Benjamin Vigneron, Author at Search Engine Land News On Search Engines, Search Engine Optimization (SEO) & Search Engine Marketing (SEM) Wed, 23 Feb 2022 20:27:38 +0000 en-US hourly 1 https://wordpress.org/?v=5.9.2 How PPC Advertisers Can Best Leverage The Research Period This Holiday Season /leveraging-research-period-holiday-season-232689 Wed, 07 Oct 2015 17:41:54 +0000 /?p=232689 Columnist Benjamin Vigneron shares thoughts and data on how to pace your search marketing budget over the course of the holiday season.

The post How PPC Advertisers Can Best Leverage The Research Period This Holiday Season appeared first on Search Engine Land.

]]>
ss-holiday-shopping-laptop

Many advertisers will be looking into their historical daily or weekly online revenue numbers in order to come up with their online flighting strategy for this coming holiday season.

While that’s a good start, you want to make sure that when leveraging historical revenue data, you are taking into account the lag from click/impression to sale. Essentially, you want to build quality traffic for the Thanksgiving weekend, as opposed to spending a lot more than usual just on those days.

Here are a few thoughts on how can you can make this happen and the potential lift you can expect from better pacing your budget over the course of the holiday season.

1. Account For The Lag From Impression/Click To Online Sale

If you’re using AdWords as your primary source of truth, then you’re only capturing transaction date online revenue. The revenue numbers are associated with the days when those transactions occurred, as opposed to when the original impressions or clicks leading to those transactions occurred.

It might not be a big deal if most transactions occur within the first 24 days following an ad impression or click; however, in most cases, this lag is likely to be significant enough for you to adjust your online flighting strategy accordingly. And the revenue lift from better pacing your budget over the course of the quarter can be massive.

The example below shows daily transaction date vs. click date during last year’s holiday season for a major US retailer across their paid search and shopping campaigns. Comparing both trends shows that there are times when transaction date revenue is greater than click date revenue, and conversely, some other times click date revenue is greater.

The pre-Black-Friday weekend research period is often characterized by lower transaction date revenue because consumers are holding off on their purchases, while click date revenue is actually pretty high during that research time, since those impressions and clicks are having a strong influence on later transactions. This is precisely what we are trying to capture here.

Transaction date vs click date revenue chart

The research period (from 43 to 47) can be identified by higher click date revenue, while the decision period (from 48 to 51) is characterized by higher transaction date revenue.

In practice, either the technology you use (such as Adobe Media Optimizer) offers impression/click date reporting, or you can still delve into AdWords’ time lag analysis in Tools > Attribution > Time Lag.

Google AdWords time lag screen

Note that this lag may vary over time. From my experience, it tends to decrease during promotion periods and increase the rest of the year. In the example below, you can see how the time lag from last click to transaction varied over time last year.

Time Lag from last click, q4 2014 chart

While the average lag from last click to transaction was 10.8 days in 2014, the lag was longer in October, then much shorter in November, especially during weeks 48 (Black Friday) and 49 (Cyber Monday).

2. Adjust Your Flighting Strategy Accordingly

Whatever your methodology for putting together a forecast, you might want to compare your predictions if using transaction date vs. click date revenue. In my case, I estimated the optimal online flighting based on historical efficiency (that is, the ROAS, or return on ad spend) where we’d want spend less whenever the ROAS is expected to be lower than average, and more whenever the ROAS is expected to be higher than average.

More specifically, you can build predictive yield curves to determine future optimal spend targets based on the predicted elasticity between ad spend and revenue, as explained in a previous post.

The outcome of this analysis is that you would naturally want to invest more during the research period, now that you are better attributing online revenue to the original impressions and clicks that lead to this revenue.

Flighting Strategy chart

Looking at click-date revenue, we’d want to invest more during the research period (mid-October through end of November), now that we are attributing revenue to the original impressions and clicks which lead to those transactions.

Another way to look at this is to compare your share of budget going to a specific month:

Share of Budget chart

In this particular case, my optimal September budget should be 11% of my budget for the whole period (Sept. to Dec.) if using transaction date revenue, as opposed to 16% if using click date revenue. October, November and December would get 18%, 38%, and 32% if using transaction date revenue as opposed to 21%, 46% and 17% if using click date revenue. Essentially, we should move a significant chunk of the December budget to October and November.

As you can see, the budget shift from December to October and November can be significant. More importantly, when running the numbers for this particular advertiser, the expected revenue lift would be +11 percent simply by pacing the same amount of money in a smarter way through the holiday season — that is, by building quality traffic during the research period.

In a nutshell, online marketers should definitely move away from transaction date reporting and instead value impression/click date revenue reporting capabilities for the online initiatives. Better attributing revenue over time can unlock significant growth opportunities.

The post How PPC Advertisers Can Best Leverage The Research Period This Holiday Season appeared first on Search Engine Land.

]]>
ss-holiday-shopping-laptop Transaction date vs click date revenue Google AdWords time lag screen Time Lag from last click, q4 2014 Flighting Strategy Share of Budget
How Page Views, Time On Site & Bounce Rate Predict For Changes In Quality Score And Revenue /pageviews-time-site-bounce-rate-predict-changes-quality-score-revenue-227898 /pageviews-time-site-bounce-rate-predict-changes-quality-score-revenue-227898#respond Wed, 19 Aug 2015 15:55:11 +0000 /?p=227898 Columnist Benjamin Vigneron explains how you can use website analytics data to derive insights that you can apply to your paid search efforts.

The post How Page Views, Time On Site & Bounce Rate Predict For Changes In Quality Score And Revenue appeared first on Search Engine Land.

]]>
analytics-data-ss-1920

Over the years, Google’s Quality Score secret recipe has become more sophisticated and more challenging to optimize against.

From Google’s blurry definition, we know that Quality Score is mostly a function of your historical CTR (click-through rate), as well as the “quality of your landing page.” However, the landing page quality piece can be hard to quantify.

In this post, I will share a few findings about how engagement metrics (that is, page views, time on site and bounce rate) can be strong predictors for both your Quality Score and your revenue metrics.

Bounce Rate & Time On Site Found To Predict For Changes In Quality Score

When running a multiple linear regression analysis based on daily Quality Score, CTR, page views, time on site and bounce rate across millions of keywords, I found that three metrics out of four were correlated with Quality Score:

  1. The bounce rate was the strongest predictor for Quality Score, accounting for roughly 2.6 to 3.9 Quality Score points. A high bounce rate (that is, greater than ~40% in this particular case) pretty much guarantees a Quality Score lower than 7. (Note that, per Google, a low bounce rate does not guarantee a boost in Quality Score.)
  2. The CTR was the second-strongest predictor, accounting for 1.6 to 2.4 Quality Score points.
  3. The time spent on site accounted for 0.2 to 0.5 Quality Score points.
  4. Page views seemed to be a strong predictor, too; however, the data were not statistically significant for this particular data set (high p-value).

Essentially, if you want to stay away from high CPCs (cost-per-click) and low impression share due to a low Quality Score, you want to address those campaigns/keywords/product listing ads with above-average bounce rates, below-average CTRs and below-average time on sites, or any combination thereof.

Both the bounce rate and CTR are strong Quality Score predictors, then Time on Site

Both the bounce rate and CTR are strong Quality Score predictors, then Time on Site

Page Views, Time On Site & Bounce Rate Predict For Revenue Changes

While the Quality Score is a relevant metric to optimize against in order to minimize marketing costs, advertisers typically focus on the end revenue metrics. One of the main challenges often is to address revenue sparsity across thousands or millions of keywords, product listing ads, and so on — and that’s really when those engagement metrics come in handy, as they can help predict for revenue.

Indeed, from the data I collected, I found the following:

  1. Bounce rates were associated with 61 percent to 100 percent of the average RPC (revenue per click). While a low bounce rate does not guarantee a higher Quality Score, it does seem to guarantee more revenue, and vice versa.
  2. Page views were associated with 2.2% to 3.9% of the RPC.
  3. Time on site was associated with 0.4% to 0.7% of the RPC.

In short, if you haven’t collected enough revenue data across certain keywords, product listing ads, devices, times of day or locations, it is definitely worth using those engagement metrics as proxy metrics for future revenue.

Bounce Rate, Page Views, and Time on Site Are All Statistically Significant Revenue Predictors

Bounce Rate, Page Views, and Time on Site Are All Statistically Significant Revenue Predictors

In a nutshell, those site-side engagement metrics are a valuable source of information when it comes to both mitigating your marketing costs and enriching your data for making more informed decisions.

Now, you might want to compare those findings with what you are seeing in your own accounts, so feel free to share what you find!

The post How Page Views, Time On Site & Bounce Rate Predict For Changes In Quality Score And Revenue appeared first on Search Engine Land.

]]>
/pageviews-time-site-bounce-rate-predict-changes-quality-score-revenue-227898/feed 0 analytics-data-ss-1920 CTR and Engagement Metrics vs. Quality Score Both the bounce rate and CTR are strong Quality Score predictors, then Time on Site Engagement Metrics vs. Revenue Per Click Bounce Rate, Page Views, and Time on Site Are All Statistically Signiifcant Revenue Predictors
Mobile Searches Now The Lion’s Share: 5 Tips Not To Miss The Boat /mobile-now-lions-share-5-tips-not-miss-boat-225010 /mobile-now-lions-share-5-tips-not-miss-boat-225010#respond Tue, 14 Jul 2015 13:18:43 +0000 /?p=225010 The year of mobile has arrived -- and with it, the need to put as much into your mobile paid search campaigns as your desktop ones. Columnist Benjamin Vigneron shares his tips for evaluating and improving your mobile efforts for better paid search performance.

The post Mobile Searches Now The Lion’s Share: 5 Tips Not To Miss The Boat appeared first on Search Engine Land.

]]>
mobile-ppc-pay-per-click-tablet-ss-1920

How is it that mobile searches now exceed those from desktop, yet so many advertisers still spend so little on mobile?

This discrepancy is probably because some advertisers are still in the process of adjusting their websites and marketing campaigns for mobile. But perhaps more importantly, it is because they are still in the process of finding ways to better measure what mobile devices truly bring to the table.

Here are some steps you might want to take for a more successful mobile search engine marketing (SEM) effort.

1. Test Your Site For Mobile Friendliness

Mobile devices with full browsers have been around long enough for advertisers to pay attention to the mobile experience. If users don’t convert on mobile devices as much as you’d expect, it might be because of a poor user experience — common issues include confusing page layouts, lengthy loading times, discouraging “sign up” or “log in” pop-up windows, etc.

How can we make best use of a smaller screen without disrupting the user experience? Google recently created the Mobile-Friendly Test tool to help webmasters with this challenge — simply plug in a URL, and the tool will analyze it for mobile-friendliness and provide suggestions on where to improve.

Mobile-friendly test

Note that the tool does not analyze the entire domain but rather the specific web page URL you enter; thus, you might want to start with those landing pages which are used the most for your SEM program.

If your pages are not mobile-friendly per Google’s criteria, then you definitely need to address this first. Proper mobile website optimization is pretty much a prerequisite to creating a high-performing mobile SEM campaign. If you pass the test, you can move on to the items below.

2. Align Your Mobile Goals With Meaningful Mobile Actions

Easy to say, not necessarily easy to do. Essentially, you want to uncover what mobile searchers are truly trying to achieve and focus on delivering that, as opposed to focusing on mobile conversions only. That’s going to mean moving away from a last touch approach, as many mobile visits are only one piece of the puzzle, and often not a closing stage in the consumer journey.

In a nutshell, you want to able to quantify mobile actions — such as calls, store locators, product views, app downloads, etc. — and use those as a proxy for future online or offline revenue. This really is the key to understanding what mobile users do after they’ve used their mobile device (once they’re back home, or at work, or in your store). In practice, instead of making decisions based on mobile conversions, you’d want to give some credit to all sorts of relevant proxy metrics so your mobile goals reflect the true value of those mobile actions.

Mobile action examples

3. Leverage AdWords’ Estimated Conversion Data

In the same vein as using mobile-appropriate metrics, you can use AdWords’ estimated conversions in order to get a feel for indirect conversions — and that’s likely to benefit mobile devices the most. While you should keep in mind that those numbers are estimates only, they can be very useful, at least directionally. Consider the following metrics:

  • Conversions. Direct conversions occurring on mobile devices.
  • Conversions across devices. Cross-device conversions occur when a searcher clicks an ad on one device (e.g., smartphone) and later converts on a different device (e.g., laptop). This can be a very strong indicator as to what mobile actually does, as many advertisers realize that they’re getting few direct mobile conversions but tons of cross-device mobile conversions.
  • Calls. If using call extensions with Google forwarding numbers, you’ll get this additional metric. You can do the math to attribute some credit to those calls, such as x% of all calls lead to a sale.
  • Store visits. For eligible advertisers, Google helps track in-store visits, and a proxy for calculating those is [Est. total conversions]-[Est. cross-device conversions]-[Converted clicks]. Getting the estimated number of in-store visits is definitely a useful metric to optimize against, and looking at this number by device is likely to indicate that mobile traffic brings a greater proportion of in-store visits, such as in the example below, where 44% of the mobile revenue occurs in-store vs. 16% for desktop/tablet revenue. Those are actual numbers for a large U.S. retailer, and they have led us to adjust our mobile goals accordingly based on cross-device and in-store sales.

Revenue Breakdown Pies

4. Use Mobile-Specific Ads & Sitelinks

AdWords’ mobile-preferred ads and sitelinks were introduced together with AdWords’ enhanced campaigns. They definitely require some time to implement; however, they can really boost your CTR (click-through rate), as well as your conversion rate. A side benefit of having mobile-specific ads in place is that you can better analyze ad performance by device, as opposed to looking at combined performance across devices.

In my experience, there is no direct benefit to just duplicating your desktop ads and making them mobile-preferred. You need to tailor those ads for mobile users and make sure your ads are consistent with the on-the-go nature of mobile users. You can also use mobile-friendly display URLs such as “m.domain.com” to help boost your CTR or conversion rate on mobile devices specifically.

Note: On the Google Display Network (GDN), note that there are specific image ad sizes for mobile devices: 300 x 250, 320 x 480, and 480 x 320. More details here for all types of mobile ads, including other relevant mobile ads such as call-only ads.

With regard to mobile sitelinks specifically, these can be useful if you want to direct mobile users to specific pages (which should, of course, be mobile-friendly). A relevant use of mobile-preferred sitelinks would be a store locator type of page or an in-store coupon page.

5. Mobile Bid Adjustments & Trade-Offs Across Devices

This is one of my favorite parts because the technology I currently use does this extremely well. A couple of years ago, I would have been looking at the average performance on mobile vs. desktop/tablet and calculated bid adjustments accordingly, such as: mobile bid adjustment=(mobile ROAS)/(desktop/tablet ROAS)-1.

While this approach makes sense and will help your mobile ROAS converge towards your desktop/tablet ROAS, it is suboptimal, as it does not take into account the revenue trade-offs across devices. And at a time when mobile is accounting for the lion’s share of search queries, “suboptimal” is no longer good enough.

Essentially, you will get more revenue overall if tolerating different ROAS levels across devices, and you will bid more aggressively wherever the next marketing dollar brings the most revenue.

Here is an example for a given ad group. In this case, we could decrease the desktop/tablet base bid down to $2.10 and increase the mobile bid adjustment to in order to account for the fact that desktop/tablet marginal returns are lower than those from mobile. This would translate in more revenue overall.

Device revenue tradeoffs

Also, if you have physical stores and have noticed that mobile is bringing lots of in-store visits, then it might be worth setting more aggressive mobile bid adjustments around specific stores. In this case, you’d want to duplicate some of your top campaigns, target your physical stores, and set above-average mobile bid adjustments there.

Conclusion

Mobile SEM is not getting as much credit as it should. What is being tracked currently is mostly the tip of the iceberg, so it’ll take lots of intuition and workarounds to see the full picture.

Ultimately, as an analyst, I value intuition the most and do enjoy all those workarounds to some extent — but I need hard numbers to convince people around me, so why not run a proper A/B test to measure what mobile actually does? Turning off your mobile efforts in SEM for a couple of weeks is likely to be very insightful and perhaps the price to pay to collect new data points and additional knowledge, which will allow you to take your mobile program to the next level.

The post Mobile Searches Now The Lion’s Share: 5 Tips Not To Miss The Boat appeared first on Search Engine Land.

]]>
/mobile-now-lions-share-5-tips-not-miss-boat-225010/feed 0 mobile-ppc-pay-per-click-tablet-ss-1920 Mobile-friendly test Mobile action examples Revenue Breakdown Pies Device revenue tradeoffs
The Missing Link: 3 Steps For Connecting TV & SEM Performance /3-steps-connecting-tv-sem-performance-223342 /3-steps-connecting-tv-sem-performance-223342#respond Wed, 17 Jun 2015 16:39:51 +0000 /?p=223342 Columnist Benjamin Vigneron shares his method for attributing changes in SEM performance over time to multiple internal and external variables.

The post The Missing Link: 3 Steps For Connecting TV & SEM Performance appeared first on Search Engine Land.

]]>
video-tv3-ss-1920

Every marketer who thinks about their marketing mix holistically not only cares about each channel individually, but also how those channels perform in combination with each other.

While this can get tricky to measure accurately, I’ll share some basic techniques to connect online and offline data — and, more specifically, how marketers can measure the impact of TV and seasonality on their SEM efforts.

1. Pick Relevant Data

Ideally, you’d want to run a test on a significant sample of your audience and compare the results with the rest of your audience. Unfortunately, that is not always possible in real life.

For example, if you run TV ads nationally, you won’t be able to target a randomized sample of the population and compare the results with the rest of population, so you won’t be able to form nice and tidy test and control groups. Instead, you’ll have to analyze how much of an impact national TV has on your online initiatives over time.

Assuming our response variable is the weekly SEM impression volume we’re getting on a selection of branded search queries on Google and Bing, then our first variable would be how much was spent on TV ads over time. Note that seasonal trends may play a major role on general SEM performance and should pretty much always be taken into consideration when attributing changes in performance over time.

You essentially want to normalize the data based on seasonal trends — this will prevent you from attributing a change to TV ads when you were actually expecting more volume based on historical seasonal trends.

Similarly, budget changes — whether they are online (SEM, Social advertising, RTB, emailing, etc.) or offline (TV, radio, etc.) — can hugely impact performance over time and should definitely be factored in.

For the purposes of this article, I’ll keep it simple and focus on the following variables: national TV spend and seasonal trends. However, the logic would hold true for more variables, as long as those variables are independent of each other.

In this case, we’ll use the following input variables:

  • Response variable Y1: weekly branded SEM impressions
  • Input variable X1: weekly national TV spend in this case
  • Input variable X2: weekly Google Trends index on top non-branded queries, which supposedly reflect the market demand

2. Run A Contribution Analysis

The next step is to run a contribution analysis (more specifically a multiple linear regression analysis, in this case) so that we can predict our response variable (i.e., SEM branded search query volume) from two independent variables: TV ad spend and seasonal trends. For the sake of this post, let’s use some hard numbers and this downloadable spreadsheet: Actual vs. Modeled (.XLSX file). Say we have nineteen weeks of SEM and TV data, as well as Google Trends data.

(Note: We could use R, which is very well suited for this type of analysis. For the sake of this post, however, we’ll just use Excel, which is by far more widely used.)

Excel offers a “Data Analysis” package, which will well help run a multiple regression analysis. Step-by-step instructions are as follows:

  1. Load Excel’s analysis tool pack once for all — see Load the Analysis ToolPak for instructions.
  2. Launch the data analysis package via the “Data” tab

data-analysis-button

  1. Select “Regression” in the Analysis Tools box.

data-analysis-menu

  1. Select your response variable (“Input Y Range”), input variables (“Input X Range”), pick a cell where you want to output the results, such as $G$1, and click “OK.”

regression-settings

Looking at the regression summary, you’ll be looking for:

  • A high adjusted R squared — that is, a value greater than 0.6-0.8, which would indicate that 60-80% of your branded impressions can be attributed to the combination of TV spend and seasonality.

summary-output-11

  • Low p-values for each input variable. A p-value greater than 0.05 is not statistically significant (it might be due to random chance, rather than a finding).
  • Positive coefficients for each contributing variable. Negative coefficients may indeed occur as a result of the co-linearity of two input variables, which means that your input variables are correlated (for example, they happen at the same time) and the regression analysis is not able to distinguish the impact of those variables individually.

summary-output-2

3. Visualize Your Predictive Model

You can now compare your model against the observed data, and more specifically look into the contribution for each individual variable.

Observed-vs.-Modeled

In the present example, you’d be able to say that 39% of all branded impression can be attributed to TV, and 51% can be attributed to seasonality.

Variable-Contribution-Over-Time

Average-Variable-Contribution

Of course, this is the best case scenario where the data is particularly clean — in real life, you might need to first clean up the data (remove outliers, normalize the data further, add more input variables).

However, this technique can be very useful in order to get a first feel for connecting online and offline data, and more generally attributing changes in performance to multiple internal and external variables — then you can test your predictive model, see how accurate it is, and fine tune it over time.

The post The Missing Link: 3 Steps For Connecting TV & SEM Performance appeared first on Search Engine Land.

]]>
/3-steps-connecting-tv-sem-performance-223342/feed 0 video-tv3-ss-1920 data-analysis-button data-analysis-menu regression-settings summary-output-11 summary-output-2 Observed-vs.-Modeled Variable-Contribution-Over-Time Average-Variable-Contribution
3 Steps To Optimizing Local Paid Search /3-steps-optimizing-local-paid-search-213686 /3-steps-optimizing-local-paid-search-213686#respond Wed, 28 Jan 2015 15:00:55 +0000 /?p=213686 Columnist Benjamin Vigneron explains how to localize your messaging and optimize your budget across your best performing locations in Google Enhanced Campaigns.

The post 3 Steps To Optimizing Local Paid Search appeared first on Search Engine Land.

]]>
maps-local-search2-ss-1920

Whether online marketers are looking to localize their messaging or optimize their budget across best performing locations (or both combined), Google’s enhanced campaigns have made it possible to set location bid adjustments (LBAs) while still being able to target specific locations.

So what’s a good set-up for your program?

1. Determine An Efficient But Scalable Account Structure

In addition to any other considerations which may impact your paid search account structure such as breaking out branded vs. non-branded campaigns, or categorizing campaigns by product categories and/or user intent, search marketers should definitely put some thinking into how localized they want their paid search effort to be. More specifically, one can assess the following scenarios:

  • Localized campaigns, no location bid modifiers: Stick to the old-fashioned way of having campaigns organized by geo, such as campaigns targeting each individual US state, as well as a couple of campaigns targeting the top cities. This typically makes sense if advertisers need to bid on localized keywords, serve localized ads, and/or redirect to localized landing pages. This type of account structure allows for the most granularity but there are a few major trade-offs: the amount of time it takes to build then maintain, as well as data dilution across all those locations, making it more difficult to make bid and budget decisions based of statistically significant data.
  • Nationwide campaigns with location bid modifiers: for example, that would be one nation-wide campaign with 50 state-level bid modifiers. This makes sense if you don’t need localized keywords/ads/landing pages, and allows for improved efficiency across the board as the cost can be better allocated across locations – states in this case. Note that one can use multiple layers of location bid modifiers, from state to DMA to city to zip code levels and those won’t stack if overlapping – instead the most granular level will be used.

Localized-Campaigns-vs.-LBAs

  • Blend of localized and nationwide campaigns: for instance, that could be a couple of localized campaigns for those top locations, and nationwide campaigns with LBAs for the rest of the country.

Blend

The main benefit of the above hybrid structure is that marketers can get true localized search capabilities in those top locations, while keeping their paid search accounts manageable. Also, this type of structure allows for local testing – from local promotions to local mobile push and so on.

2. Get Started With Location Bid Adjustments

If you still haven’t got a chance to use those location bid adjustments, then you might want to consider putting together some cost and revenue geographic data from your favorite SEM platform and see whether there is room for improvement there. The odds are that there is a lot of room for improvement out there – the question is not whether there is any room for improvement or not, but how much, and is it worth the time spent optimizing local performance.

While estimating the time required for more localized campaigns can be discussed separately based on the overall business strategy and the above scenarios (localized campaigns vs. LBAs), the room for growth can be assessed by looking at the ROAS distribution across campaigns and locations, such as DMA-level ROAS for instance.

To simplify, if you are seeing the same ROAS across all DMAs, then there is no opportunity; however, if the local ROAS vary with locations, there is an opportunity. Essentially you want to see whether there are areas with a high ROAS where you’d want to bid up and get more revenue volume, and other areas with a low ROAS where you’d want to bid down and cut down cost.

Note that when measuring local ROAS, you might need to include other factors such as the impact of online advertising on offline sales, whether those occur over the phone, through an app, or in-store. It is obviously crucial to make bid and budget decisions based on true revenue numbers, as opposed to the short-term online returns you can see in AdWords, which might not reflect the true business value of your local traffic.

More specifically, once you have some relevant geographic data together, you can measure how ROAS are distributed by location, such as:

Local-ROAS-frequency1

In light red those are locations with significantly lower or higher ROAS compared with the mean, more specifically the ROAS is more than 1 standard deviation away from the mean, which is not exactly statistically significant but it definitely indicates there’s room for improvement. More specifically:

  • Locations with a ROAS more than 1 standard deviation away from the mean: those can be optimized as there’s a 68% chance that they are not performing as well as they could.
  • Locations with a ROAS more than 2 standard deviations away from the mean: those should definitely be optimized as there’s a 95% chance they are not performing as well as they could.

For calculating location bid modifiers, you can use a previous post about it – long story short you want to calculate those location bid modifiers for each campaign/location and use the campaign average performance for reference, so if a given campaign is targeting the US and want to apply state-level LBAs, you can use the following formula for California:

[California LBA]=[California ROAS]/[Average Campaign ROAS]-1.

In addition to this previous post, you might want to consider the distance from the mean and apply softer LBAs when the local ROAS is close to the mean, and more aggressive LBAs when the local ROAS is further away from the mean.

3. Analyze & Predict Local Ad Spend & Returns

After you have implemented location bid adjustments based on non-biased geographic data, that is without any LBAs in place, you’ll want to look into the LBA impact and what a good next step is. From my experience that’s where lots of marketers get stuck: most of them understand how to re-allocate their budget across locations based on historical geographic performance, however it can get tricky to measure geographic performance while accounting for LBA changes over time.

Ideally, you’d want to build cost and revenue models for each campaign/location pair, by LBA, where you are getting a sense of the cost and revenue change when applying different LBA values. In order to put this together, you can run a test across some campaigns/locations and look at the changes in clicks, average CPC, and conversions/revenue. In order to isolate the impact of LBAs from any other bid changes (base bids, mobile bid adjustments, etc..) you can set the intercept to zero assuming a LBA of 0% should not impact your test campaigns/locations:

LBA-vs-main-KPIs

In this example, using a linear regression which is far from perfect however a decent starting point:

Simplistic-LBA-cost-and-revenue-models

Then you can build those models for each campaign/location pair, maybe move away from linear models for something more sophisticated, then compare the marginal returns by location and update your LBAs accordingly. That might seem like a lot of work but that will be worth the effort at least for those top cost locations where the ROAS is significantly lower or higher than average – hope you’ll find this useful!

The post 3 Steps To Optimizing Local Paid Search appeared first on Search Engine Land.

]]>
/3-steps-optimizing-local-paid-search-213686/feed 0 maps-local-search2-ss-1920 Localized-Campaigns-vs.-LBAs Blend Local-ROAS-frequency1 LBA-vs-main-KPIs Simplistic-LBA-cost-and-revenue-models
Tackling AdWords’ New Default Close Variants Matching Behavior /tackling-adwords-new-default-close-variants-matching-behavior-202975 /tackling-adwords-new-default-close-variants-matching-behavior-202975#respond Wed, 10 Sep 2014 13:29:47 +0000 /?p=202975 Contributor Benjamin Vigneron explores the new default behavior and finds an efficient negative keyword strategy is more important than ever.

The post Tackling AdWords’ New Default Close Variants Matching Behavior appeared first on Search Engine Land.

]]>
google-adwords-yellow2-1920

There is no shortage of literature about ‘pure’ exact match type going away some time soon. Savvy search marketers love control, and this AdWords update clearly feels like a downgrade.

Back in May 2012, I was actually finding it beneficial be able to opt in to or out from Google AdWords’ close variants, depending on your preference: opt in to easily increase coverage at a fairly stable cost-per-click while keeping an eye on search queries, or opt out if you are seeing poor performance from close variants.

Whether you decided to opt in or out over the past two and a half years, you might be upset by no longer having the option to opt out, especially for those ambiguous keywords with irrelevant close variants you want to filter out.

Below are some thoughts about the potential effects, as well as suggestions for combating AdWords’ new default matching behavior.

The Conversion Chain At Stake

As shown below, the relationship between search query and keyword is the first step in the conversion chain.

In fact, not only is it the first step, it is also one of the levels where search marketers have historically had the most options available, from exact to phrase to broad match types, then with modified broad in 2010, allowing for a more sophisticated match type mix based on performance and scalability.

Conversion Chain

The relationship between search query and keyword is the first step in the conversion chain.

Furthermore, the search query to keyword relationship cannot be analyzed without considering what comes next: the associated ad copy and landing page.

If you lose control over the search queries associated with your exact keywords, you’re potentially also losing control over the ads that are served and the corresponding landing pages as a result.

Close Variants Matter More On Mobile Devices

I collected some data across multiple advertisers (mostly retailers), and I saw some interesting trends.

For those campaigns using exact and phrase close variants, those variants accounted for roughly 8.5% of the impression volume, with a remarkably higher proportion on mobile devices at 9.1%, most likely due to frequent typos.

Impressions by Match Type and Device

Impressions by Match Type and Device

So mobile impressions are likely to be impacted the most. Could this be another intentional attempt to push the industry into increasing mobile investment? Well, it sure looks like it.

Ambiguous And Branded Keywords To Be Most Impacted

Besides impressions, which indicate how Google matches queries to keywords, I also wanted to look into the average CPC. I found that:

  • Close variants may have similar or even lower CPCs. This could reflect lower competition on those close variants as not all advertisers are using close variants currently. That shouldn’t last long, though!
  • Close variants can have tremendously greater CPCs for branded keywords, which indicate that branded keywords’ close variants are… no longer branded!
Normalized CPCs

Normalized CPC for branded and non-branded keywords

Further Funneling Allows To Mitigate Close Variants’ Negative Effects

First of all — and you might have noticed this yourself — remember that negative keywords do not behave like positive keywords!

Exact positives and exact negatives are not associated with the same queries; same thing for phrase and broad. Essentially, positive matching will cover more queries than its negative counterpart, which means there is a gap — and this gap is getting bigger with close variants being the new default.

Search marketers can potentially leverage this gap by expanding their negative keyword lists in order to exclude irrelevant or poor-performing close variants even for exact-only campaigns or ad groups.

In general, the following campaign structure can be suggested to best tackle the new default matching behavior (note that a similar process can be applied at the ad group level):

Campaign Level Funneling Structure

Also, for ambiguous keywords where the singular and plural forms are significantly different (from a user intent and performance standpoint), you might want to have an ad group for the singular form where the plural form is excluded, and another ad group for the plural form where the singular form is excluded.

Setting that up is a tedious process, so you’ll definitely want to prioritize and only to do this for your top exact ambiguous or branded keywords.

Conclusion

Unless you had already opted in for close variants over the past two and a half years, you should expect some incremental impressions at higher CPCs in the near future, with a potential emphasis on mobile devices and branded or ambiguous keywords.

In all reality though, it is hard to quantify the potential effects as it all depends on your current match type mix, bids, and negative keywords. What’s for sure is that you should definitely continue to add more negative keywords across your campaigns and/or ad groups, even across exact ad groups and campaigns.

In a nutshell, an efficient negative keyword strategy is getting more crucial now than ever. Also, with close variants now matching for all phrase keywords, is modified broad still worth the effort?

The post Tackling AdWords’ New Default Close Variants Matching Behavior appeared first on Search Engine Land.

]]>
/tackling-adwords-new-default-close-variants-matching-behavior-202975/feed 0 google-adwords-yellow2-1920 Conversion-Chain The relationship between search query and keyword is the first step in the conversion chain. Impressions-by-Match-Type-and-Device Impressions by Match Type and Device Normalized-CPCs1 Normalized CPC for branded and non-branded keywords New-Campaign-Level-Funneling-Structure2
3 Opportunities With Those New Google Shopping Campaigns /3-opportunities-new-google-shopping-campaigns-194562 /3-opportunities-new-google-shopping-campaigns-194562#respond Fri, 27 Jun 2014 13:12:51 +0000 /?p=194562 For those who are not necessarily familiar with Google Shopping campaigns — let’s clarify things straight away. Product listing ads (PLAs) are remaining the same as an ad format for end users, whether search marketers are using “old PLAs” or new Shopping campaigns. The novelty lies only on the campaign management side of things, including […]

The post 3 Opportunities With Those New Google Shopping Campaigns appeared first on Search Engine Land.

]]>
For those who are not necessarily familiar with Google Shopping campaigns — let’s clarify things straight away.

Product listing ads (PLAs) are remaining the same as an ad format for end users, whether search marketers are using “old PLAs” or new Shopping campaigns.

The novelty lies only on the campaign management side of things, including a layering system based of the Google Merchant Center feed directly in AdWords. But in my mind, some of the new Google Shopping features should have a significant impact for both advertisers and end users – here are a couple of initial thoughts.

1. More Control Over Search Query To Product Mapping

Your PLA campaigns will get simplified only after reorganizing your PLAs within the new Shopping campaign world.

While Google will provide a tool to transition from old PLAs to new Shopping campaigns, you’ll definitely want to review the new structure carefully to make sure you are leveraging all new features coming along with those Shopping campaigns.

Besides the look and feel in the AdWords interface, there are mostly two new features that I would like to shine a light on.

First of all, the way those catch-all product targets behave is changing. As Google says:

If an original ad group doesn’t have an “All products” product target, the ad group in the new campaign will be created with an excluded product group for “Everything else.”

In other terms, the new “Everything else” product targets should not overlap specific product targets, which should help mitigate intra-account cannibalization. This is a big change.

(Click to enlarge.)

(Click to enlarge.)

Secondly, you are now able to set campaign priorities from to “Low” to “Medium” to “High,” which should theoretically force query mapping. The time when you had to bid up to receive impressions across specific product targets might very well be over — at least in theory.

Layered on top of an intelligent use of those new catch-all targets, priorities can potentially help address the overlap issue, although one should not need priorities as long as your Shopping campaigns are mutually exclusive.

Also, note that negative keywords for PLAs are not going away, fortunately. They should still be crucial in general, but you should no longer have to worry about cross-product competition, and you can instead concentrate more on regular search query filtering.

So altogether, online marketers are getting much more control in terms of what ads are being served, thanks to those features that help address the overlap issue.

2. Improved Product Mapping Involves A Shift From Head To Tail Products

Because of the above reasons, search marketers should be able to serve better ads to end users. More specifically, here are a couple of effects online marketers can anticipate:

  • Because of the mapping issue being addressed — at least theoretically — marketers should be able to serve more specific ads with more specific promotion lines related to more specific queries.
  • If you are using the new Shopping structure correctly, you should expect a shift from a couple of generic product targets and the current “All Products” product target to more granular, long-tail product targets and those “Everything else” product targets. This should help with both the average click through rates and conversion rates due to improved relevancy in general.
  • Since every layer of product targets needs to have an “Everything else” product target, you’re likely to get more impression volume if you didn’t already have a catch-all ad group. That could be a hit or a miss, and you’ll want to make sure your “Everything else” product targets are doing what they are supposed to, i.e., catching incremental queries, as opposed to stealing impressions from the rest of the account.

As a result, one can expect an overall increase in impression and click volume due to those new ‘Everything else’ targets. And from a marketplace standpoint, one can reasonably expect increased cost-per-click levels since more advertisers are likely to show their entire product catalog due to the ‘Everything else’ product targets being active by default across the board.

3. Bidding On PLAs Is becoming More Complex & Long-Tail Heavy

Because of the expected shift from a few head product targets to many specific product targets as described above, building data models for PLAs can also be anticipated to get more sophisticated in the near future. So, if the new Google Shopping features work the way they should:

  • There will be more product targets with some cost and revenue data, hence more bid decisions to be made overall.
  • More product targets will have a limited amount of data, making it more relevant to be able to borrow data from other product targets in order to use statistically significant data models across the board.

Long story short, there is definitely a challenge ahead of search marketers in order to best leverage increased PLA data granularity, and this is something we see as an opportunity at my company as we are used to building sophisticated data models in a complex auction-based environment.

Conclusion

What seemed to be mostly an AdWords interface update in the first place might turn out to have significant effects on the PLA environment. More consistent product mapping should improve overall ad relevancy and user experience, and potentially increase traffic quality, too.

But online marketers will only take advantage of those new Google Shopping campaigns if using the appropriate campaign structure and bidding solutions out there. To be continued…

The post 3 Opportunities With Those New Google Shopping Campaigns appeared first on Search Engine Land.

]]>
/3-opportunities-new-google-shopping-campaigns-194562/feed 0 Old-vs.-New-PLA-Campaign-Structure (Click to enlarge.)
5 Steps To Leverage Those AdWords Bid Simulations For Maximum Return /5-steps-leverage-adwords-bid-simulations-189373 /5-steps-leverage-adwords-bid-simulations-189373#comments Fri, 18 Apr 2014 13:47:32 +0000 /?p=189373 Though the basic concepts have remained the same, a lot has changed since Hal Varian released his bidding tutorial back in 2009 (video autoplay), especially with enhanced campaigns and those new mobile and location bid modifiers. One such change is that keyword-level simulations are now available for download in AdWords. For any search marketer interested in marginal […]

The post 5 Steps To Leverage Those AdWords Bid Simulations For Maximum Return appeared first on Search Engine Land.

]]>
Though the basic concepts have remained the same, a lot has changed since Hal Varian released his bidding tutorial back in 2009 (video autoplay), especially with enhanced campaigns and those new mobile and location bid modifiers.

One such change is that keyword-level simulations are now available for download in AdWords. For any search marketer interested in marginal cost and revenue numbers — essentially how diminishing returns affect your paid search program — those simulations can be a great source of knowledge. Below are some thoughts on how to leverage them.

1. Pull Those AdWords Keyword Bid Simulations

In AdWords, you want to add all columns from the “Bid simulator” section. Also, you might want to use a long enough time period so you can get a fairly significant conversion rate and average-order-value (AOV) numbers at the keyword level — we’ll use those later.

Bid-Simulator-Columns

For the sake of this exercise, I am using a 12-week-long report. Also, my AdWords account has conversion and revenue data, which is convenient — otherwise, I’d have to pull a report from a third-party tool, then plug it in here.

2. Put The Data Together In Excel

Essentially, we want to map those AdWords simulations, which provide weekly click and cost simulations, together with our average conversion rates and average order values (from AdWords or any third-party tool). Based on those additional click and cost estimates, we can calculate the estimated CPC, additional conversions, revenue and ROAS (Return-On-Ad-Spend) for all keywords.

This can be done using simple formulas, or calculated fields if you are more pivot-table adept. Either way, we are going to need the following calculations:

Conv. Rate = ‘Conversions/Clicks’

AOV = ‘Total conv. value/Conversions’

Those two calculations will be used in the simulations, assuming that both the conversion rates and AOV remain stable. Then, the below calculations will be used in conjunction with the conversion rate and AOV numbers:

Est. CPC (+50% bid) = ‘Est. add. cost/wk (+50% bid)’ / ‘Est. add. clicks/wk (+50% bid)’

Est. add. conversions (+50% bid) = ‘Est. add. clicks/wk (+50% bid)’*’Conv. rate’

Est. add. revenue (+50% bid) = ‘Est. add. conversions (+50% bid)’*AOV

Est. add. ROAS (+50% bid) = ‘Est. add. revenue (+50% bid)’/’Est. add. cost/wk (+50% bid)’

Etc…

The process is tedious the first time, so you might want to build a clean template that you can re-use in the future. Basically you should end up with the following type of pivot table, where you can see your current performance (12 weeks of data in this example), and AdWords simulations (always 1 week of data):

Est.-add.-calculated-fields

3. Determine Bid Simulations’ Coverage

While this is somehow optional, you might want to know how often Google is able to provide bid simulations, given that lots of keywords drive a very small amount of impressions and clicks.

In order to measure this, you can simply count the number of keywords with a simulation over the total number of keywords in your account. Interestingly, it seems those simulations are now available across most keywords, which is great.

Coverage

4. Aggregate Simulations & Understand High-Level Trends

For those keywords with available simulations, we can easily aggregate all those (-50% bid) simulations, (+50% bid) simulations, (+300% bid) simulations, and (top page bid) simulations, and visualize where the program would be, cost- and revenue-wise, if we followed AdWords’ suggestions.

Note that we still need to translate those weekly additional estimates into weekly total estimates. Those high-level numbers are interesting as they provide some hard numbers on the relationship between cost and revenue, as well as theoretical room for growth (+300% bid).

However, those should only be considered as directional, as it would obviously be sub-optimal to apply similar bid changes across all keywords.

weekly-add-and-totals

Then you should be able to visualize these high-level simulations type as a yield curve:

Yield-Curve

5. Delve Into Keyword-Level Marginal ROAS For Improved Insights

As noted earlier in this post, those aggregated simulations can only be directional, as every individual keyword should be analyzed separately, cost- and revenue-wise, in order to determine the optimal bid.

In the below example, I have sorted my keywords by descending “Est. add. ROAS (+50%)” — that is, those keywords with the greatest marginal ROAS out there if increasing the bids by 50%. This is where I want to invest more money and get incremental revenue from.

Note that the keywords with the greatest marginal ROAS are not necessarily those with the greatest average ROAS. For instance, keyword B has the 2nd highest marginal ROAS while it only has the 5th highest ROAS. As a result, those search marketers still using AdWords to manage their bids can now easily optimize those based on the marginal ROAS, as opposed to the old-fashioned average ROAS.

keyword-marginal-roas

Conclusion

Hopefully, this post will help search marketers take advantage of AdWords new bid simulator columns, or at least get familiar with basic marginal cost and revenue concepts. Keep in mind that, as Google says, those simulations “aren’t meant to serve as predictions or guarantees of future ad performance.”

Also, from my experience, those simulations are mostly reliable for non-branded exact keywords, not so much for phrase/broad keywords nor branded keywords in general. That being said, it is nice to see that Google is becoming more transparent and now provides better tools for search marketers to make more informed decisions.

The post 5 Steps To Leverage Those AdWords Bid Simulations For Maximum Return appeared first on Search Engine Land.

]]>
/5-steps-leverage-adwords-bid-simulations-189373/feed 4 Bid-Simulator-Columns Est.-add.-calculated-fields Coverage weekly-add-and-totals Yield-Curve keyword-marginal-roas
3 Tips To Fine-Tune Your Paid Search Program For The Holidays /3-tips-to-fine-tune-your-paid-search-program-for-the-holidays-175736 Fri, 01 Nov 2013 13:02:44 +0000 /?p=175736 With Black Friday and Cyber Monday around the corner, it is high time search marketers got their paid search program ready to scale up efficiently. In this column, I’ll outline and dive into three tips you should consider to get the most from the holidays in paid search. From my experience, most of the success […]

The post 3 Tips To Fine-Tune Your Paid Search Program For The Holidays appeared first on Search Engine Land.

]]>
With Black Friday and Cyber Monday around the corner, it is high time search marketers got their paid search program ready to scale up efficiently. In this column, I’ll outline and dive into three tips you should consider to get the most from the holidays in paid search.

holiday-stockings-300x114

Modified stock image used by permission from Shutterstock

From my experience, most of the success during that time of the year really comes from only three things: building out seasonal campaigns to best leverage incremental search queries; rotating holiday ads closely derived from your champion ads; and using day-parting to massively scale up and down during short time periods.

For more details, check out my full column on Marketing Land.

The post 3 Tips To Fine-Tune Your Paid Search Program For The Holidays appeared first on Search Engine Land.

]]>
holiday-stockings-300×114
How To Use The New AdWords Estimated Cross-Device Conversions /how-to-use-the-new-adwords-estimated-cross-device-conversions-173683 Mon, 07 Oct 2013 12:10:14 +0000 /?p=173683 A couple of months after the paid search world had to transition to AdWords Enhanced Campaigns in the name of simplified and more relevant cross-device ad management, we are finally getting some initial food for thought with regard to cross-device performance. I called it the “next frontier for online marketers” in a previous post, and we are now officially getting […]

The post How To Use The New AdWords Estimated Cross-Device Conversions appeared first on Search Engine Land.

]]>
A couple of months after the paid search world had to transition to AdWords Enhanced Campaigns in the name of simplified and more relevant cross-device ad management, we are finally getting some initial food for thought with regard to cross-device performance.

column-set-600x497I called it the “next frontier for online marketers” in a previous post, and we are now officially getting there!

In AdWords, a new column called “Est. total conv.” (aka Estimated Total Conversions) was made available for some beta accounts recently. 

How should marketers be looking at this newly-available data? Find out what I think in my latest Marketing Land column.

The post How To Use The New AdWords Estimated Cross-Device Conversions appeared first on Search Engine Land.

]]>
column-set-600×497