The odyssey of measuring a brand campaign
My good friend and online expert Daniel D. is working on a framework of KPIs to measure brand campaigns. We’ve been talking a lot about that in the recent days, so I decided to capture some of the thought we came up with. My intention is to give my two cents, especially with the “but how can we measure that?” part.
Let’s start tackling the problem from the very existential perspective…
What do we run a brand campaign for?
Big Bugdet Owners get scared when you tell them that you are planning a brand campaign… What they understand is basically a pretty expensive media spending campaign but whose result cannot be measured in sales uplift. Moreover, they remember the last time they asked for results and got some cryptic KPIs with fuzzy results back.
Basics first! You are running a brand campaign because you got money and you -I hope so- got some evidence that investing this money in running a not-strictly-trade campaign is going to pay off good returns. Full stop. But how?
I’m a big fan of the simplicity. Breaking down complex problems into simpler pieces is the best way of understanding the problem and providing actionable solutions with reduced scope. Let’s examine the ultimate goal of a brand campaign
You run a brand campaign…
To reach the maximum number of people you might be relevant for
What for? To attract them, to invite them to engage with your brand
What for? To make them aware of your brand proposition
What for? To improve the positioning over the competition for a product or a service….
We could continue with a few more “what for?’s”, but you get the idea… Now that we are clear about the purpose, let’s have a look at the ideal result of a properly run brand campaign to see where the changes manifest and where we have the opportunity to quantify the outcome:
The way a (good) brand campaign manifests
You might have a hard time attributing sales uplifts, call center call reduction, churn decrease, or whatever your goals of your campaign are. The effects might not be immediate, might be blurred by the next campaign, might not come at all… Yet there are many places where the effect of a properly executed brand campaign manifests. Let’s have a look:
(Brand) Traffic increase
Congratulations, you managed to attract more people to your dependences (online or offline) … After seeing your ads they are literally coming for more! This is the materialization of the reach.
How can you measure that? Well, you are facing here an attribution problem… You are interested in understanding “What’s the proportion of my visits (or even better visitors) that are coming because of the campaign?” Even if it can be modeled with a baselining exercise (what would it look like without the campaign) and the uplift vs. the baseline, you can do much better… put the required mechanisms in place to flag the campaign traffic (offer a dedicated landing page, a campaign banner on your home, a dedicated GO-TO url, etc.) Obviously it’s more difficult if you want to measure footfall increase in your stores, but even if that’s workable at a lower precision level (shop agent data gathering, coupon redemption… or even placing at the store entrance a one-click device to identify the ones that came because of the campaign -similar concept like the happy-or-not feedback gathering device you can see in the Heathrow airport-, etc.) you are going to have a hard time justifying the ROI of the required investment, so let’s focus on the online domain for the time being.
With your online marketing hat on, you are going to tell me that traffic is not enough, you need high-quality traffic. One step at a time… First of all, try to separate the campaign traffic from your natural traffic, then if your brand campaign was meant to address particular segments, try to classify your visitors accordingly… if your campaign is not focused on any particular segment, try to discover them!
Impact on behavioral metrics
As I said, I’m a big friend of simplicity, and for capturing customer value behavior, I really like the Recency, Frequency and Monetary approach. An (again) properly run brand campaign should result into higher Recency -number of days since the last visit per user-, higher Frequency -more touch points- and as Monetary, as we the primary intent is not expressed in number of sold items, I suggest taking for example Dwell time or Avg number of pages per visit for online. Measure behavioral metrics for physical stores is way more challenging and only feasible using Wifi Mac Address tracking, Cam-based solutions, etc (worth seeing following video)
What to me works best is an aggregated of all three components (let’s call it engagement index), which we baseline over time to quantity the uplift attributable to our brand campaign.
Echo in the Social Media
Social Media is the space where you get really measured… where you constantly get unfiltered feedback from your existing customers, from the customers you lost, from your prospect customers that are about to acquire one of your products or services, from the specialized press… actually from everybody. Of course a properly run brand campaign makes no exception here… but how does it echo in the social channels?
- Your Youtube channel experiments an unprecedented views, likes, shares or even dislikes increase.
- The Youtube videos from the band who created the music of your TV spot increase their daily view number
- Someone not related to your company uploaded to Youtube a video called “[Your_Brand_Name] TV Ad”.
- The number of Twitter mentions over time increases vs. the baseline.
- The weekly number of new Twitter followers of your company’s official account goes up.
- The weekly number of new Facebook followers increases.
- The share of all Social Media interactions related to your campaign vs. all created on the market increases.
- Specialized blogs devote a post to give you a kudos or to take your campaign into pieces
- … and a quite long etc.
And do you know what? Everything is measurable and with the proper data science behind, even pretty accurate. Your ToDo is to define your metrics and to put the right mechanism in place to compute the uplift vs. the baseline.
Your search share increase
Pure brand searches experiment a significant uplift, but this is relevant for you to certain extent only. What you need to monitor are the “[Product_Name] [Brand_Name]” combinations for your brand and for the competitors’ brands (like in both examples below, Lufthansa + London or Air Berlin + London).
Depending on whether you run your campaign country wide or regionally, you should try to analyze the outcome at different geographic granularity levels (nation wide, region wide, top 10 cities, etc). As you can see in the Google Trends charts below, the results can be very different for Berlin compared to Bavaria. Do you know what this peak in May for the Bavaria chart is? I’m sure you do if you are a football fan! Otherwise find it out here
There are other ways of measuring your share, for example based on daily or weekly mentions in the news, in the social media channels or number of natural interactions with your publicly available media assets.
If you have done a decent job, you might also observe a pretty notorious increase of searches related with terms that are not strictly related to your brand, but to your spot (e.g.: think of campaigns like the singing cat from Three in UK back in February 2014, the searches for “singing cat” increased massively but not in combination with the brand name).
In general terms, your search share is just a particular case of something much bigger… As the result of a successful brand campaign, you gain positions in terms of relevance vs. your competitors… If a customer needs a product or service that you and three or four other competitors are offering in the market, you need to understand where you are in this particular ranking and measure to which extent your campaign made you climb up how many positions… Not an easy one!
Your little steps toward a bigger goal
If you are in the online marketing domain, you probably know what I’m talking about… yeah! the so called micro-conversions or
A micro-conversion metric shall encapsulate the main reason why your campaign was created… Remember! It is not a trade campaign! The traditional conversion rate metric doesn’t work here because your campaign has not been designed for conversion optimization!
Think off-portal… say you display rich-media adds… a micro-conversion might be defined in terms of dwell-time, in terms of seconds viewed if you display a video. Think on-portal… your campaign landing page… you could set up a metric for the share of visitors who scrolled down to read more content, or measure the time on page, the content consumption, the net promoting index, or even created simple call to action and measuring the CTR… do you see where I’m coming from?
If you are using a cross channel tracking tool -you should!-, you could go and define micro-conversions in terms of assisted macro conversions for digital assets…
Further ideas for micro-conversions, depending very much on the purpose of your campaign, could measure the uplift vs. baseline of
content views / downloads, content/channel subscriptions, ratio of post to comments/replies, etc.
Your engagement (with Avinash’s permission) increases
If you define engagement by increased number of new visitors reaching your site (see point 1 about reach) or increase number of touch points per visitor (see point 2 about behavioral metrics), you got your engagement! So, YES, you can measure the engagement, and YES, you can create a combined metric to quantify it, but there’s nothing new on top of what we already discussed. Moreover, complying with Avinash’s discussion about sites’ particular existence purpose, you could even model your engagement by means of a set of micro-conversions (see previous point)… So nothing here to look at!
The not that tangible effects
That’s the crucial part but where you are going to struggle the most… After a properly run brand campaign you should see a reduction in the sales acquisition costs… For example, the cost per qualified lead should go down, as well as the cost per contact or cost for customer acquisition. You could go even further and say that the cost per click through should experiment a significant decay. Obviously, measuring it requires a lot of modeling decisions based on assumptions that are sometimes poorly implemented in the measuring and controlling systems, but that’s what makes it so attractively challenging.
Putting all together
If you came to this point, you are probably aware of the measuring possibilities for a brand campaign. Yes, you can define a lot of metrics in many areas and you can measure the before/after, the change vs. baseline, perform time series analysis, etc. In order to make it manageable, you are better off with a combined KPI, where each and every metric we presented before is weighted by its contribution. Finding the aggregation function is not a trivial task and no sure formula exists for that.
Considerations for this model:
- Think of the comparability. Performance cannot be established in absolute terms, but related to previously executed campaigns that you take as reference. Your model shall comply with this requirement
- Different brand campaigns should have specific micro-conversions goals, that can be individually modeled or previously aggregated into a factor
- You need to be constant in your assumptions, otherwise the comparability might be affected
- It’s a good practice to play with different models to find out the one that better adjusts for your business.
It’s really difficult to summarize into a few simple sentences the breath of ideas captured in such a long post, but here we go:
The output of a brand campaign is measurable and quantifiable. You just need to establish the right metrics where the effect manifests and create a clever model to combine these metrics into a single KPIs. There’s no standard applicable model so you need to iterate and find the one that works for you
P.S.: I take a TODO for me, which is to summarize the set of statistical models and algorithms you might need in your modeling exercise