1000 locations KW research
Basically, I need to do KW research for about 1000 locations in 41 states. Each location needs its own click and cost estimations. The locations can have up to a 50 mile radius, but (because of double serving) it cannot overlap geos with another location, which could be in the next town over. Each location is an independant franchisee, so we cannot have the franchisees in the same account, each will have its own AdWords account. This is also because we don't want the owners having access to each other's accounts.
Any thoughts on how to go about this?
I really wish that we could use the radius targeting to help get estimated click volumes, but alas, the keyword planner can't do that. One method that we might try is to target by counties the city is in, which might be good enough for our estimates. Trying to piecemeal a geo fence around cities is very spotty.
We tried to do a general estimates by city size, but the problem that we ran into is that even though the city that location is in is small, it might be close to a larger city. That can significantly affect our data.
The reason we need decent click/cost volume estimates is because each location is individually owned and operated, the owner will determine how much to spend for his/her area. So, if we give general estimations, there is a good chance that we will over/under estimate their traffic by a lot. My guess would be that 70% of locations are rural, 20% are decent sized cities, and 10% are big cities.
Re: 1000 locations KW research
Hi Joe, that's some challenge you have there and unfortunately my reply is unlikely to solve your specific question.
The key issue here is that even if you were able to easily gather estimates for all these locations, that's all they'd be - estimates. In actual live situations CPCs and clicks can vary quite dramatically from these estimates, especially CPC since it's so subject to local variation.
How I would approach this situation is to try and educate the franchises to approach AdWords in terms of goals, testing and measurement. For example, I'd suggest each franchise set a target number of conversions, a specified period of testing - with a reasonable budget - and a process of evaluating actual performance against those goals.
If you say to Franchise A that you estimate they'll get 1000 clicks a month and their CPC will be $1, when they only receive 300 clicks at $2.50 they'll be upset. If you provide them with a planned approach to setting goals, measuring them and adjusting budgets accordingly the ball is in their court and they are responsible for their own decisions.
If you are managing their Accounts, you can still plan a testing period and encourage sensible spending - and by sensible I mean here enough budget to achieve close to a 100% impression share for the testing period. Franchise A may then say - "We'll allocate $2000 for a two week test, for which we're looking for 300 conversions." Franchise B can say "We'll allocate $500 for a week, for which we want 100 conversions" and so on. You will need to use your expertise to caution franchises against unrealistic goals and spend but I would steer clear of quoting specific CPC or click estimates.
Re: 1000 locations KW research
It seems like statewide data is fairly reliable, but if we did search data on some of these little townships, it would be off (if any data existed at all). We thought we would take the statewide cost/click data for each state and then weight the click data by population of the cities the franchises they are in. For example, let's say that Alabama has 2 locations (there are really about 25) with populations of 9,000 and 1,000. The statewide data is 1000 clicks for $1000.
We would add the two location populations (=10,000), divide the location by total population (1K/10K and 9K/10K respectively), and then multiply the locations by the click and cost data. The smaller location would have an estimated 100 clicks and the larger location would have 900 clicks.
There are several reasons why I don't think that this is ideal, but it is about 200 man hours faster than doing it location by location and trying to find the right targeting. The locations are all over the states, so we are expecting at the state level to get all those clicks; it is just a matter of how to hand out click estimates. The targeting is not going to overlap, so we figure that the finite clicks would probably go from most to least populated cities.
In the end, like you said, it is just an estimation (and we will for sure let the client know that). I think this might be a good middle of the road solution between accuracy and speed.