Bid Management Lookback Options
PPC bid management tools. Do look back. Your competition may not be gaining on you.
PPC bid management tools. Do look back. Your competition may not be gaining on you.
Bid management programs seek to simplify search marketers’ jobs by automatically adjusting PPC (define) bids to meet ROI (define) goals. Most manage bids by examining data over time and adjusting bids as data comes in from campaigns.
Business rules specified by the person managing the PPC campaign comprise the major “artificial intelligence” component of bid management technologies. Business rules define the goals of a campaign. For example, a campaign might have a goal of 130 percent return on ad spend (ROAS) For example, you’d earn $1.30 for every $1 spent. Other business rules include be cost per acquisition (CPA) or cost per lead. That goal might be acquiring leads at a cost of, say, $10 or less per acquisition.
There are many ways to set up business rules. The options for creating business rules vary from tool to tool. Most bid management technologies offer business rules, while some simple tools still rely on “bidding rules” to make decisions.
Lookback Time Period
In an ideal world, bid management programs would calculate ROI on a keyword by keyword basis. However, this is not as simple as it may seem. One of the key factors is how far back in time you need to look at the data to get meaningful results.
For example, you may decide you want to look back 30 days. It sounds like a reasonable amount of time. Right? Wrong. The problem: there’s no one time period that works for all keywords. To illustrate the point, let’s look at some live sample data:
Keyword
|
Clicks/
Month |
Conversions/
Month |
Clicks/
Week |
Conversions/
Week |
Clicks/
Quarter |
Conversions/
Quarter |
1
|
876
|
48
|
219
|
12
|
2,628
|
144
|
2
|
772
|
42
|
193
|
11
|
2,316
|
126
|
3
|
291
|
12
|
73
|
3
|
873
|
36
|
4
|
289
|
4
|
72
|
1
|
867
|
12
|
5
|
288
|
14
|
47
|
4
|
864
|
42
|
6
|
188
|
6
|
41
|
2
|
564
|
18
|
7
|
164
|
5
|
40
|
1
|
492
|
15
|
8
|
158
|
7
|
38
|
2
|
474
|
21
|
9
|
152
|
1
|
33
|
0
|
456
|
3
|
10
|
132
|
5
|
33
|
1
|
396
|
15
|
11
|
131
|
6
|
32
|
2
|
393
|
18
|
12
|
129
|
6
|
30
|
2
|
387
|
18
|
13
|
120
|
3
|
28
|
1
|
360
|
9
|
14
|
111
|
5
|
26
|
1
|
333
|
15
|
15
|
105
|
2
|
26
|
1
|
315
|
6
|
16
|
102
|
5
|
26
|
1
|
306
|
15
|
17
|
102
|
3
|
26
|
1
|
306
|
9
|
18
|
88
|
0
|
22
|
0
|
264
|
0
|
19
|
86
|
3
|
22
|
1
|
258
|
9
|
20
|
86
|
7
|
22
|
2
|
258
|
21
|
Source: Marin Software |
This sample data is a snapshot of real data, although the exact keyword names are obscured. The data shows a wide variance in click and conversion rates. If you believe you need at least 30 clicks for a given keyword in order to do a meaningful analysis of the results for that keyword, you see only some of the above keywords meet that metric.
You need to take care in picking a time frame for analyzing the data. For example, if you pick a month for your time interval, some of your keywords in the above table will have provided enough data in just a week.
If they’re expensive keywords losing money at a rapid rate, this is a problem. Or, if you have a very profitable keyword you are making money on, but you can increase your profits by raising your bids, why would you want to wait for a month if there already is a way to tell that after one week?
However, it’s more complex than this. To show how, let’s look at another slice of the data:
Keyword
|
Clicks/
Month |
Conversions/
Month |
Clicks/
Week |
Conversions/
Week |
Clicks/
Quarter |
Conversions/
Quarter |
221
|
1
|
0
|
0
|
0
|
3
|
0
|
222
|
1
|
0
|
0
|
0
|
3
|
0
|
223
|
1
|
0
|
0
|
0
|
3
|
0
|
224
|
1
|
0
|
0
|
0
|
3
|
0
|
225
|
1
|
0
|
0
|
0
|
3
|
0
|
226
|
1
|
1
|
0
|
0
|
3
|
3
|
227
|
1
|
1
|
0
|
0
|
3
|
3
|
228
|
1
|
0
|
0
|
0
|
3
|
0
|
229
|
1
|
0
|
0
|
0
|
3
|
0
|
230
|
1
|
1
|
0
|
0
|
3
|
3
|
231
|
1
|
0
|
0
|
0
|
3
|
0
|
Source: Marin Software |
If you look at keyword 230, you’ll see it has three conversions on three clicks on a quarterly basis, or a conversion rate of 100 percent. The average conversion across the entire campaign is just over 4 percent.
If you were to make bidding adjustments based on the quarterly performance, you would raise your bid through the roof. Not a good idea. Just as bad would be to take the keywords with no conversions and assume they’re all duds.
One way to try and deal with this: decide you were going to sit on every keyword bid until it generated at least 100 clicks. The problem? For some of these keywords you’ll have to wait a long time.
Your campaign might have many thousands of underperforming long tail keywords sucking the wind — and the profit — from your campaign. You just can’t afford to wait that long.
Moreover, there may be a seasonal aspect to your campaigns (for example, sales of ski boots). Waiting too long a period of time will therefore run into problems because the expected performance will differ greatly from summer to winter.
The Solution
The lesson: one size does not fit all. You can’t pick one lookback time interval for all of your keywords. You need to subdivide the problem to a finer level of granularity.
The solution: use several different lookback scenarios in addition to a basic 30 days lookback scenario. Marin Software, who contributed the raw data for this article, recommends the following approach:
Here are some of the things that can be done to enhance the results for these keywords:
Conclusion
In tests conducted by Marin Software, this modification to the approach yielded an increase from 1,300 clicks per week to 3,000 clicks per week for the client whose data was sampled above. That’s an outstanding gain in performance, one that may be the result of a more sophisticated approach to looking at available data.
Note: Marin Software has an application for larger search marketers ($50,000 and up in monthly spending). The software provides slice-and-dice analysis, but also has bidding capabilities. It calculates “best-guess” bids for low-volume keywords (90 percent to 99 percent of all keywords) using Bayes’ Theorem.
Join us for SES Chicago from December 3-6 and training classes on December 7.