Attribution Digital Marketing

How to avoid getting Screwed by your Remarketing Platform

Many popular Remarketing Platforms purposely report misleading metrics. Their goal is to fool advertisers into increasing ad spend. Here are the tactics they use and the ways you can protect yourself.

Common Deception Tactics

View Through Conversions

  • A View Through Conversion is when a user sees your ad, DOES NOT CLICK on the ad, and later makes a purchase.
  • A Click Through Conversion is when a user sees your ad, CLICKS on the ad, and later makes a purchase.

For a sales driven remarketing campaign you should exclusively focus on Click-Through Conversions and ignore View-Through Conversions. The incremental sales generated by a simple ad display in a remarketing campaign is almost zero.

Remember that we are working with a Remarketing campaign. We are targeting customers that are aware of your brand. Customers who have already visited your site. In many cases, customers that have added a product to their cart and are very close to purchase.

Consider this scenario: A long time customer visits your site, quickly finds the product he is looking for and starts the checkout process. In the middle of checkout, he receives an e-mail from a friend. He opens the e-mail and clicks on a link. That link takes him to a news site to read an article. Since he is in your remarketing audience, that news site shows him an ad and records a “View”. The customer finishes reading the article and returns to your site to complete his purchase. Your remarketing campaign has now taken credit for that sale because an ad was displayed to this customer.

If your customer visits any webpage during the purchase process, or if they have another browser tab open, they will likely be exposed to an ad view. That ad view will get View-Through credit for that conversion.

The number of View Through Conversions recorded in a remarketing campaign can be more than ten times that of Click Through Conversions. It is no wonder that remarketing platforms and ad agencies are eager to include View Through Conversions in ROI calculations. If you include View Through Conversions you will see fantastic performance results. This will cause you to double or triple your budget. It’s easy money for them.

Here is what a fictional remarketing campaign with a typical distribution of view vs click conversions might look like:

 Ad Spend Click Conversions Click Sales View Conversions View Sales
 $12,000 50  $10,000  400  $80,000

If we use Click Conversions to calculate performance, we lose $0.17 for every $1.00 in ad spend. We have a negative ROI. The campaign is not profitable. We need to DECREASE our bids and budget.

 Ad Spend Conversions Sales CPA ROAS ROI
 $12,000 50  $10,000  $240  0.83 -16%

If we add View-Through Conversions to our calculation, we come to a totally different conclusion: The campaign is now performing fantastically well. We profit $6.50 for every $1.00 in ad spend. That’s a 650% ROI! The campaign is super profitable. We should INCREASE our bids and budgets. This is exactly what your remarketing platform wants you to do, but is exactly the OPPOSITE of what you should do.

 Ad Spend Conversions Sales CPA ROAS ROI
 $12,000 450  $90,000  $26.60  7.5 650%

The (unfortunate) reality is that the 400 view-through sales above were not a result of your remarketing campaign. The ad display is simply poaching attribution from your other channels.

Note: View Conversion metrics are valuable and useful in many scenarios. However, in the context of a sales driven remarketing campaign they should be excluded.

Cookie Bombing and Fake Impressions

Compounding the problem of View Through Conversions are fraudulent ad impressions and cookie bombing.

Over 50% of Display Ads never actually appear on the user’s screen. They are fraudulently served in a hidden iframe, or are stacked underneath other ads, or are compressed into a 1×1 pixel image, etc… These hidden fraudulent ad impressions still record a “View” in the user’s remarketing cookie. When that user eventually purchases on your site, that invisible fraudulent ad gets credit for the purchase.

If a particular ad network’s website can generate many conversions for advertisers, then those advertisers will want to spend more with that website. Fraudulent sites engage in “cookie bombing” to maximize the number of conversions they can “capture”. They serve as many ads to as many unique users as possible, so that as many users as possible have their cookie. The more users have their cookies, the more conversions will be credited to their sites, and the more ad budget money they will be paid.

To be clear, these sites are not actually serving any visible ads to any users. They are generating fraudulent ad impressions to set cookies on as many users as possible in order to “poach” conversion attribution (and budget) from your other channels. If a user with their cookie ever makes a purchase in the future, that site will get a view-attribution for that purchase, causing more ad budget money to flow to them.

All these hidden impressions distributed to as many users as possible cause View Through Conversion metrics to skyrocket. As mentioned before, View Through Conversions reported by an ad platform can easily be more than 10 times the number of Click Through Conversion reported. It is no surprise that ad platforms want you to believe in and count View Through Conversions.

New Attribution Models: View Throughs Disguised as Click Throughs

The simple way for an ad platform to fool you into counting View Through Conversions is to simply have a single “Conversions” column and lump all conversions in there. The platforms are however getting more sophisticated at transforming view conversions into click conversions, using new attribution models as a guise:

  • YouTube TrueView Conversion is any conversion that occurs after the user has watched 30 seconds of a video. Even if no click happened. TrueView conversions get counted in the Conversion column in AdWords along with the regular Click Through conversions. This makes it seem like an actual click happened, when none did. What makes this even more confusing and misleading, is that YouTube will still report View Conversions in a separate column. Since there are two separate columns, one tracking View Throughs explicitly, you will assume that the Conversions column must only contain clicks. This is not the case, and if not careful, you will significantly overspend on your YouTube remarketing campaign.
  • Steelhouse’s “Verified Visit” model counts any View-Through visit to your site within 1 hour of seeing an ad as a “Verified Visit”. Any conversion that happens within 30 days of this “Verified Visit” is counted as a Click Conversion (even though, no click ever occurred). Steelhouse will skirt the View Through question by saying that they only count conversions from “Verified Visits”. This will mislead you to think that they only count Click Through Conversions. But that is incorrect. The stats they report definitely include View conversions, which inflate your ROI, and cause you to overspend.

I spent months believing that SteelHouse’s numbers were based solely on clicks and post-click conversions. SteelHouse’s conduct made it appear to perform better than it actually was performing. Toms made spending allocation decisions and marketing vendor choices based on SteelHouse’s inflated performance.

Anna Hordov
former Online Demand Generation Manager, Toms

How to protect yourself from View Through Conversion “fraud”?

  • Insist that your platform report on Click Through and View Through Conversions in two separate columns. It is OK to include View Conversions in a report if they are in a separate column. It is NOT OK to combine View Conversions with Click Conversions in the same column.
  • Validate all data with your site analytics. Your analytics platform should report the actual accurate click through conversion data. Always compare the reports your remarketing platform generates with your own analytics data.
  • Don’t run Conversion focused YouTube remarketing campaigns. There is currently no way to opt-out of the TrueView model, therefor your conversion column will be overstated. Read more about TrueView…
  • Don’t use Steelhouse. Their Verified Visit model is designed to fool you. They also use their tracking pixel to inject data into your analytics. Scary stuff. Avoid.
  • Don’t use Criteo. They won’t tell you where your ads are displayed and their network has much more fraudulent sites participating in cookie bombing than anyone else. Avoid.

Fraudulent Automated Clicks

Fake clicks are an evolution of Cookie Bombing, except in this case clicks are auto generated onto the invisible ads, sending invisible visits to your site, and claiming Click credit for your purchases.

How Fake Clicks Work

  1. A legitimate customer visits your site but does not purchase. He is now placed into your remarketing list.
  2. This customer now visits another site on the internet. That site serves up an invisible ad and automatically “clicks” on that ad.
  3. That click causes an invisible visit to your website, by that customer. That customer’s remarketing cookie records a click and a site visit.
  4. Your legitimate customer now has a remarketing cookie set that says he saw an ad, clicked on that ad, and visited your website from that ad.
  5. When that customer comes back to your site to make a purchases, that ad will take credit for the conversion.

There are two issues here: The first is that you’ve paid for a fraudulent click. The second more important issue is that a conversion has been attributed to that fraudulent click. As a result, your remarketing platform appears to be performing better, meaning you will likely increase your budget in order to buy more of those fraudulent clicks.

Obfuscation and Lack of transparency

Anytime the remarketing network is hiding data from you, that’s an opportunity for them to fool you. For example, Criteo refuses to be transparent with the list of publisher sites where you ad appears and where your clicks and conversions come from.

3.6% of Criteo’s ‘users’ generate 25% of its clicks. Such behavior by real human users is highly unlikely. This behavior is indicative of adware, bots, click farms, or other code created by Criteo or its affiliates to generate clicks and drive up Criteo’s click-count numbers.


How to protect yourself?

  • Insist on detailed and transparent reports on where your ads were served, and where the clicks came from. If certain sites seem to have unusually high click and conversion rates, dig deeper into those sites to see if they are legitimate or not.
  • Avoid using platforms that lack transparency. Don’t use Criteo for this reason, as they refuse to provide transparent reports of where your ads are served, preventing you from doing a proper audit.

Hijacking Visits and Overwriting Analytics

When you install 3rd party javascript pixels on your site, you give the owner of that script a lot of power. In particular that script can overwrite referral data, popup hidden pages, and generate clicks to your site. It can even directly write to your analytics platform!

You may think that your analytics data is safe, but think again. When you install any 3rd party javascript on your site, you also expose your analytics account for injection.

This is super alarming and you should be very concerned about this capability. Any Advertising Platform that has any javascript installed on your site can be overwriting and modifying your analytics data. They can override referral data for direct and organic visits to make it look like they are coming from their network. They can highjack e-commerce transactions and attribute them to themselves. They can make it seem like they are out-performing all your other marketing channels.

SteelHouse used “malicious code to make it appear as though an internet user clicked on a SteelHouse-placed advertisement, even though no such click occurred.”

Leah Bliss
former Global Retargeting Manager, VistaPrint

Steelhouse, under their Verified Visit model, uses their javascript pixel to generate clicks on ads, visits to your site, and to attribute conversions to themselves directly in your analytics.

How this works:

  1. A legitimate customer visits your site but does not purchase. He is now placed into your remarketing list.
  2. This customer now visits another site and is shown an ad. The customer DOES NOT click on the ad, but the remarketing cookie records a View.
  3. An hour later, the customer revisits your site from an organic source. The pixel script on your site detects that the customer has a recent ad view recorded.
  4. The pixel script loads an invisible iFrame, in the iFrame it loads the original ad and generates an automatic click and visit to your site.
  5. Your legitimate customer now has a remarketing cookie set that says he saw an ad, clicked on that ad, and visited your website from that ad. Your analytics data also says that this customer came from an ad click, and not from the original organic source.
  6. When that customer comes back to your site to make a purchases, that ad will take credit for the conversion.

Worst still, when you look in your analytics, it will appear that your Remarketing Platform is sending you lots of visitors and conversions, which will cause you to spend more with them. The reality is that these visitors are organic visits that were hijacked by the remarketing pixel.

“SteelHouse’s practice of inserting a code into an internet user’s browser to make a view of an advertisement appear indistinguishable from a click on Decker’s Adobe Analytics system is not, in any way, common or acceptable industry practice. Nor should it be because it is deceptive.”

Graham McCulloch
Director of eCommerce, Deckers


How to protect yourself?

  • Insist on an image only tracking pixel. Although this is rarely a viable option, try to avoid 3rd party javascript when possible.
  • Keep your analytics private and hidden. The less they know about your analytics platform and setup, the more difficult it will be for them to inject data. Avoid giving them any access, and ideally don’t even tell them what platform you use.
  • Get a written declaration from the Remarketing Platform that they will not inject data into your analytics, that they will not modify referral data, that they will not generate clicks, nor do anything that will simulate visits to your site or in your analytics.
  • Don’t use Steelhouse

More steps to protect yourself

Be very skeptical

Be very skeptical of all data provided by the 3rd party platform. By default assume they are lying to you and are trying to take your money. Always rely on your in-house analytics data first (although even this can be messed with via 3rd party pixels) and always try to deeply understand how their platform and algorithms work. Explicitly ask what they are doing about ad fraud.

Beware of secret sauces, black boxes, and magical artificial intelligence powered predictive self optimizing algorithms. If something looks too good to be true, then it probably is.

Never ask the barber if you need a Haircut…

The ultimate selfish goal of any Advertising Platform is to MAXIMIZE your ad spend. You are THEIR customer and they want YOUR money. They want you to allocate as much of your budget to them as possible. They are incentivized to fool you.

Therefore be very skeptical with the data they give you and the stories they tell you.

Recommended Platforms

I would explicitly AVOID using either Steelhouse or Criteo due to first hand bad experiences with both. Here are the only three remarketing platforms that I currently recommend:

  • Google Display Network with my recommended remarketing campaign setup and segmentation. Also continually audit the ad placements site list for fraudulent sites.
  • Facebook but ignore the “Results” metric and configure your attribution model for Click Through Conversions only. You can configure this in Ads Manager by clicking on Columns > Configure Columns > Attribution Window > Comparing Window and only selecting 1, 7, or 28 day click attribution.

More Reading about this topic:

Analytics Attribution Digital Marketing Google Ads

YouTube Campaigns are not performing as well as you think!

I recently ran a conversion focused YouTube video campaign that performed extremely well: According to AdWords I spent $10,000 and generated $100,000 in revenue.

However, looking at our analytics data showed a much different picture: That same $10,000 spend only generated $3,000 in sales.

Why the huge discrepancy? Something wasn’t right…

View Through vs Click Throughs

First a quick explanation of the difference between View-through Conversions and Click-through Conversions:

View Through Conversions (VTC) occur when someone sees your ad but does not click on it, and later buys your product. These types of conversions generally are only counted in the “View-through Conversions” column in AdWords.

Click Through Conversions (CTC) occur when someone sees your ad, Clicks the ad, and then later buys your product. These types of conversions are counted in the “Conversions” column in AdWords (and are generally the only conversions tracked by Google Analytics).

YouTube Campaigns count Conversions differently from other Campaign types

Standard Campaigns

For all campaign types except YouTube, a conversion is only counted in the “Conversions” column if someone clicked on your ad and then purchased your product. If a user sees your ad but does not click, but purchases anyway, that conversion is only counted in the “View Through Conversion” column.

This is the normal expected behaviour.

AdWords Search and Display conversion attribution

YouTube Video Campaigns

For YouTube campaigns, things behave differently. Click-throughs behave as expected, but View-Through conversions are handled very differently, and are divided into two categories:

Impression View-Through: A user briefly sees your video ad, does NOT click, but then buys your product. This conversion is counted in the “View-Through Conversion” column. (This is normal expected behaviour)

Video 30s View-Through: A user watches your video ad for at least 30 seconds, does NOT clickbut buys your product. This conversion is counted in the “Conversions” column, just like a Click-Through conversion. THIS IS NOT EXPECTED BEHAVIOUR!

YouTube conversion attribution

From Google:

Keep in mind: An impression is different than a “view” of a video ad. A ‘view’ is counted when someone watches 30 seconds (or the whole ad if it’s shorter than 30 seconds) or clicks on a part of the ad. A ‘view’ that leads to a conversion is counted in the ‘Conversions’ column.

Misleading Data

A “Video View” is more like an Impressions than a Click

A 30 second video view is much more like an impression that it is like a click, so why are these conversions being lumped in with click-through conversions? They should clearly be lumped into the View-Through Conversion column, or if YouTube wants to explicitly report on this metric then create another column type. Don’t lump it in with click-through conversions.

Unexpected Behaviour

The problem with all this is that suddenly the “Conversions” column behaves differently in one campaign type vs another. Suddenly the clean click-through conversion data is being polluted and mixed in with View-Through conversion data. This will make the campaign appear to perform better than it should. Which will lead you to incorrectly increase spend.

In our specific case, this issue led to AdWords over reporting campaign revenues by 3000%

This could also affect your regular display campaigns

This issue could also affect your regular display campaigns if they are setup to serve ads on the YouTube network. Any of your regular display ads displayed on YouTube network will count View Through Conversions as if they were Click Through Conversions. This leads to attribution poaching, and makes your display campaigns appear to perform much better than they actually do. This ultimately causes you to increase bids and budgets, and overspend.

To exclude YouTube placements, go to: Campaign Settings > Additional Settings > Content Exclusions and select all of the following for exclusion:

  • Live streaming YouTube video
  • Embedded video
  • In-video


  • This should only be an issue if you use Google AdWords Conversion tracking. If you import your conversions directly from Google Analytics, then this should not be an issue (as Analytics only counts click throughs)
  • If you are running  a pure brand awareness campaign, this is probably less of a concern for you.
Analytics Digital Marketing Google Ads

Why doesn’t my Ad Spend Scale?

Consider this Scenario:

You spend $1,000 on a new Ad Campaign that generates $10,000 in revenue. “That’s a great ROI” you tell yourself, “Let’s double the spend!“.

When you double the budget to $2,000, your Campaign only generates $12,500 in total sales and not the $20,000 you were expecting. Why?

SpendSalesCost of SalesROAS
Ad Campaign #1$1,000$10,00010%1000%
Ad Campaign #2$2,000$12,50016%625%
Increasing spend by $1,000 only resulted in $2,500 in additional revenue: An incremental cost of sales of 40% and an incremental ROAS of 250%

Why doesn’t it scale?

By scale I mean that your ROI should be linear: If the first $1,000 generates $10,000 in sales, then the next $1,000 should also generate $10,000 in sales.

The issue is that rarely is your campaign performance evenly distributed. If your drill down deeper into your initial $1000 Campaign, you might see the spend broken down into something like this:

$1000 Campaign$1,000$10,00010%1000%
Majority of the sales are being generated by a small subset of the overall campaign. A classic 80/20 scenario. The performance of the Branded ad set is subsidizing the cost of the unbranded ad set.

The performance of the Branded subset is subsidizing the cost of the Unbranded subset. 80% of your sales are coming from only 20% of the spend, while 80% of your spend is going towards an underperforming segment.

When we try to double the budget to $2,000, here is how the budget gets allocated:

$2,000 Campaign$2,000$12,50016%625%
When budget is doubled, most of the spend goes towards the underperforming segment, resulting in disappointing incremental sales.

With this new data in mind, we should probably:

  1. Decrease budget on the Unbranded segment
  2. Increase budget on the Branded segment

However, it is probably the case that your performing segment is already receiving 100% reach/impressions. So spending more is usually not possible. (In particular for Google Search Ads targeting your branded term: How much you can spend is a function of how many people are searching for your brand. Once you reach everyone, spending more can’t get you more people).

If you want to increase your spend, the only place to do so is in the underperforming unbranded campaign. But at least you’ll have a better expectation of the results.


  • Avoid making budget decisions on aggregate data. Always try to segment and dig a little deeper.
  • Don’t let underperforming segments ride the coat tails of your top performers. Look for 80/20 campaign and ad group performance and analyze those individually.
  • For Google Search Ads, always separate your Branded search terms and Unbranded search terms into separate campaigns.
Analytics Shopify

Fix Product Performance Reports in Google Analytics with Shopify

By default, Shopify sends transactions to Google Analytics with a unique product title for each product variant. This causes the Product Performance Report to be split on the variant level instead of at the product level as it was intended.

This is how Shopify data shows up by default in the Product Performance Report. Notice that the various “Trillium Parka” variants are ungrouped because of the different size and color information in the product name. This makes it difficult to see “Revenue by Product” for all “Trillium Parkas”.

If, for example. you are selling winter boots, and someone buys a size 10 in Black, Shopify will send the product name as “Winter Boots – 10 / Black” instead of just “Winter Boots“. This is a bug as as the variant details are already included in the Google Analytics “Product Variant” column.

The Solution: GA Custom Data Import

The solution to this problem is to overwrite the Shopify data using the Google Analytics Custom Data Import tool.

1: Export your Product Data

First we need to export all our product data – we can accomplish this by creating a custom Collection Template that generates a CSV report instead of the standard HTML.

a. Create a new Collection Template

Call the new collection template csv-ga-product-feed and paste the following code:

{% layout none %}{% paginate collection.products by 1000 %}ga:productSku,ga:productName,ga:productVariant{% for product in collection.products %}{% for variant in product.variants %}
{{ variant.sku }},{{ product.title | replace: ',','' | remove: '"' | remove: "'" | strip_html | strip }},{{variant.title | replace: ',','' | remove: '"' | remove: "'" | strip_html | strip }}{% endfor %}{% endfor %}{% endpaginate %}

(also available on GitHub here)

b. Create a new Collection based on your csv-ga-product-feed Template

Select the products you want to include in this feed (probably all your products). These will be the products whose values will be overwritten in Google Analytics. Call your collection “Google Analytics Product Data Import” or something similar and save it.

c. Download your Product Feed

  • View your new collection in your store (eg:
  • View source in your browser and save as HTML
  • Rename the file with a CSV extension (eg: google-analytics-product-data-import.csv)

2: Setup and Import the data into Google Analytics

WARNING! You can really mess up your Google Analytics data if things go wrong. I highly recommend that you duplicate or backup your Google Analytics view and do a trial run before working with your live data. Once you upload this new data and overwrite there is no UNDO!

a. Setup the Data Feed

  • Go to Google Analytics > Admin > Account > Property > Data Import
  • Click the red “+ NEW DATA SET” button
  • Select “Product Data”
  • Give your Data Import a name: “Product Name Override”
  • Select the Google Analytics Views you want this import to affect
  • Setup your Data Set Schema: Product SKU is the mandatory key, but select Product and Product Variant as the additional fields.
  • Overwrite Hit Data: Choose Yes (but read my warning above)
  • Click Save and Done

b. Upload your data feed

  • Click on “manage uploads” beside your new Data Feed definition
  • Click the blue UPLOAD button
  • Choose your CSV file and click UPLOAD again
  • And now wait for the upload an update to be complete

3: Verify your new data

The data upload will only affect data from this date forward. So your old data will not be fixed. But your future data will be nice and clean… Until you add new products to your store, in which case you will have to repeat this process.

You will need to wait at least a day before you start seeing the new data coming in. If you add new product SKUs to your store, you will also need to regenerate and reupload a new file in order for the new product data to be fixed.

Attribution Digital Marketing Facebook Ads Google Ads

The Value of View Through Conversions

My personal experience with View Through Conversions (VTCs) is that they provide very little real world lift in sales, in particular when dealing with lower funnel remarketing campaigns. The value increases if you are advertising to a brand-unaware audience, but the value is likely still very small.

But what is the actual value of a View Through Conversion vs a Click Through Conversion? You can theoretically calculate this by running a placebo A/B test of your actual ad vs a Public Service Announcement ad.

Below is such an A/B test that was run on the AdRoll network back in 2013:

Placebo A/B Tests to Measure View Through Conversions

Display Network A/B Test

You will likely need assistance from your Ad Platform or agency to properly run this sort of test, and to properly segregate the A and B groups.

Setup a Placebo A/B Test

  1. Create 2 separate (but equal) non-overlapping campaigns
  2. Campaign A will serve your “normal” ad
    Campaign B will serve a Public Service Announcement ad
  3. Run both campaigns for a few weeks
  4. Calculate VTC “lift” as follows:
Valid VTC Formula

Real Life Example

Campaign A
(Real Ad)
Campaign B
(PSA Ad)
Data from a real world A/B test performed in 2013 on the AdRoll network

VTC “Lift” = (329 – 261) / 329 = 20%

In this real life example, the irrelevant PSA Ads still managed to generate 80% of the View Through Conversion volume that the real ad generated. To be more clear, 261 people saw an ad to adopt a cat, and later went to to purchase a boating license.

So 80% of VTCs can be given a value of zero. Of the remaining 20%, more analysis is needed to figure out exactly how they influenced sales. The Real Ad did seem to generate more VTCs than the PSA ad, but the real question is those VTCs result in extra incremental sales or are they simply tracking sales generated by another channel such as e-mail? I don’t have a scientific answer to this, but anecdotally, I think the answer is that the VTCs did NOT result in incremental sales.

NOTE: A few years later we ran a similar Placebo A/B test on Facebook with the help of SocialCode. That test showed ZERO lift from VTCs, and the PSA Ads actually outperformed the real ads from a VTC standpoint!

So how should you value View Through Conversions?

Here are my recommendations:

  • Give View Through Conversions a value of ZERO. Unless you can prove otherwise via an A/B test, assume VTCs are not providing any lift to your campaigns. This is especially true for re-marketing campaigns as the visitors have previously visited your site, and may be actively engaged in checkout when the ad is shown.
  • Run your own Placebo A/B test. If someone insists on using VTCs in performance metrics, then you should insist on running an A/B test to calculate the true value. You should run at least two tests: One for remarketing audiences and another for brand-unaware audiences.
  • Be cautious and skeptical of anyone pushing the value of VTCs especially if they are an ad agency or ad platform that will benefit from including the extra VTCs in their performance metrics.