Using Facebook to Win!

Do you think we should be using Facebook more aggressively?  Let’s eavesdrop on a conversation between some people who do.  First some basics:

Let’s start with the rough and ready approach:

Remainer 1

As I said the other day the best approach is to find the local community group on f/b and post something tame, then let the leavers rant, prod them occasionally with facts to keep it going. This means the post stays at the top of the page and more people see it. Then you just harvest the remainers. So for example I posted Thursday about the march, jesus the abuse but… 23 people have liked and there are >18 comments. Most comments -ve but a few remainers sticking their head above the parapet. I will then look at the likes and have a look at their profiles and then send them a DM asking them to join a local group. So for Cars for example, set up a dummy profile, ask to join the “There’s no place like Halesowen” group, leave it a week and then post, “I’m really concerned about if JLR will scale down production more or move to Slovakia. You will get a lot of “you silly xyz,”  etc but people will like the post and some will agree with you. Then do a second post a few days later and correlate the likes and hates and you will start to map the active users and their views. You can then invite the people who are worried to a special group you have created. “Halesowen what happens if JLR leave”. You can then use this group as a market research group. test out messaging get people to discuss their hot buttons and then use those to target posts in the full Halesowen group. And repeat…. It is what CA were effectively doing but a) legal and b) free and c) yes slightly devious but…

Remainer 2

I think this is a really useful suggestion. Obviously it is time and resource heavy so would need a hell of a lot of staffing but if someone is willing to take it on that would be a fabulous prong in the multi pronged attack. But who can coordinate this? The Facebook methodology was also well used by Cambridge Analytica on both elections. I understand that they were running out around 7,500 message variants a day. Big political campaigns test and refine from a bank of messages running well into six digits.

and moving on to Facebook advertising:

Remainer 2

You can run ab style testing where you give Facebook maybe 30 or so variants of an ad. What we typically do is start with a table of lead messages (the copy hook), sub headers, calls to action (you must have calls to action eg click to xyz, click if you xyz) and several images. You set the ad up so that every possible combination of words and images starts out in the test. Facebook will start to automatically show the most successful ads more and hide the less successful ads. This is where you need to be canny though. The less successful ads may still offer you powerful segmentation data and this is very much worth whacking into a spreadsheet for analysis. For example, you may notice that your less successful ad actually seemed to attract clicks exclusively from older men with higher educational attainment. On closer inspection you could find that group in a minority within those responding to the most successful ad. Don’t throw that low performing ad away, but target it only at the segment that we’ve seen from the test phase will respond to that ad. This is how you do the targeting. Then run ten copies of that ad only within your narrow well defined target audience but now with small changes to wording if you want to further refine to get the most powerful ad you can to that audience.

Remainer 1

I wish we could have got everyone there yesterday organised into their local groups. It is only people on the ground who will know the local hot buttons, central “scare” messaging doesn’t work, x local jobs lost, their favorite bakery, Local Mway turning into a lorry park etc that gets people interested.

Remainer 2

Yes that is very important indeed.  Local advice should add to this process.  But not replace it – ie they may say that certain scare tactics wouldn’t work but you could run a threatening ad and find it performs well.

But yes they may say xyz is massive in rugby and you would never have known it otherwise.

One other thing on the subject of cars in rugby.

There’s some very juicy data points on the car industry in some reports I downloaded. Typically, emotional messaging is key. But I’d still throw in a couple with data on eg “Car production down 18% in the UK this year. What can you do to safeguard your future?” – you may find something more number-led appeals to middle management level employees for example.

It’s worth knowing that the text that goes above and below a Facebook ad is actually put into a special form in the FB ad generator so you don’t need to include it in your image. That way you can create that matrix test format where different images are matched with different headlines and different calls to action. You’re looking for the message and image magic combo. So I would be advising that you test more calls to action than just the one.

Testing and refining online is ongoing and constant. It’s not like you test then launch. It’s part of the campaign management process. The important thing is to develop some granularity about whom you will target with which piece, which should become apparent during testing. At a crude level, you should begin to spot differences between old and young, male and female. You then save the money by specifically targeting the ad that appealed to older males only at older males so you optimise the performance of that ad. But you then create four slight variations on that ad and continue to refine based on performance. But the important thing is to look closely at the many parameters you have access to in your digital targeting to define your target audience for each ad in increasingly tight delimiters. This is a micro dramatically scaled down version of what Cambridge Analytica were doing.