Measurement Protocol 101: Improve Your Google Analytics Data | CXL Institute Revenue optimization services Blog Search Start 7-day trial for $1 Subscriptions Minidegrees Sprints Online courses Free courses Resources

Measurement Protocol 101: Improve Your Google Analytics Data

Measurement Protocol 101: Improve Your Google Analytics Data

Google Analytics and Google Tag Manager have limitations. Often, these limitations are outside your control (e.g., technical aspects of a website that can’t be changed).

Fortunately, many issues—such as updating a user status from your CRM or sending refund data—are solved by using Google Analytics’ Measurement Protocol

The Google Analytics Measurement Protocol allows developers to make HTTP requests to send raw user interaction data directly to Google Analytics servers. This allows developers to measure how users interact with their business from almost any environment.

Don’t let the “allows developers” phrase scare you. Using the Measurement Protocol is nothing more than adding a bunch of parameters to a URL. Therefore, it should be equally simple for people who aren’t developers to add it to their toolstack.

This article shows how marketers can use the Measurement Protocol without coding—structuring data in Google Sheets or Excel instead. Still, there are advantages from being able to code (even a little bit), or having a developer friend to help with some of the solutions below.

Parameters and why they matter

With the Measurement Protocol, you can mirror the data and collection method of Google Analytics. On any page running Google Analytics, simply go to the console under Network and type “collect.”

Reload the page, and this will show up:>m=2wg9p0NX56DK&cd4=%22%22&cd5=583336462.1570739164.1570739177647.pageview&z=1459566673

url that contains information that's sent to google analytics.

At a glance, this might seem incomprehensible; however, if you deconstruct the URL from the query parameters, you’ll see it as a list of information that’s passed to Google Analytics:

  • v: 1
  • _v: j79
  • a: 27896466
  • t: pageview
  • _s: 1
  • dl:
  • ul: en-gb
  • de: UTF-8
  • dt: Optimization and Growth by CXL
  • sd: 24-bit
  • sr: 1440x900
  • vp: 885x789
  • je: 0
  • _u: SCCAgAALQ~
  • jid:
  • gjid:
  • cid: 583336462.1570739164
  • tid: UA-486673-28
  • _gid: 521630025.1570739164
  • gtm: 2wg9p0NX56DK
  • cd4: ""
  • cd5: 583336462.1570739164.1570739177647.pageview

Here it can be seen that:

  • Document location (DL) is
  • User Language (UL) is English.
  • Document Encoding (DE) is UTF-8.
  • Client ID (CID) is 583336462.1570739164.
  • Tracking ID (TID) is UA-486673-28.
  • We are tracking (T) a pageview.
  • And so on.

The URL contains the exact information sent to Google Analytics. You use the same approach to send data through the Measurement Protocol—emulating what happens each time you do something on a webpage that’s tracking user behavior.

To see all available parameters, visit the Measurement Protocol Parameter Reference guide.

Sending data via the Measurement Protocol

Most programming languages are capable of sending data via the Measurement Protocol through Post messages.

I recommend using the tool that’s most natural for you to use. If you like using R, Mark Edmonson created a package that lets you do it. To execute requests, the only thing you need is the data.

But you can also do it via Python, PHP, and tools such as Zapier or Google Sheets.

Note: Before executing any instructions for the use cases below, please read the caveats and advice section that follows.

4 practical examples for using the Measurement Protocol

For marketers, four use cases highlight the potential value of the Measurement Protocol:

  1. Correct missing transactions;
  2. Adjust duplicate purchases;
  3. Send refund data;
  4. Update a lead status.

1. Correct missing transactions

With Google Analytics, proper quality assurance can be difficult. You may not realize that ecommerce tracking isn’t working—until you discover that data is missing.

To add missing transactions to your Google Analytics Property, you need the following data points:

  • Google Analytics data with order ID’s and date;
  • Actual sales data for the time period that needs to be recovered.

This is the flow that makes it possible to extract, transform, and load data back into Google Analytics:

flow for data using the measurement protocol.

Here are the steps to do it:

Extract the data

To begin with, we need the correct data from a sales database. Once we have the correct transactions for the missing period, extract the sales data from Google Analytics through the API, or export the data into Google Sheets:

google sheets export from analytics.

Transform the data

Once you have the data, create a sheet (Excel or Google Sheets—doesn’t matter) and merge the transaction IDs and revenue. The table should look like this:

You can then use this formula to see if there are any duplicate values in the dataset: 


removing duplicate transactions from spreadsheet data.

Filter out any transaction that already exists in Google Analytics (i.e. select rows only with the value “0” in them).

To finish creating the upload dataset, add the necessary parameters. Use the Google Sheet template here to align your dataset with the image below:

spreadsheet with curl request data for measurement protocol upload.

The last step to transform the data is to copy the information from the “Request” column into your favorite code editor, such as Visual Studio Code:

curl examples.

Load the data into Google Analytics

You need to send cUrl requests with the hits created in the previous step. To do so, open your terminal or GIT Bash. Copy all the code and press Enter.

The following view should now appear:

using the bash terminal for measurement protocol.

Press Enter one last time to get the last row, and look at the Real-Time report (“Last 30 min”) to confirm that the hits have been sent:

real time view in google analytics to verify that measurement protocol hits have gone through.

Unfortunately, Google’s data model doesn’t allow you to specify a date for the uploaded data; it’s determined based on the time the data is sent. Another side effect of this approach is that you lose the users’ Client ID and, therefore, the referral data for the transaction.

A way to manage this is to give a weighted distribution on your calls “traffic source” and “medium” based on historical data. You can also give out custom values, such as calling the traffic source “Measurement Protocol.”

The most important thing is that it fits your business. And, of course, the optimal solution is to QA your Google Analytics setup after any site releases that might affect tracking.

2. Adjust duplicate purchases

A frequent issue in Google Analytics is duplicate transactions. Duplicate transactions can happen if people revisit the receipt page, or if something is wrong with the technical implementation.

When auditing Google Analytics, you sometimes find that a site isn’t tracking Enhanced Ecommerce with Events. This causes ecommerce data to be sent with a Pageview, and—if an Event occurs on the transaction page at the same time—it will generate two transactions.

Correcting this data can fix revenue figures for weekly or monthly reporting. However, because we need to add a negative transaction to correct the purchase amount, it creates an additional transaction, inflating the total number of transactions on the site and skewing the ecommerce transaction rate. 

Thus, I recommended this approach only:

  1. To correct large revenue discrepancies;
  2. If you use Google Analytics to report revenue.

The approach is similar to the one for the previous use case of adding missing transactions. This time however, only Google Analytics data is needed (not sales data). You’ll use the following variables:

  • Transaction ID;
  • Transactions;
  • Revenue;
  • Client ID (not mandatory but will improve accuracy).

 The flow to create negative transactions looks like this:

flow for negative transactions to integrate for measurement protocol.

Extract the data

As in the previous step, download the data from the API or create a Custom Report with the necessary Metrics and Dimensions, then export it to Google Sheets.

Transform the data

Not all transactions are likely to be duplicates. Create a filter and only include transactions greater than “1”:

filtering in a spreadsheet.

Then, calculate the refund amount using this formula:

(Revenue - (Revenue/Transactions))*-1

This should result in a view similar to this one:

example of how to add refund data to a spreadsheet.

Finally, make the cells match this Google Sheet and follow the same approach as before, copying the data into a code editor.

measurement protocol worksheet for curl request.

Load the data into Google Analytics

As before, go to your terminal and paste in the lines from the spreadsheet. Check that data is coming in correctly to your View.

3. Send refund data

Sending refund data to Google Analytics is not the same as reversing a transaction.

Instead, refund data appears separately in your Enhanced Ecommerce reports to help analyze how many transactions and products are refunded and, in turn, whether something is wrong (e.g., the way products are displayed on the site).

Getting refund data into Google Analytics automatically takes resources; many ecommerce companies skip the process altogether. This makes it difficult to calculate Return on Marketing Investment (ROMI) when comparing Google Analytics data to actual sales numbers.

If possible, the Client ID is great to have—it’s the most precise way to connect the refund to the user. If it’s not available, simply generate a unique one.

The process for creating refunds is the following:

example of data flow to match crm data with analytics data and send via measurement protocol.

Extract the data

Refunds added to Google Analytics will occur on the date they are sent. To get the most precise refund date, send them in often to get as close to the actual refund date as possible.

Transform the data

The Measurement Protocol hit for refunds is almost identical to the one for sending in purchase data. The only difference is the value for the parameter “pa” (Product Action), which is named “refund.”

To create the dataset, use this template and replace it with the refund data that you’ll copy and send through cURL requests:

measurement protocol spreadsheet template.

Load the data into Google Analytics

As in the other examples, copy the “Request” column, open the terminal, and paste the values:

bash console that sends hits to google analytics via the measurement protocol.

Finally, check the Google Analytics Real-Time report to ensure that data is coming in correctly. The data should be available under Ecommerce > Sales performance.

4. Update a user status from your CRM

Ecommerce data can indicate how an online store is performing. If a business’ primary KPI isn’t selling products but converting leads, however, it can be difficult to assess the efficiency of marketing efforts. 

Lead scoring—one way to assess those efforts—isn’t always feasible via standard Google Analytics data collection. Here, it’s up to developers to update the status of users and send that data back into Google Analytics.

That data offers marketers some new analytical options:

  • Understand the time lag from when a user first visits the site to when they become a sales-qualified lead.
  • In terms of attribution, see the marketing efforts that most efficiently converted leads.
  • For leads that get declined or exit the lead-scoring funnel, the data can help segment against them, preventing wasted marketing spend.

Here’s a hypothetical flow for bank leads:

sample flow for a lead.

The next steps show how to acquire the Client ID and use it to update the user status via the Measurement Protocol.

Collect the Client ID

To connect on-site user behavior to Google Analytics, the Client ID must first be collected:

capturing the client id via a hidden field in a form.

A hidden field should collect the Client ID from the ga cookie. This value needs to be stored in the CRM/lead-scoring system.

Create a Measurement Protocol trigger

Once the data is collected, developers need to set up the system to fire a hit each time a user status updates. The hit needs to be formatted this way:{{the client id from the CRM}}&ec=measurement+protocol&ea=lead+updated&el={{the lead status updated}}&cd1={{the lead status updated}}&ni=1

The “cd1” parameter is the Custom Dimension. If you need to set it to another Custom Dimension (because you’ve already created others), you can use any number between 1 and 20 (200 if you have Google Analytics 360) after the “cd” variable, such as cd17 for Custom Dimension 17.

3. Set up the Custom Dimension

To receive the data from the Measurement Protocol hit, you need to activate the Custom Dimension. As we’re updating the users’ lead status, the scope should be set to “User” level:

add custom dimension in ga.

Click “Create.” The following information should appear with the Custom Dimension number:

custom dimension setup in google analytics.

From here, the data should be available as a Custom Dimension in reports when selecting Secondary Dimensions:

example of custom dimension in google analytics.

Caveats and advice

This section illustrates areas for which I wish there were more guidance in online communities, or even in the Google Developer documentation. These are things I’ve learned through trial and error.

Read this section thoroughly before implementing any solutions, as these caveats could potentially break your tracking setup if you’re not careful.

Set up filters for View-specific data

When working with multiple countries or business units, your tracking setup might look like this:

potential division of google analytics account.

However, sending a Measurement Protocol hit to your site will add the data to all Views in a Property. This means that if you try to send in transactions for Country A, this data will also be sent to Countries B and C.

In many cases, you need to split up the information to populate specific Views. To do this, add an Event to your hit to filter the data.

For example, send an Event to your Google Analytics Property with the following information:

  • Category: Measurement Protocol;
  • Action: [the action the hit produces];
  • Label: [which Views it should send to].

If we use the country example mentioned above, a filter that takes data only to Country C and the roll-up View could look like this:

custom filter to segment data import from measurement protocol.

It’s a simple way to exclude unwanted data and to use segments to view data delivered via Measurement Protocol hits.

Uncheck Bot Filtering

An easy mistake is forgetting to uncheck Bot Filtering when sending hits. Bot Filtering will, in most cases, block Measurement Protocol hits from being sent to the Property.

Go to View Settings and make sure the checkmark box is empty. If it’s checked (as in the example below), uncheck it.

bot filtering toggle in google analytics settings.

Formatting currencies

Always specify the currency for the value being sent. When getting data from your CRM or other platforms, convert the currency to a single unit, or make sure that each row of your data specifies the right currency.

To see how currencies should be formatted, visit Google’s help center.

Be careful with the Client ID

When sending data to Google Analytics, it’s best to have a unique Client ID. If you use the same one every time, all the data will be attributed to one user, which will skew your data. Instead, I recommend one of two solutions:

  • Send the original Client ID for the data that you’re trying to upload. This is the most accurate way.
  • Generate a Unique Client ID for each hit to distribute data evenly across unique users.

Keep GDPR in mind and protect users

The last thing worth mentioning about Measurement Protocol is to respect users. Once you have a Client ID and an email, the marketing possibilities are endless.

You can understand exactly who does what on your website, run segmented retargeting campaigns—even for people who don’t want it—etc.

Always have documented permission and be compliant with Measurement Protocol guidelines.


This post has focused on how to use the Measurement Protocol with as little coding knowledge as possible. But there are many (more) ways to work with it.

Build a workflow that aligns with the technical strengths of the user sending the requests. (To see an example of how Measurement Protocol can be automated with the statistical programming language R, check out this blog post.)

Measurement Protocol can be an asset to improve data collection. However, it can also cause more harm than good to your Google Analytics Property if used incorrectly. Send the information to a test account before putting it into production.

Finally—and unfortunately—others can also use the Measurement Protocol for your site, making it easy for third parties to obscure your data with misleading information. When you’re not sending Measurement Protocol data, filter out bot traffic in View settings and your filters.

Featured image source

Related Posts

Join the Conversation Add Your Comment

  1. Totally agree with you on whether or not to perform these actions at all. It has to “fit your business”. I’m leaning more towards simply communication with your eComm team what the calculated difference is of over-reported revenue. Hell, they should be looking at actual store sales in the eComm system rather than GA as ‘source’ truth. There is a risk to Paid Media however who may have campaign optimizations and bid strategies that run on eComm conversions.
    I ran the sheet with your formula on our business and in one Property, it turns out that 5% of all sales was over reported. I think the variance on duplicate transactions totally depends on your consumer tendencies to leave tabs open or re-visit order confirmation pages. I think its also a good idea to move the shipping information notifications to another page altogether, if you haven’t already.
    So with that said and moving on… I don’t think this is good practice to run the batch corrections. Here is why and just my 2 cents.
    If you do MoM or YoY comparisons, you are screwing yourself if you have to re-process a bunch of Duplicate transactions to their correct transaction amounts. The date issue as you mention is a huge obstacle. If you re-process the dup transactions, it makes the reporting even more unreliable. Yes, according to the GA data model, all of the corrected transactions show up on one date so if you have a bunch of negatives to reverse, the result is that you have a “really bad eComm day”.
    I know you mention using an event label to call these out but even then its a huge amount of money shifting. Depending on the amount of duplicates how it equates to X% of the overall sales, you could be shifting millions of dollars to a single day of sales.
    Also, after running a few test removals of the duplicate transactions and looking at the data in GA, there is another challenge with the dates. You would always have to include the date range you ran the ‘corrections’ because the math of the re-processing must exist in order to do a fair comparison for MoM, WoW or YoY reporting. If you don’t include the corrected date range, then you are still looking at the original transactions however many times it was duplicated in your date range by the user visiting the page repeatedly.
    For anyone looking to do this, I implemented Simo’s method of the customTask to check for transaction cookie in the client and also highly recommend a code fix to NOT fire the EEC purchase event if that Order page has loaded already for the user or any user. Do both to prevent over-reporting as client side GTM isn’t 100%.
    Great write up and also exposes some really cool methods to poison competitor’s GA data. ;-)

    1. Danny Mawani Holmgaard

      HI Lorenzo. Really strong and well put arguments on your comment =)

      In regards to having your Google Analytics data aligned as closely as possible, the only real good solution is to use measurement protocol server side, or else there are so many things that can go wrong. In many cases, people relying on Google Analytics data is the marketers that needs to allocate their spend best way as possible, which is extremely difficult to do through CRM alone. What we have done for some of our clients is to create a dashboards that takes the transaction table from the CRM / ERP and used the transaction id as a key for the channelgrouping (and other relevant metrics). In general there are so many ways to do everything, and I guess it is up to the individual team to assess how they solve it the best way.

      As a rule of thumb, you should never be able to refresh the page again, or at least the developers should add some tracking parameter in it so you can block the tracking script. In many cases, this has not been possible to implement, and I have also used Simo’s solution (with some level of success, since ITP and localstorage can be an issue).

      In terms of batch calling, the only reason that I have been able to report the correct numbers, has been because we have managed to add the right information back to GA within the week when things have gone wrong. Here I highly recommend using annotations to ensure that it you remember what has been done to the property. The “pasting into the terminal” way of doing it is not the best way of working with it in a smart way, however it was definitely the simplest way possible for me to illustrate how the process works.

      Thanks for the great reply, I am happy you enjoyed the read =)

Comments are closed.

Current article:

Measurement Protocol 101: Improve Your Google Analytics Data

Apply for the Conversion Research Sprint

Applications are open for our first live interactive training program. This program is designed to give you advanced-level skills in conversion research. There are limited spots available.

Apply now

We’re running the CXL Sprint program in four other disciplines:

  • Analytics frameworks
  • CRO program management
  • Growth experiments
  • Differentiation

Sign up here to stay updated with upcoming program dates and enrollment info.