Marketing is sales
I am a firm believer that the main role of marketing is to drive sales (as long as you represent a commercial company). Yes there is more to marketing than just sales, like branding, positioning, segmentation and so on. But these parameters are not the end purpose. They are marketing tools you use to achieve the end goal – sales.
One important aspect then is how you measure your success often called Return On Marketing Investment (ROMI) or Return On Investment (ROI). I still get surprised how many marketing professionals who say they don’t believe that the effect of marketing can be measured, especially against sales achievements. I have a previous blog post about this.
For simplicity the rest of this article makes the assumption that you believe it’s possible to measure the effect of ROMI (on sales). At least on a level of believing that it is better than just using your gut feeling.
Importance of measuring effect
In the last couple of years the importance of measuring the effect of marketing on sales has grown. During the resent financial crises the pressure on all business units, including marketing, to provide proof of how they contribute to the company bottom line has increased. And it’s a well known fact that “when the going gets tough”, the marketing budget is one of the first to get cut. Last year Harvard Business Review had some articles about why many big companies don’t have a marketing director as a part of their top management group. And in one of their podcasts they went as far as saying that a major reason is that marketing is viewed as a “soft” discipline and their contribution to the bottom line is so difficult to measure.
The measuring strategy
Basically there are three, somewhat broad, topics you need to consider when working with measuring effects, and these are:
- What to measure
- How to measure (methode)
- How to interpret
- What to measure
What do you want to achieve? Setting clear objectives for you campaign or activity will make it much easier to measure later on – And according to Peter Field and Les Binet “campaigns that set clear campaign objectives are more effective than those that don’t.” (Marketing in the Era of Accountability by Peter Field, Les Binet – IPAdataMINE -Learning from the IPA effectiveness cases 2008)
Other learning here is that the benefit is greater the more hard objectives, prioritized, there are (with some limitations).
Their hierarchy of hard objectives is as follows:
1) Business objectives
2) Behavioral objectives
3) Attitudinal objectives
Hard measures should always come first, and the harder the better. Measure of financial return is the ultimate evaluation measures for all commercial campaigns.
If you have set up your campaign goals in the right way, then you have established what effect you are aiming at and so setting your measurements is more or less done. Many posts on this subject mistakenly just use the immediate effect of traditional marketing and Social media as the effect that should be measured. Be it traditional measurements for TV, Radio or newspapers, or number of likes, friends or followers in social media.
A recent post on Mashable asked three experts about how they measured Social Media Marketing Success. And because they mainly were representing digital agencies the answers were based mostly on how their customers measure their (the agencies) effect, so the answers got a bit abstract (as one of the comments points out).
All of these measurements are important and fairly easy to measure. But it doesn’t really say that much about the effect on business targets like, profit, market share or sales. In general what we do is using these measurements and if they exceed our expectations and there is an increase in sales during the period of the campaign – we assume that there is a connection.
The problem with this assumption is that there are a lot of factors that affects sales and you can’t say anything about the effect of marketing in this mix until you understand how strong the other factors are and how influential they are.
A classic example: An ice cream producer is getting ready for the season, but decides to start earlier to see if they can expand the season. So they start a marketing campaign and distribute the ice cream earlier than they have done before –and before their competition. The sales goes through the roof (compared to the year before) and the conclusion is that starting the season earlier was a very good idea and that the campaign did a very good job, since that was the only thing they did differently than last year. So they decide to do this again the year after, but then the sales are slow, way below the previous year. It was a copy past of the last year campaign so they hadn’t changed anything. What went wrong?
After some research they found the following explanations: In the record year the temperature was much higher than normal. None of their competitors started earlier last year, but they did this year, based on our success last year. The price could also have been changed between the years.
What we see is that many factors (important factors when it comes to sales of ice cream) has changed and might even have been a more important reason, than marketing alone, for the increased sales the previous year. So it turns out that it’s impossible to measure the effect created by the marketing campaign alone. Or even to say that the first year’s campaign was a success compared to the following campaign.
2. How to measure
There are no shortcuts on how to measure the marketing effect. Like so many offers you see on the net – if it sounds too good to be true – it probably is.
In my experience you can come a long way with just starting with small measurements that you can check yourself – but at the same time be aware of the limitations in doing it this way.
The only viable way (and that your management group will have a hard time fighting, and they would understand), is sales modelling (or econometric modelling)
Econometric modelling will isolate the actual contribution of a factor (in this case marketing) to sales, adjusted for all other drivers.
Step 1 (Identification of goals and needs)
- Understand the KPIs and market situation
- Identification of goals and needs
Step 2 (Data collection)
- Data Brainstorm
- Data collection and validation
- Feasibility study
Step 3 (modeling)
- Construction of models
- Test of models
- Calculate the contribution from the significant drivers
- ROI calculations
Step 4 (Implementation)
- Convert the results from the model into strategic recommendations
- Integrate to use the model for optimization
- Test of planning-scenarios
Step 1 – Identification of goals and needs.
Here you have to identify what you want and what you have to measure. Since we are talking about ROMI it would be obvious to define the goal as “how much of the total sales can be connected directly to my marketing activity”.
Based on this need you should start to identify the most important drivers that affect sales. Start with a brainstorming and build models like the one below (this will differ between companies, products and market segments):
This model will be a mixture of general- , industry specific- , micro- and macro effects. The importance of each of these might also vary, between industries and products and organisation.
If you sell one product to all target groups you will only need one model. But for most companies the customer groups are divided into target groups with specific products towards each group. Then you might need one model for each of your groups.
As an example: We (a business school) target three segments. The bachelor students (young students starting their higher education for the first time between 18-22 years old). Master of Sciences (young people with a bachelor degree between 20-25 years old). And the Executive market (Adults needing additional education between 25-45 years old. So we operate with three different models.
Step 2 – Data collection
Now you have your list of drivers, so now you need to gather data about them. This could be your marketing spending, macro data, often from governmental sources, competitor data like marketing spending. Basically quantify everything that affects sales.
If you just have data for one year this will limit your models accuracy, so if possible you should gather data for 4-5 years back in time. As an example: We managed to gather solid data for 4 years back in time.
It goes without saying that the quality of your data will influence the quality of the analysis. You probably know the term “Garbage in – Garbage out”.
Econometric models is build with non-linear multiple regression methods, and once the different factors has been weighted and fed into the model you will get a prediction of how accurate it is. This accuracy is often expressed as R2. Say you have an R2=70% (0,7), which means that the factors you used in the model explains 70% of marketing’s effect sale, and the higher R2 the better.
Then new factors can be added until you reach the highest R2 possible. Only factors that has a high effect on the R2 are included so the factors that you put in that don’t alter the R2 is therefore removed. The goal is not to have as many factors in the model as possible. Only the factors with a significant affect will be used.
This will also tell you if you got the right factors. With a very low R2 there is factors out there that affects your sales, but you are not measuring them in the model.
The R2 will vary between different industries and product, depending on the complexity of the market and product. For our part, “selling” Higher Education, we managed to get an R2 between 86-95%. We achieved the lowest R2 in the Executive segment/model, which makes sense since this is the most complex target group.
To get so good numbers in a segment that can be described as: A complex, high price, high involvement service, is very good.
Step 3 (modelling)
This step is done by the agency used for the modelling. What you get back is a full analysis of all your models. But still you as the costumer have an important role in this process. This is where the statistical geniuses meet the market experts.
It’s important to test different preliminary findings in the modelling and check it against common sense, and your experience in the marketplace. If something sounds way off, it often is. Then one has to redefine the factor or the measurement of the factor until it makes sense.
Your model could look something like this (this is our executive market):
Step 4 (implementation)
This is where you decide what action to take viewed against the modelling. This could mean a different market focus when it comes to channels, spending level and so on.
It’s at least three levels you can implement the results of modelling into your organisation:
- Here you use it as a campaign management system. You change your media strategy based on what gives effect and what don’t. Could be changing you message, timing of the activities or the level of investments.
- In the bigger marketing picture you can begin moving money between your different activities (media spending is just one of the measurements). More on direct sales, DM, social medias and so on
- On the business level you can use it as a sales forecast and building different scenarios like “…what if?…”
In our case our Media Company handles the first level. They come up with new media plans based on the learning from the modelling. On the marketing and business level we do the analysis our self and running scenarios.
The big question is of course how Social media measurement fits into all this.
And the answer is; it depends…
If you are doing Social media activities to support your sales, it is possible to implement it into sales modelling. What needs to be done then is to see how the effect is measured. How is a sale connected to an activity? The easy answer is that timing decides. So if you start a campaign, lets say on a Monday, and by the evening your sales rises above the base level (which is a important metric in sales modelling – it explains what level of sales you would have without doing something (in the short run)). Then it’s normal to assume that this extra sale is motivated by the campaign.
So if you use Social Media in a way that it can fit into this system this activity will be just another factor.
The problem lies in the fact that social media activities not normally has sale as it goal. A more used goal might be “create engagement” around my product or brand.
This has a longer time horizon than short-time sales increases, and would not be easily measured in the example above. But it is possible. The model would then measure engagement instead – not sales. What you need then is to find a good measurement on “engagement”. One way of measuring it could be to use the Klout score.
This article has been written with contribution from our agency, Brand Science who did the modeling for us, and our internal analyst, Laurie Bloome Jacobsen.
This article is an extended and more detalied version of a presentation I did at a ROMI conference in Oslo, Norway.