How to Spot Bad Data Within Your PPC Campaigns & What to do About It

upside-down-roadsOptimizing PPC campaigns with faulty data is like driving with an upside down map. You’ll end up in the exact opposite direction of your destination.

Irregular performance data can be caused by technical glitches, seasonal fluctuations, other marketing channels, or even natural disasters. As a paid search manager you need to know how to spot trends that may indicate compromised tracking and how to mitigate the damage to your PPC campaigns.

Categories of Bad Data

Compromised, unreliable data can appear within your campaigns in a numerous ways.

“Bad” data can relate to any performance statistics that may misguide your optimization efforts. In other words “bad data” or “abnormal data” is performance statistics that have extenuating circumstances. These circumstances may cause you to make reactionary changes to your PPC account which can hinder long-term performance.

Below isn’t an exhaustive list of bad data causes (it would be impossible to list them all!), but these are a few to watch out for:

  • Misplaced tracking code: This is one of the most common reasons for improper data. Usually someone on the development team is making changes to a confirmation page that have nothing to do with a tracking pixel, but somehow the code disappears. Obviously, if tracking is removed from a website this is going to throw off your data.
  • Website errors: Websites can be fickle beasts. Sometimes out of the blue a page will just stop working or render completely wrong. When critical pages on a website (especially if they on the path to conversion) have an issue, this can negatively impact your performance. This impact on performance isn’t indicative of a downward trend, but rather it’s simply an error.
  • Conversion optimization testing: We’ve seen conversion rates increase and decline dramatically due to testing on the website/landing page. When conducting tests this type of fluctuation can be expected. However, it is mission critical that all stakeholders know that there is testing being conducted. For example, we were working on a specific campaign that had been struggling for a couple of weeks and we couldn’t put our finger on the problem. Then the client mentioned that maybe the test they had launched within their shopping cart was the issue – and of course we didn’t even know a test was running! Communication is key.
  • Special offers/deals: When looking at data you should most certainly take into account any offers/specials/etc. that may have been offered on the website. This isn’t necessarily “bad” data, but performance during a special promotion is usually the exception and not the rule when looking at long-term data.
  • Natural disasters: Keep in mind where your audience is located and what may be going on around them. We had clients that were significantly impacted by Hurricane Sandy. Our client isn’t located on the East Coast, but that’s where our core audience resides. When our traffic and conversion rates started to decline we knew it wasn’t a result of a change we made in the account. However, we made notations within Google Analytics so that we have notes when looking at long-term trends so that we understand what was going on during this time.
  • Fluctuations in other marketing channels: No marketing channel exists in a vacuum. All channels reside in a symbiotic relationship with each other (some more so than others). For one client we noticed that our brand traffic started to degrade quickly but everything looked OK; average ad position was holding steady, impression share hadn’t changed. After some discussion we learned that the client has stopped all of their television ad campaigns. As soon as they relaunched their television campaigns, our campaigns picked back up to their normal levels.
  • Political upheavals: We have clients that rely heavily on government contracts. Over the past quarter we had noticed that overall conversion rates had been depressed when compared to our averages. We started to conduct some research and realized that for many of our client’s potential customers their budgets had been frozen due to “fiscal cliff” negotiations. And many of their budgets are still in jeopardy. Again, the performance trends weren’t indicative of changes made within a PPC campaign, but this was a bigger-picture concern. The theme here is the same as some of the topics on this list: this isn’t necessarily, “bad” data but we need to take into account other external factors that are impacting our performance.

How to Spot Bad Data

Spotting “abnormal” performance changes can be difficult. The easiest issues to spot are website-related. In these instances the numbers should be cut-and-dry: one day everything is OK and the next day things are way off – sometimes traffic and conversions even drop to zero. That’s a pretty clear indication that you have a problem.

One easy way to monitor performance trends for extreme issues is to set up automated alerts within Google Analytics or AdWords. You can establish alerts to compare timeframes in order to find sudden or unexpected shifts in trends.

When setting up alerts you need to take into account normal fluctuation within your account. For example, if a 20 percent drop in revenue compared to the prior day is normal, then you may want to set up an alert that triggers if revenue drops by more than 40 percent.

Also, you may want to establish alerts that focus only weekends and others that compare weekends. For most accounts, transactions dip dramatically on Saturday when compared to Friday – but this is normal. You certainly don’t want a bunch of alerts sending out false panic messages (the whole, “boy who cried wolf,” thing).

Difficult issues to detected and attribute are related to external factors that don’t make your performance fall off a cliff. Rather, these issues can take a few days or even weeks to surface. The most effective way to combat these issues is to monitor recent and long-view trends closely.

Consequences of Bad Data

The biggest danger of “abnormal data” is that it can guide your team in the wrong direction when it comes to optimizing accounts.

There may be a short-term external factor affecting your performance. If you make hundreds of changes to your account to offset the issue then your campaign may be in poor shape when the issue is resolved (hopefully).

Immediate Actions for Bad Data

Depending on the urgency of the “irregular data” there are a few actions you may want to take so that these (hopefully) short-term issues don’t send your campaign off course:

  • Review any automated efforts: Campaign automation is based on performance data. If the data is skewed or compromised, the software will make changes that aren’t in the best interest of the account.
  • Make proper notations in analytics: Most analytics packages allow you to make notations regarding performance. You should certainly take plenty of notes as issues arise and when they are resolved.
  • Start working on solutions: After you have spotted a problem and you have determined the root cause, then you need to set a plan to fix the situation.
  • Communicate to stakeholders/team members: Notify everyone that needs to know about performance fluctuations within the account. Ideally, you’ll be able to describe the issue and the resolution plan.

Some of these examples aren’t necessarily “bad” data but they are “abnormal” data that can skew your PPC optimization efforts. A skilled PPC manager will be able to detect the difference between “normal” issues within an account that need to be addressed with campaign optimizations, and “irregular” issues that are impacting performance.

Remember that PPC campaigns don’t operate within a vacuum. If your recent statistics look way off, then there could be an external factor at play. It’s worth your time to conduct a bit of additional research before making knee-jerk adjustments within your campaigns.

Related reading

Five ways blockchain will impact search marketing
Using IF functions on Google Ads to improve productivity
Tips to lower brand CPC for greater profitability