Here is David Rodnitzky's original presentation from SEMpdx in February 2013.

It's a bit hard to get, so I want to break this down for you because it's a very strong lense through which to view PPC. Both in theory and in practice. Here's a 5 minute video that explains the theory:

I like to explain the Alpha/Beta method with the analogy of a strainer.

You scoop out a pile of modified broad keyword dirt (beta keywords) and you shake the strainer to reveal the ROI positive search query gold nuggets (alpha keywords).

The only way to find out if a keyword or a placement will be profitable for your business is to test it. Here's how you test:

So lets say we're building an account from scratch. At first we have no idea what's going to work. We're going to build out what we'll call beta campaigns.

Sidenote: The only pre-requisit to all of this is conversion tracking. If you don't have conversion tracking running, stop reading and go get that done at all costs.

Beta campaigns are composed of modified broad keywords. We use these modified broad keywords to pull in a diverse, themed group of search queries. It's like an incubation unit. Its primary purpose is to discover winning queries which we will then transfer to an Alpha Campaign (gold nuggets). It's not meant to be a profitable campaign.

In practice, we'll be digging into the search queries inside of beta keywords and eliminating the clearly loosing queries and transferring the winning queries to an alpha campaign. The final step is to add an exact match negative of the winning alpha query to the beta campaign. We don't want Google to mistakenly trigger the beta modified broad keyword instead of the exact match alpha keyword.

As an AdWords account matures, Beta campaign spend becomes a good indicator of how aggressively an account is investing in growth vs. maximizing net profit. For example an account spending 50% of its budget per month in beta campaigns is looking to grow pretty aggressively. In comparison, an account with 10% of it's budget in Beta Campaigns is looking more to generate net profit at its current pace, perhaps to re-invest profit in other channels or other areas of the business entirely.

If AdWords spend needs to downsize, downsize beta campaigns first before any other campaign. They're not meant to bring in business, they're just a method to feed and grow your Alpha Campaigns. By reducing your Beta Campaign spend, it's as if you were cutting your AdWords R&D budget. You would want to leave the Alpha campaigns running as much as possible seeing as by definition they're the cream of the crop - profitable.

Ok so hopefully by now you get the jist of this Alpha/Beta mentality. Lets walk through a concrete example. Lets imagine we're selling online video guitar lessons.

I'm serious.

A quick, easy and effective way to do keyword research is to reverse engineer your competitor's keywords via tools like SEMRush or SpyFu. WhatRunsWhere is fantastic if you're running on the Display Network.

Don't get excited by the number of keywords/placements you can download and upload into AdWords. Ideally, you want to hand pick 10-30 keywords to test (depending on your budget). If you upload too many, and you're budget can't handle it, you'll be budget capped. Ideally, you don't want to be budget capped because AdWords will fairly randomly allocate your budget. Too much dirt in the strainer and the gold nuggets stay hidden. Often, for reasons unknown, Google seems to allocate most of your budget to 1 keyword when budget capped. It doesn't necessarily try to allocate your budget evenly across keywords.

So lets say that based on our keyword research we identified 6 keywords we want to test. We're going to be using modified broad, so our keyword list would look like this:

Now we want to group these keywords into themes. We would need more keywords for me to give you a telling example here, but for now we can imagine that we could group these 6 keywords like this:

AdGroup: learn guitar

AdGroup: online guitar

AdGroup: beginner guitar lessons

To get a head start, use the google keyword tool to see how Google would semantically group your keywords. But don't just copy paste what the keyword tool give you into AdWords though - it's only meant to give you some ideas of what could make sense for broad keyword themes. Always hand-make your keyword lists. It takes a bit of time, but it's very important.

These groups will be your ad groups. It's important to have relatively tightly themed ad groups so that we can write ads that are relevant to the user's query. High relevancy = high CTR = high Quality Score = lower CPCs = lower cost per conversions, higher ROI and all that good stuff.

If there's a big difference between groups of keywords, you can consider creating a different campaign for them entirely. For example if we also sold online piano lessons, we would definitely create a separate campaign for that.

We also want to break out campaigns by network, device and country (because as we know, AdWords averages lie). We don't mix search vs. display, US vs. UK and desktop vs. mobile.

Here's my recommended nomenclature to name your campaigns:

S | US | BETA | Guitar Lessons

S stands for search (the network we're running on - vs. Display), US for the country, BETA for the campaign type. Add a M or MOB if it's a mobile campaign.

Assuming we started with the US (which usually has more than enough traffic for a test) and then expanded to other english speaking countries, our campaign list would look like this:

You'll also want to tag the campaigns with either Beta or Alpha, that way, later, you can easily view stats broken down by all Alpha/Beta campaigns, like this:

I'll leave writing ads and adding extensions for another post. Once we're live, we need to wait for some juicy data to accumulate.

Once we have some data, we're going to go into the search query report (for search) and the "detail placements" (for display) to look at conversion metrics.

We may decide to pause a modified broad keyword entirely if the relevance feels low and overall performance looks unworkable.

Sometimes, you'll find that a search query or placement is eating up most of the budget of the containing modified broad, and pausing that specific poorly performing search query or placement shifts the budget to new, better performing, potentially Alpha material - so don't throw anything away too quickly, do some diagnosis first.

Lets assume that we're getting queries like "how to play bass guitar" eating up considerable amounts of our budget. We're not interested in targeting people who want to learn the bass guitar (totally different instrument) so we'd add "bass" as a negative keyword. I like to use shared negatives lists - explained below. We could call this list "General Negatives" and apply this list to all campaigns.

Some keywords should get traffic but don't. There are many reasons this could happen. Is the bid too low? Are its queries being triggered by other keywords? If you don't know what's wrong get on a quick hangouts call with Google Support, they're fantastic for this type of thing.

Lets assume we've found that the query [acoustic guitar lessons for beginners], triggered by the modified broad keyword +guitar +lessons +for +beginners in the US is really showing some promise.

We're going to want to transfer it to a SKAG in an Alpha Campaign. SKAG stands for "Single Keyword AdGroup".

So we're going to create a new campaign:

S | US | ALPHA | Guitar Lessons

Inside that campaign we'll create an AdGroup called "acoustic guitar lessons for beginners" inside of which will be the exact match keyword [acoustic guitar lessons for beginners]. We're doing this so that we can write an ad specific to that keyword to get even better performance and milk the keyword as much as possible.

In this case, the ad might specify that this course teaches you how to play the acoustic guitar in particular. We might also want to increase its bid to increase volume.

If it's a really high performer, we might even consider building a landing page (perhaps with a picture of someone practicing the acoustic guitar) specific to that keyword to further increase its performance.

The final step is to make sure that from now on, when someone searches for [acoustic guitar lessons for beginners], AdWords triggers our acoustic specific Alpha ad and not our more generic Beta ad.

To do this we create a shared negative keyword list in AdWords. You can find it in the shared library here:

We can call this list "Alpha Keywords". Inside this list we'll add our negative exact match keyword, like this:

The final step is to apply this shared negative keyword list to all beta campaigns. You do this below by clicking on the "apply to campaigns" button. It will reveal your campaigns, and you can select all of your beta campaigns and hit save.

If we don't do this last step, our campaigns will compete against each other to show their ads to search queries. This is bad because the data is likely to be really messy if the same query is triggered in multiple places.

And now you can structure AdWords accounts like a boss!