Over the past year, I have come across more and more fancy-looking graphs and charts in the online marketing space.
This is most likely because strategists, analysts and executives are getting more savvy in general, but also because there are more and more easy-to-use analytics packages out on the market.
Although this is great, this analysis is often times very high-level and informational. It’s sometimes hard to understand the bottom line.
Whatever analytics package you or your team currently uses, I want to share a couple of ways you can look at your keyword data in SEM. I have found the below analysis to be useful when troubleshooting paid search performance, and beyond that, a good first step towards the resolution of potential issues.
1. Daily Impressions by Match Type, Including Modified Broad
Typical Insight: The below graph shows the impression breakdown by match type over time. It also includes “modified broad” which means you need to first label all those broad keywords using the ‘+’ sign as modified broad keywords in the data.
A high percentage of broad impressions (like those in the below example) are unhealthy because you have limited control over which ads are served by the search engines. If you notice the percent is trending up like in the example below, this indicates that you are losing control of your program. You want to keep the percentage of broad impressions low (somewhere around 20-40% depending on clients and based of efficiency) and make sure they are stable over time.
Potential Fix: There is always a way to add more exact keywords or negative terms to serve more targeted ads and filter out irrelevant impressions. This process should be repeated until the amount of broad impressions are “acceptable” as a percentage, which typically happens when all keywords are active in exact match type and the exact bids are greater than the phrase/modified broad/broad bids.
Note that a single broad keyword with a high bid could steal impression from the rest of the account, so watch intra-account cannibalization!
2. Average Cost-per-Click By Query Length & Match Type
Typical Finding: The below graph shows the average CPC (cost-per-click) by query length and match type, with the same modified broad tweak as in the previous analysis. In general, long-tail keywords are less competitive and cheaper as a result. One expects a lower CPC on keywords with more terms in them, which is not exactly the case in the below example where we see fairly high CPCs on 6-term-long keywords.
Potential Fix: You want to level off those peaks, especially for phrase/modified broad/broad match type. More specifically, this high-level analysis requires at least one extra step where you actually identify those keywords with more than 3-4 terms with an unreasonably high CPC.
3. Keyword-Level Impression Share Vs. Cost Per Acquisition
Typical Finding: First of all, note that not all data visualizations require a chart! Sometimes, a table is more relevant than a graph, especially when using colorful conditional formatting.
The below table shows the count of keywords, their conversion volume and efficiency level (CPA) by impression share level. The data was also split into branded and non-branded keywords. Essentially, you want to maximize your impression share on those terms with a low CPA.
The below table helps identify those keywords with a low impression share and a low CPA – those are the opportunity keywords. In the below example, 341 non-branded keywords have an impression share of only 60% while their CPA sits around $30. This is way more efficient than most non-brand keywords. One could also point out that most non-branded with an impression share greater than 70% also have a CPA greater than $95 – those are inefficient keywords you want to address, as well.
Potential Fix: Low impression shares could be due to three main reasons: low bids, low quality score or low daily budget (or a combination of the three). In short, if the campaigns in which the keywords sit are capped due to budget, you want to lower the bids on those inefficient keywords (high impression share, high CPA), allowing more budget to be spent on opportunity keywords.
However, if those campaigns are not capped, you’d first increase the bids on those opportunity keywords while lowering the bids on those inefficient keywords. Of course, for those keywords with a low quality score, you want to keep testing new ads until you see better performance.
While this fix is very typical, it could also be applied based on the average position instead of the impression share. Impression share can be a more intuitive indicator of the room for growth for opportunity keywords.
Conclusion
There are lots of ways you can look at your keyword data, so keep in mind the above graphs are just examples. My point is that in general, the purpose of visualizing data should not be to spit out fancy-looking graphs. Instead, those graphs or tables should be as actionable as possible – they should help understand what is going on and how to potentially tackle a problem, or improve what is already doing alright.
While I understand that sometimes there cannot be any simple takeaway, I feel it cannot hurt to regularly ask yourself or your team the following question: “Could there be a better, more actionable way to look into this?”
Hozzászólások