How to check why PPC performance changed

There are lots of moving pieces in PPC campaigns and that can make it challenging to pinpoint the cause when performance changes. An investigation into the cause of the change can either be top-down or bottom-up. Here we’ll cover both ways to investigate the root cause and we’ll also share tools that make this process faster.

The difficulty with investigating PPC performance changes

What makes it so hard to know why results in PPC change is that all ads run through an auction every time a search happens and every auction has different parameters — like where the searcher is located, what time it is, what else they’ve been researching, and much more. Every time the advertiser makes a change to settings, it changes how they participate in the auction and that can lead to different outcomes for the main KPIs. A few examples:

1. The advertiser changes their bid and this causes a change in performance because their ad is now shown for a different set of queries.
2. The advertiser enables a new ad type like RSAs (responsive search ads) and the performance changes when Google’s Machine Learning system starts to show the ad to a new audience that wasn’t exposed to the old type of ad before.

Top-down investigation

Investigating why performance changed usually starts with a question about a primary KPI. For example, you might ask, “why did we have fewer conversions last month than the month before?” Pulling this data can be done straight in the Google Ads interface by adding a second date range to the campaign pages. It’s simple enough when you want to know the size of the change but it gets cumbersome quickly when advertisers start looking for interactions between metrics in an effort to connect the dots to determine the underlying cause.

A manual investigation of why performance in an account changed involves several steps in the Google Ads interface

For example, if you find that conversions have gone up, you may want to learn why this happened. As we all know there are two direct drivers of conversions: clicks, and whether those clicks converted, i.e. whether they had any conversion rate. Next, if we want to know why clicks changed we have to look at the two primary things that drive clicks: impressions and CTR. This quickly becomes difficult in the Google Ads interface where everything is in a big table that is very wide and can’t be read without scrolling from left to right.

A tool like Optmyzr’s PPC Investigator greatly simplifies this type of top-down performance investigation. Its visualization closely mirrors the steps a PPC pro would normally go through manually.

The PPC Investigator tool from Optmyzr connects the dots so an account manager can quickly understand the main reason why their KPIs are changing. In this example, conversions are down primarily due to a drop in conversion rate.

The tool has many filters that enable a deeper analysis, like filters for campaigns, labels, date ranges, networks and device types. It takes just a few clicks to re-run the entire analysis when one finding leads to further questions. For example, when an advertiser finds a drop in conversion rate, they may wonder if this is due to differences between their mobile and desktop sites.

This further analysis is just a click away by adding a filter for device types. With a clear visualization, the advertiser can quickly understand that a poorly optimized mobile site is a big reason for declining performance. As more search volume shifts from desktop to mobile, their bad conversion rate on mobile devices is leading to fewer conversions despite the overall click volume being relatively steady.

The PPC Investigator also allows quick drill-downs into a root cause analysis where the user can get a sense of what elements of an account are major drivers of big changes in the metrics. For example, if it appears that a drop in conversions is due in large part to a decrease in impressions, a single click on the box for impressions brings up the root cause analysis that shows top movers for impressions.

With a single click on a metric that has changed, the user can find the main components of the account that are responsible for this particular shift.

Once an advertiser determines the metrics that need an investigation, they can drill down to the root cause analysis tab to see which parts of the account are most responsible for the shift.

The Optmyzr root cause analysis tab in the PPC Investigator highlights the main elements of the account responsible for a change in performance, e.g. which campaigns, ad groups, keywords, networks or device types are causing a change.

Bottom-up investigation

Another way to investigate why PPC performance changed is a bottom-up approach that starts from the most granular elements of a PPC account. While a top-down approach may hide good and bad changes so long as overall averages are steady, a bottom-up investigation will uncover these more granular changes.

An in-depth bottom-up investigation usually requires processing lots of data through a spreadsheet, a time-consuming task that PPC pros are all too familiar with.

The process of finding things that changed usually involves these steps:

1. Download data for the date range where results changed
2. Download data for the same entities from a previous date range where things were considered ‘normal’
3. Combine the two sets of data by doing a vlookup
4. Add a few formulas to calculate the amount of change and add these in additional columns to the spreadsheet
5. Add filters and sorting to bring the most important changes to the top

While this process works great in spreadsheets, tools like Optmyzr’s Rule Engine can make it faster and more repeatable, both important considerations for time-strapped agencies and in-house PPC teams.

The Rule Engine automatically grabs the necessary data from the ad engine and offers a simple graphical UI for building if-then statements with it. The following example shows a rule where Optmyzr automatically grabs data for two date ranges, does an automatic joining of the multiple date ranges and then presents it in an easy to read report.

The Optmyzr Rule Engine lets advertisers create advanced If-Then statements to automate complex optimizations and analysis of PPC accounts.

Example rules for investigating PPC performance

Optmyzr has created predefined ‘recipes’ in its Rule Engine so that PPC pros can run common investigations and optimizations with a single click while still allowing access to the underlying methodology to those who want to customize the logic.

Drop in impressions – due to lost queries

The primary way to target ads in PPC is through keywords. But those keywords are really just a means to target the queries that users are doing. A simple change like a new bid for a keyword can impact what queries that keyword shows for. And when the results look different, it’s hard to know if the bid or the new query mix is the main cause. To investigate this, advertisers can use the Rule Engine to compare queries across two date ranges. Of particular interest may be queries that went from low volume to high volume and vice versa.

A prebuilt Rule Engine ‘recipe’ helps advertisers find queries that used to perform well but that no longer do. By writing the logic into a rule, advertisers can run the same analysis automatically on a regular basis, and apply the same methodology across many accounts.

Slow decline in performance

Another good investigation is to find entities in a PPC account that are slowly but steadily trending the wrong way. In a week-over-week or month-over-month investigation, most attention is usually devoted to things with big changes. But that lets slowly degrading components of the account slip past the manager’s attention and eventually these small changes can really add up.

Google wrote a sample script for this, the Declining Ad Groups report. And while scripts are great — Optmyzr even has many that you can simply copy-and-paste into your account — there are still many advertisers who’d rather not work with scripts. The Rule Engine is a great solution that gives advertisers the power of scripts in a more familiar interface for specifying their logic.

Rather than having to write the logic for the investigation in JavaScript code, it’s now possible to create the logic in a graphical rule builder. Taking the example script from Google, what if the advertiser wanted to find declining keywords or shopping product groups? What if they wanted to set a threshold for when a decline is significant enough to worry about? Making these improvements to the methodology is easy using Optmyzr’s Rule Engine.

The Optmyzr Rule Engine provides an easier way to create a report of declining performance than using a Google Ads script. The logic shown here looks for 4 straight weeks of declining CTR where impressions were significant enough to make the CTR meaningful.

Close variants

A bottom-up investigation doesn’t always need to start with a question about performance. It can also analyze whether ad targeting is remaining focused enough. In this example, a rule can be created to flag cases of exact match close variants that need to be investigated. In September of 2018, Google made close variants broader than before by allowing exact match keywords to show ads when the query is considered the ‘same meaning’ as the exact keyword.

This Rule Engine recipe finds queries that close variants for exact match keywords in a Google Ads account. It can be extended to add negative keywords when needed.

While this is just a report, the beauty of a Rule Engine is that it’s easy to take findings and act on them. For example, this rule could be enhanced and use performance attributes to decide when to add a negative keyword for an undesirable close variant.

Conclusion

Explaining why performance in an account shifts can be tremendously time-consuming for PPC teams. Whether you prefer a top-down or bottom-up investigation to find the root cause, there are tools like Optmyzr that can streamline the process and make it more consistently repeatable so that PPC managers can provide the high level of service expected by the companies whose marketing they support.