We’ve been testing the performance of Facebook advertising on our
Facebook optimisation platform, and how it performs against search for a
test sample of brand clients.

We did this by running two simultaneous
campaigns across search and Facebook for each client (both campaigns are
designed to work together, with a similar message and content). We’ve
measured the impact of each on conversions (predominantly sales and
registrations) on each brand’s website.

The first step was to build and optimise the brands’ search campaigns across Google, Bing and Yahoo!. Then, we built and optimised Facebook campaigns designed to achieve the same goals. This lets us look at how Facebook and search campaigns perform together, and against each other, and what the different benefits are of each.

Our goal for these test clients was to distribute spend across each channel in the most efficient way possible, to effect the most conversions at the lowest possible cost-per-conversion.


Of course, each campaign is built to the specifications of each client. For example, search keywords relate to the types of deals that the client offers. If the brand is a retailer; the keywords will reflect this season’s offers, sale prices, limited offer availability and so on.

This is done using our Advanced Campaign Management system, which connects the search campaign to a data feed provided by the brand, so that the information in the search ads is always up to date with the latest inventory or deal information.

The most effective use of Facebook is to focus campaigns on a handful of specific deals (ads are created using our Facebook Ad Builder, a tool we have developed to create ads in bulk.

We use an ‘auto-split’ function to divide the ads to show different versions to individual groups of Facebook users – by demographic, interest or geographic region, for example.) Our automated bidding technology assesses the performance of each group and then bids an appropriate amount to target it.

Initial test result: how the two campaigns perform against each other

Once the campaigns are up and running, and optimised to the most efficient level, we can analyse the spend distribution across the two channels. A simulation model shows the performance of each campaign at various spend levels, based on the models that our bidding technology creates, for each campaign.

This examines how each campaign measures up against its original objective (in our test cases, to get the most effective cost per acquisition, rather than maximising the reach of the campaign).

We are then able to compare the CPA results for a number of different spend levels (for example, how each campaign performs on 20 different spend levels between £100 and £5000, depending on client budgets).

We use our optimisation technology to calculate the CPA across each spend level. The test results so far have been interesting:

  • For brand advertisers spending around the £1,000 mark, the CPA for Facebook campaigns is marginally lower than it is for search.
  • For higher-spending brands (those spending more than around £2,000), Facebook starts to give a lower CPA rate, and so becomes the more effective channel to meet the CPA goals.
  • At this level, the optimum spend ratio between the two channels is between 60 and 75% Facebook, and 40-25% on search.

These are initial tests, and we need to do more to establish whether they will be reflected across all brands, but they do show that Facebook is, for the first time, starting to perform as an acquisition or sales channel, and is a real challenger in performance marketing.