Staying Organised When Advertising

How to structure campaigns, why to not delete data, when to schedule reviews

This article is part of my Entreprenerd: Marketing for Programmers book, which is currently available to read for free online.

It’s altogether too easy to become overwhelmed with your online advertising efforts. The accounts you create on various advertising platforms gradually fall into a state of disorder and disrepair. Without active measures to stay on top of things, you’ll one day reach a point where you’ll dread modifying even the tiniest aspect. For me, it once got so bad that I would procrastinate maintaining my advertising accounts by instead filling out my tax return, which I considered at the time to be the lesser of the two evils. Predictably, this period of sustained neglect caused my advertising performance to dip into free fall.

Structuring and Labelling Your Accounts Sensibly

Online advertising platforms come equipped with somewhat standardised tools and terminology for structuring your accounts. Typically you have a couple of “campaigns”, which contain a few “ad groups”, each of which is itself composed of a selection of different “ads” and is “targeted” to a certain audience. Let’s continue by connecting the abstract structure just given to how it gets used in practice.

Campaigns

Starting from the highest echelon of this organisation chart, campaigns most commonly correspond to business goals.

For instance, an online retailer that sells shoes and socks would normally have one campaign for each major category of products: one for selling shoes and another one for selling socks. If the retailer starts stocking(!) some other class of products (e.g., shoe polish), then it just creates a third advertising campaign. Now, if the advertising for one product class tanks, the retailer can switch its budget to a more successful campaign.

The business goals of a two-sided marketplace (as expressed through advertising) normally reflect the priority such businesses give to recruiting people for both the supply and demand sides of their markets. For example, a startup building a marketplace for piano lessons would have one campaign to recruit piano teachers and another to recruit piano students seeking lessons. This way, advertising budget can be shifted towards the side of the market that most needs a boost.

Other businesses divide their efforts between branding and getting sales. Thus, a luxury watch manufacturer would have one campaign corresponding to gaining admirers and another to sell watches at their store.

The names of advertising campaigns should reflect the business goals behind their creation; for example, good campaign names for the above examples would be “Sell Shoes”, “Sell Socks”, “Recruit Piano Teachers”, “Recruit Piano Students”, “Build Brand Awareness”, or “Sell Watches”.

One advertising platform (e.g., Facebook or Google Adwords) is rarely enough these days. Instead, you will likely create parallel campaigns for the same goal with various other advertising providers. Make sure you keep the campaign names the same across all these platforms. If the campaign for recruiting piano teachers is called the same thing on Twitter, Facebook, and Adwords, it is much easier for you to turn it on and off, to change its overall budget, or to compare how well each platform performs at reaching it. Without this kind of parallelism in naming, managing campaigns spread out across many platforms gets thorny and becomes unacceptably error-prone.

Ad groups

Moving down to the middle-management level of account structure, we find ad groups.[^7-3-1] These typically correspond to two ideas that often show up inseparably intertwined: 1) targeting options and 2) finer-grained versions of the business goal of the ad group’s holding campaign.

With Google Adwords, advertisers usually create one ad group for each cluster of related keywords, such that different ad groups are triggered depending on the keyword someone searched for. For instance, a shoe retailer would have ad groups corresponding to each individual product (e.g., UGG boots or Birkenstocks).

Of course, some businesses only have a single product/service on offer, as would be the case with a furniture transport company. Its ad groups should instead correspond to specific customer concerns as ascertained through the keywords their customers typed. Google queries about “insured furniture transport” ought to trigger ad groups highlighting the company’s comprehensive insurance coverage, whereas queries about “last minute furniture transport” should trigger ad groups that emphasise the transport company’s quick response time and punctuality.

Due to differences in available targeting mechanisms across platforms, Facebook’s ad groups (or “ad sets”, as they call them) may instead be associated with targeting options other than keywords, such as the social network’s members’ relationship statuses or the subjects they studied at college. As such, the retailer selling university textbooks might create an ad group corresponding to each individual field of study—one for math, another for law, another for medicine, etc. Through using the available targeting options, the retailer breaks down their campaign goal into focused subunits encapsulated in ad groups.

With this connection between ad groups and targeting in mind, I believe the most helpful way to name ad groups is according to their targeting. For instance, a language translation service would have ad groups named “Legal Translation (T:keywords)”, “Movie Script Translation (T:keywords)”, and “Interpreter Service (T:keywords)”.

Notice how I included the mechanism of targeting within the ad group names, communicated with “T” (for “targeting”) and followed by the specific targeting mode (“keywords” in this instance). The extra targeting information is helpful to have in the name because otherwise we’d be unable to tell the difference between ad groups that advertised the same thing via different targeting options. Suppose we advertised law textbooks through three different targeting mechanisms: one that targeted people who studied law at university, another that targeted people who interned at law firms, and another that targeted people who wrote about law on their timelines. Without differentiating between these targeting options within our ad group names, our performance reports wouldn’t communicate how effective each targeting option was at its advertising job. But this is a critically important data point in marketing intelligence. The choice of targeting mechanism is a huge determiner of advertising effectiveness. Perhaps targeting law students was a waste of money, but targeting law interns was pure gold. Without the suggested naming convention, the only way we could answer these basic questions of effectiveness would be to open up the settings of every ad group on analysis day and painstakingly compare targeting options so as to determine the differences. A little upfront organisation would have saved us a whole lot of chaos.

As with campaigns, it’s important to work on keeping inter-platform and intra-platform consistency in naming ad groups. There’s not much to say here, other than advising advertisers not to guess what they called an ad group in another platform, but rather to go back and look it up.

In this section I showed you a slightly idealised scenario. The truth is a little messier, mostly because Google Adwords fails to provide a clean separation between campaigns and ad groups along targeting lines. The problem is that most of the targeting options in Google Adwords are set at the ad group level (consistent with other online advertising platforms), but some targeting options are set only at campaign level—for example, geographic region. If an advertiser on Google Adwords wishes to show different ad groups to people in different locations (e.g., “UGG Boots” only to USA residents and “Birkenstock” only to German and French residents), then they would have to have two separate campaigns. I would advise the advertiser to name these campaigns so as to communicate the location within the name (e.g., “Recruit Piano Teachers L:USA” and “Recruit Piano Teachers L:Germany&France”). (“L” denotes location.)

Ads

At the lowest level of the campaign structure are the individual adverts. These contain the actual creatives that the public sees. Every ad group must thus contain at least one ad, otherwise there’s nothing to display when the ad groups get triggered. But I wouldn’t recommend putting more than one ad in each ad group, unless you are consciously serving one of these two purposes: 1) protecting against ad-blindness by refreshing your creatives, or 2) optimising the creatives themselves (as opposed to optimising your targeting, which is done at the ad group level, and which generally gives you more bang for your buck).

Not all platforms let you name your individual ads, but on the ones that do, use the opportunity to communicate how the ads are unique with respect to their headlines, bodies, or images. I have a naming convention whereby “I” stands for image, “H” for headline, and “B” for body. By following it, I named one ad “I:coffee-face,H:better-grades,B:instant-download”, another “I:clever-cat,H:better-grades,B:instant-download”, and another again

“I:coffee-face,H:better-grades,B:free-samples”. This naming convention helped me notice patterns about what images, headlines, and copy bodies beat the averages, even when I had hundreds of combinations to compare.

Don’t Delete Data

Once a marketing campaign has run its course, there’s a temptation to delete it on the platform hosting it so as to declutter your account. Resist this temptation. Whenever you delete advertising data, you also delete the original records and results of past advertising forays. You might believe that this data contains no further insights, but this would be a premature and somewhat self-centred conviction.

The conviction is premature because future improvements in your skill as an advertiser (or future enhancements to the reporting capabilities of advertising platforms) may enable you to learn new lessons from old data—assuming that old data is still around. Furthermore, deletion of data is self-centred because someone else without your direct experience may one day inherit your advertising account. Without your old data to go on, your replacement may be doomed to repeat many of your past mistakes.

Many advertising platforms let you hide deleted campaigns from the UI, which should be enough to declutter your workspace without the need to permanently remove the data. Where this hiding feature isn’t available, there may be the possibility to export your old data for future reference. With a backup handy, there is no issue at all with deletion.

Change History

When the engineering team modifies the website source code, they rely on specialised “source control” systems (e.g., Git or Subversion) to preserve a precise record of how the code used to look and what their changes were. (These source control systems are like the revision history features of Wikipedia or Google Docs, except pumped up on state-of-the-art feature steroids.) Thanks to source control technology, your engineering team can surgically roll back problematic changes whenever a bug is detected. Instead of struggling to remember how the previous code looked, they see a side-by-side comparison on their screen.

Sophisticated online advertising campaigns can reach complexity levels that rival or even exceed the source code of the websites they promote. The slightest change to a campaign can upset a delicate advertising balance in ways that only become apparent weeks after the incident. As such, there is a parallel need for source control technologies in online advertising. This, luckily, is starting to be provided by some of the big platforms. Google Adwords has an account history feature that lists all modifications made by any team member (complete with 30-day undo),^7-3-2 and Facebook has something similar (minus the undo button).^7-3-3

When these features are not available on an advertising platform (or when they are but you want to supplement them to be thorough), I would advise you to create your own change log for jotting down the most momentous of changes. Within this log, you would note down not only the changes made, but also the reasoning behind those changes. Without some means of retracing your steps—both physical and mental—there’s a risk that your advertising campaign will one day cease to be effective and you won’t have a clue how to repair it. The log will then be to your advertising campaign what the trail of breadcrumbs was to poor Hansel and Gretel.

Schedule Review in Future

Always, always, always schedule a future review for checking up on your online advertisings changes. As you close the tab for today’s session with your online advertising platform, pull out your calendar right away and book a review date with yourself a week or so from now. This isn’t negotiable—the changes you just made to your campaign will affect it in ways that you can’t possibly anticipate, no matter how good you think you are. Without reviewing the changes, you will wrongly presume that everything worked out as expected—even when it didn’t.

For instance, I’ve made “improvements” to my Adwords campaigns, presuming they were for the better, only to log in months later to find that I’d chewed through thousands of euros in waste. Don’t be like me.

[^7-3-1]: Ad groups have been around for a while in Facebook and Google Adwords, but are relatively new in Twitter. At the time of writing, you couldn’t modify Ad Groups in the Twitter UI, and instead had to modify them through exported and imported spreadsheets. I would guess that Twitter will update their UI by the time you read this though….


More Articles: Click here for full archive

The Bug Slip

A bug hunting ritual that makes things a little easier


Silent Failures Must Be Given a Voice

The most dangerous bugs are those you remain chronically unaware of–as all too often occurs with failing scheduled scripts


The Conversion Surgeon: A Guide to Conversion Rate Optimization

A High-Level Catalogue of The Main Areas to Fiddle With When Optimizing Your Conversion Rate