexity to it. But with some thoughtful planning, you can build a campaign with keywords, ads and landing pages that lead to quality scores to be envied.

@OptiMineInc

Tweet

This is the fourth and final excerpt from  “Open the Black Box”, OptiMine’s white paper about finding transparency and control in paid-search automation. You can see what you’ve missed by going here for part 1, here for part 2, and here for part 3. In previous installments we looked closely at the significant improvements you can realize in your paid-search programs by using an automation tool that also provides transparency and control. Today we conclude with the conclusion. 

The Bottom Line

Automation is crucial for large paid-search advertisers, but the type of automation matters greatly in terms of maximizing performance. Automated systems that do not (or cannot) treat each keyword uniquely and essentially as its own unique market will inevitably bid some keywords too high and others too low. This is inherently true for automated systems that use clustering, rules or simple modeling. The amount of money left on the table without the use of keyword-level modeling can be 25 percent or more of the true potential of a large paid-search program. For large advertisers, that can mean millions of dollars annually in lost revenue or excessive paid search costs.

The best automated systems provide both transparency into how individual keyword bid levels are set and the ability to override these bids through the user interface when human analysts with industry-specific domain knowledge know of opportunities that no set of software algorithms can possibly predict. It is only through automation, transparency and control that keywords can be treated as the individual markets they are, allowing the performance of each to contribute optimally to the success of the whole.

There you have it. Look at “Open the Black Box” as a primer for evaluating, or re-evaluating, automated paid-search systems. Think beyond automation and consider the importance of transparency and control. Is it important for you to see not only the results, but also the “why”? Is leveraging your team’s domain expertise critical to paid-search success? If the answer to either of these is “yes”, then simply automating won’t be enough. 

Tweet

Today we bring you the third of four excerpts from  “Open the Black Box”, OptiMine’s white paper about finding transparency and control in paid-search automation. If you’d like to catch up, you can go here for part 1, and here for part 2. Part 2 was about transparency and how having it adds context to the results your paid-search system is delivering. Today we take the natural next step. Understanding why a system acts is important, but, without control, you are there is little you can to do about it. That’s what today’s post takes a closer look at. 

Taking Control

Individual-keyword modeling can produce better performance than clustering, and in most cases, advertisers will be satisfied to let such systems work automatically. But at other times, they will want to take advantage of the keyword-level transparency built into such systems in order to take more direct control based on factors that the software can’t predict. This is why transparency and control are two sides of the same coin, so to speak. Transparency is the ability to see and understand what the software is doing. Control is the ability to apply domain expertise in response to dynamically changing business and market conditions that the software can neither predict nor respond to quickly enough. The best software automation systems not only provide transparency but also anticipate what kind of control analysts will need, giving them maximum flexibility to respond to changing conditions as frequently as
necessary.

A classic situation is inventory reduction. For example, a retailer has an excess supply of widescreen TVs and plans a big sale for the coming week. The promotion will of course affect clicks, conversion rates, sales, revenue, profitability and other metrics. The software will factor in these changes eventually, but it is likely to react over a number of days rather than immediately at the start of the sale – especially if it weights bid levels based on a seven-day moving average – missing revenue during much of the sale period and wasting money on excessive bids after the promotion is finished. The advertiser knows from experience that during sales of this magnitude, value per click changes by a certain percentage, so bids should be raised by the same percentage during the sale to capture more traffic without sacrificing efficiency or other campaign constraints. The advertiser also knows that after the sale, the software will adjust bids based on data inflated by the sale effect, and so requires the ability to push the bid levels generated by the software down a fraction in order to stay on target.

The ability to take control and “drive” the software is essential for many online retailers to account for short term discontinuities such as special offers and fluctuations in stock levels.

Transparency and control are crucial for achieving paid-search success. Without them, having an automated system is nice, but not much more. The fourth excerpt will conclude this series with a bottom-line look at automation, transparency and control, and how, together, all three can drive significantly improved financial results for your paid-search programs. If you don’t want to wait for excerpt 4, you can download the white paper at OptiMine.com now. 

@OptiMineInc

Tweet

Welcome to the second of four excerpts from  “Open the Black Box”, OptiMine’s white paper about finding transparency and control in paid-search automation. If you’d like to catch up, you can read the first excerpt here. Today we look at context which, in this case, means it’s not enough to know the results your paid search system is producing if you can’t do know the why behind them. 

Seeing the Context

As noted, bid automation is essential for companies with complex paid-search programs, but with black-box bid automation systems (that is, most of automation systems), it’s hard to determine the “why” behind their paid-search performance. The performance might be acceptable, but there’s no way of knowing if it could be better.

The problem is especially acute with cluster- and rules-based automation systems, where individual keyword performance is lost because groups of keywords are treated the same. Does your system allow you to filter and sort individual keywords based on highest or lowest cost, sales, revenue, profitability (and any other metrics you can think of), and drill down into the factors – such as day-of-week effect, or days until the next holiday – to determine the variables that are driving those individual keyword results? Some systems do provide bid values and results for individual keywords but limit access to the variables that were used to drive those results.

This absence of context might be acceptable to some advertisers, because set-and-forget automation is superior to manual bid optimization, but it will be inadequate for advertisers looking for the significant performance improvements attainable with systems
that provide detailed and filterable data on individual keywords.

How do these systems achieve such results? Consider the case of two keywords: “silver amethyst rings” and “sterling silver amethyst rings.” Most automation systems would cluster such similar-sounding keywords and calculate an average bid amount based on the value per click for the cluster as a whole. Other systems apply the same rule to each, potentially masking unique performance attributes. A system that provides transparency into individual keyword performance might reveal these strategies to be sub-optimal and identify very different driving forces that make these keywords convert. One might be more influenced by a particular day-of-week effect, while the other is weighted more heavily on a one-week average cost per click. The reasons for these differences might be obscure or even unknowable, but the differences are nevertheless very real.

Software automation systems that employ advanced multivariate regression analysis modeling at the individual keyword level will pick up such differences and bid each keyword appropriately, based on a historical analysis of the factors that drive individual keyword performance. This feature is especially important in managing mid-tail and extreme-tail keywords, which often account for as much as 90 percent of a campaign or ad group keyword portfolio, to maximize whatever revenue is available in the tail – revenue that is typically lost with clustering or general rules.

With better transparency giving providing context, control is the next step and, coincidentally, the next post in the series. If you don’t want to wait for excerpt 3, you can download the white paper at OptiMine.com now. 

@OptiMineInc

Tweet

Over the next few posts I’ll be publishing excerpts from “Open the Black Box”, OptiMine’s white paper about finding transparency and control in paid-search automation. People who mange large, complex paid-search campaigns often do so with the assistance of a tools that automate some, if not all of the processes. But automation does not guarantee a higher level of transparency and control. In fact, most automated systems are severely lacking in both areas. “Open the Black Box” shines a light on the importance of automation, transparency and control, and how to find a solution that offers all three. 

Background

For online advertisers running large, complex paidsearch campaigns, automation is virtually essential for optimizing keyword bids on campaigns and ad
groups that might involve hundreds of thousands or even millions of keywords. Optimizing such campaigns manually simply outstrips the ability and capacity of
human analysis. Bid automation software typically employs highly complex mathematical algorithms that can analyze millions of transactions per second, producing scenarios for the best cost per click (CPC), return on ad spend (ROAS), revenue goal, profit margin, or other campaign goal or business objective. The best automation systems help advertisers optimize the 10 percent to 15 percent of head terms that drive most of their sales or revenues while improving performance from mid-tail and extreme-tail terms.

But most automation systems for paid-search advertising have limitations. For one, most are basically “set-and-forget” black-box systems. They
don’t provide business analysts or marketers insights into why the software bids the way it does. They do provide some control over campaign or ad group
setup variables, but once these parameters are set, the process by which the software arrives at the bids remains a mystery to the advertiser. If these systems
cluster keywords and bid on clusters as whole, any distinctions in performance among keywords within the cluster will be invisible. This means advertisers
will inevitably spend too much on some keywords and not enough on others – particularly extremetail terms, which often have the greatest potential
for performance gains if the optimal bid can be determined.

This lack of transparency leads to the second significant limitation of most bid optimization white paper: Transparency And Control systems: There’s no functionality to enable human analysts to alter or override what the software does. Software automation isn’t a replacement for humans applying their experience and domain knowledge to a paid-search campaign at a particular time. Often, analysts are aware of an upcoming event or market shift that will change the performance of a keyword or
keyword cluster in ways that no algorithm can predict or react to quickly enough. At other times, businesses need to push inventory or manage fluctuations in stock levels through short-term offers. This requires transparency with regard to how bids are determined and the ability to override what the software is doing, based on both industry-specific experience and short-term business needs. As an analogy, consider the various computers in your car and how they monitor and regulate dozens of metrics related to fuel consumption, braking systems, etc. They are technological marvels, but someone still has to drive. The same is true in bid automation software.
As we unwrap this white paper we’ll detail:

• The need for transparency and control in bid,automation systems;
• How transparency and control can only come through automation,
• but why automation alone isn’t enough to guarantee them; and
• How to find a solution that strikes the right balance among all three elements

In the next post, we’ll put the need for transparency and control into the context of paid-search campaigns, so you can begin visualizing the benefits of having more of both. If you don’t want to wait for excerpt 2, you can download “Open the Black Box” at OptiMine.com now. 

@OptiMineInc

Tweet

A study commissioned by the IAB (Interactive Advertising Bureau) and conducted by Forrester Research shows that “advanced techniques for attributing value to digital media channels, based on specific campaign goals and detailed data analysis are replacing simplistic first and last click measures.”

“Digital Marketing Comes of Age” also reveals several trends that are pushing advertisers to use more sophisticated attribution schemes to determine publisher compensation. Of the five outlined in the IAB press release, one is near and dear to those of us at OptiMine: Optimization.

From the article:

Optimization will become more tightly integrated with media buying and execution: As the industry gets better at attribution and optimization, developing the right checks and balances, many elements of a media plan will likely be executed in a much more automated and efficient way. (Underlining is mine)

Pricing is, of course, an important element of any media plan and in the world of digital marketing – paid search, display, social – it is an element that is complex and challenging.

More from the article:

A related Forrester survey of interactive marketers* featured in “Digital Attribution Comes Of Age” took a snapshot of the landscape and finds that 34 percent of respondents currently use a rules-based approach, followed by 30 percent relying on a first or last click method. In addition, 11 percent said that they use algorithmic attribution models. 

As attribution grows in importance, marketers will continue to demand tools and techniques that are increasingly sophisticated and capable helping them achieve their their financial goals.  As with bid optimization in paid search, multivariate modeling done at the most granular level will ultimately provide the best improvement in financial results.

@OptiMineInc

Tweet

This is the fifth and final excerpt from “Bid Testing Best Practices”, a white paper from OptiMine Software. In this post, Jason looks at two specific applications of value-based bid testing using the same data. You can read earlier excerpts about position-based bid testing and current bid-based bid testing, or visit optimine.com to download the whitepaper.

The best approach is value-based bid testing: basing test bids on the present value per click and making small changes up or down to see how they impact campaign goals. This strategy takes time due to the need to test a small bid change, observe the results, and repeat the process until you find the bid that ideally matches your campaign goals. However, it is time well spent because it avoids wild swings in click volume and maintains control over the budget. How much of a nudge to give bids depends on how much risk can be absorbed at any given time, balanced against whether the incremental cost incurred is outweighed by the incremental gain.

The question of whether to test up or down depends on the campaign goal. If the goal is to maximize conversions, test bids should be incrementally increased and then evaluated against your goal. If the goal is to maximize efficiency, test downward to see if the same results can be obtained at a lower cost. Testing downward is an often overlooked but extremely important bid test strategy. For example, if a bid of $0.85 meets the goal on average as well as a bid of $1 per click and the lower bid is applied across thousands of keywords, the cost savings quickly become substantial.

Below are two interpretations of the same hypothetical data, illustrating various outcomes of changing bid levels incrementally based on your goals.

Goal 1: Maximizing Conversions

If the goal is maximizing conversions, for example, bids should be increased to raise the ad position. In this example, bid testing shows the higher the bid, the higher the conversions. And even though the ad already occupies a high position, incrementally testing the bid upward might result in additional modest gains in conversions. Something to note is that profit falls as bids are increased, which might be an acceptable trade-off, given the goal of maximizing conversions. This is bid testing in its simplest form, but there is far more to it than this.

Goal 2: Maximize Profit

If we change to goal to maximizing profit, the ideal bid changes. In this case, testing shows that bidding lower accomplishes the objective. Fewer conversions are generated than with higher bids, and we are in a lower average position – but the cost per conversion is lower, and consequently the profit is higher. This example shows that, while bidding upward to achieve a higher position generates the greatest number of conversions, it comes at the cost of reducing profitability.

The Bottom Line on Bid Testing

Internal changes to financial goals combined with the constant changes of a dynamic marketplace make bid testing necessary to keeping a paid-search program competitive and profitable. Choosing the proper bid testing approach is as important to paid-search success as the choice of keyword bid optimization method. Of the three methods explored above, only value-based testing maximizes performance improvement while minimizing financial risk. In contrast to its bid-testing cousins, the goal of valuebased is to find the maximum value of a keyword, regardless of current bid or position.

There are several compelling advantages to testing small incremental changes and testing frequently. Small changes minimize risk because the potential impact on budgets is small and controllable. Moreover, because they represent relatively small potential risk to budgets and/or performance, small changes can be tested with greater frequency. Frequency is an important component of bid testing because it delivers much greater long-term benefits than sporadic testing. Frequent testing is also important for companies in highly competitive environments and companies whose products are highly seasonal. It also allows you to more quickly and proactively identify marketplace changes rather than simply react to drops in PPC performance with dramatic and often expensive bid changes.

In determining what to test, the simple answer is that every aspect of a PPC campaign should be tested to make consistent improvements in performance. But if you’re not testing now, a good place to start is determining which aspect of your campaign has the potential for the greatest and quickest returns – either in terms of improving important performance metrics or increasing the effectiveness of your ad spend. Once you’ve tested and optimized this low-hanging fruit, create additional and perhaps more difficult goals and start testing again.

Implementing such a testing regime can seem like a daunting prospect for companies that manage tens of thousands or millions of keywords. Fortunately, there are applications that can help automate the testing process. However, any tool you choose should allow you to control the frequency with which tests occur and the acceptable range in which you are willing to allow bids to be adjusted – either in terms of a fixed dollar increment or in terms of a percentage. The best tools allow advertisers to spread out the testing over a long period of time to gather a variety of performance observations at different bid levels, all while protecting budgets and profit margins to the maximum extent possible.

@OptiMineInc

Tweet

The following is the fourth in a series of excerpts from “Bid Testing Best Practices”, a white paper from OptiMine. In it author Jason Mulvey explores the three primary methods for conducting bid testing in paid search.  In earlier posts we covered position-based bid testing and current bid-based bid testing. In this post, Jason looks at value-based bid testing and how it differs from its two cousins. Visit optimine.com to download the whitepaper.

Value-Based Bid Testing

Value-based bid testing differs significantly from its two cousins. Whereas position-based and current bid-based approaches each test around a point that has no relation to actual keyword performance, value-based bid testing uses keyword performance as its guide. Regardless of position or bid, value-based testing seeks to test around the observed value of a keyword.

Properly implemented, value-based bid testing reduces the risk of loss while maximizing the probability of improved performance. In practice, value-based testing works by moving up and down around a bid that matches the keyword’s average value per click. For example: If a keyword receives 100 clicks and returns $100 in revenue, the average value per click is $1. Value-based bid testing will use that average value as a starting point, testing up and down in a reasonable range and monitoring keyword performance.

Where value-based testing becomes difficult is in estimating the value per click for every keyword. Not every keyword receives a click and, of those that do, there is always a percentage that does not experience a conversion event. In cases where there is no conversion or value data, predictive modeling is required to estimate the value. While difficult, estimating the value per click can provide immediate improvements to performance, especially in face of changing goals and external conditions.

By nature, the appropriate predictive models used for bid testing are exploratory. Traditional predictive models do not provide estimates outside the range of values that have been seen in the past. Therefore, if a keyword has never had a bid as high as $2, a predictive model should not predict a $2 bid. Exploratory models have that restriction removed, allowing them to test in new ranges of bids that have not been tried before. The disadvantage to this exploration, however, is that the predictive accuracy of the performance is greatly decreased because there is no keyword performance data for the bids being tested.

Next up, Jason will look at value-based bid testing in practice and how the same data can result is very different results depending on the goal. If you don’t want to wait, you can have the white paper in its entirety by downloading it today.

Tweet

The following is an excerpt from “Bid Testing Best Practices”, a white paper from OptiMine. In it author Jason Mulvey explores the three primary methods for conducting bid testing in paid search.  In the previous post, Jason looked at position-based bid testing. Today he covers the pros and cons of current bid-based bid testing. Visit optimine.com to download the whitepaper.

Current Bid-Based Bid Testing

As the method states, the baseline used for testing is the current bid, whatever that happens to be. This is the most conservative of the three strategies and is executed by varying bids a small amount above and below the current bid. In contrast to position-based bid testing, the risk of loss is greatly diminished through current bid-based testing. However, where the conservative approach reduces risk it also decreases the probability of driving dramatic financial performance improvements. If the current bid is $0.50 and the sweet spot is $2.00, even doubling the current bid will leave the keyword well short of its optimal bid, missing potential revenue.

In the next post, Jason will look at the value-based bid testing. In the meantime, you can have the white paper in its entirety by downloading it today. 

@OptiMineInc

Tweet

The following is an excerpt from “Bid Testing Best Practices”, a white paper from OptiMine. In it author Jason Mulvey explores the three primary methods for conducting bid testing in paid search.  In this post, Jason highlights position-based bid testing. Visit optimine.com to download the whitepaper.

Position-Based Bid Testing

Historically, position-based bid testing has been one of the most popular methods because of the questions it answers: What do I have to pay for position X, and does the conversion rate or value per click change at different positions? Another reason for its popularity is the ease of execution: increase bids until the desired position is reached, and observe the changes in conversion and value data.

Although popular, position-based bid testing can be very expensive, especially if the positions you are testing for are well above the value per click of the keyword being tested. In such cases, you are likely to drive up the number of clicks without a corresponding increase in conversions, leaving you with significantly higher costs and little, if any, incremental revenue.

A new “modern” variation on position-based bid testing is the use of the Google first-page bid estimate. In this case, Google will estimate the bid necessary for an ad to appear on page 1 of a search matching the specific keyword. The danger, again, is in bidding a keyword up beyond its value and incurring costs far beyond its revenue potential. As tempting as the Google bid estimate is, with the potential to drive more conversions and revenue, conversions and revenue don’t drive profit. If a keyword is worth $0.50 per click on average, bidding it up to page 1 for $5 per click may result in real revenue increases, but the financial losses incurred will be just as real. Simply put, the chances of finding a golden keyword with 10 times the average conversion rate are as good as the chances of winning
a lottery.

In the next post, Jason will look at the current bid-based method of bid testing. In the meantime, you can have the white paper in its entirety by downloading it today. 

@OptiMineInc

Tweet