Sunk Costs: throwing good money after bad

I ski, but I am not a good skier. I do not take lessons as I am just there to enjoy myself. If the weather is a bit yucky I stay in the hotel and read. I am generally the only person in the hotel; everyone else having gone out because they have spent the money is determined to ski. Often they come back and tell me that they didn't enjoy themselves. When asked why they went, they refer to the money they have already spent.

The money they have spent on the skiing trip is what is generally referred to as a sunk cost. That is the money has been spent, and going skiing will not get it back. Not going skiing will not get it back either. The money is spent, sunk, end of story. I have a better day than the people skiing and my happiness (or utility as the economists have it) is better optimised. It makes you wonder why they want to throw good money (well time is money) after bad.

Retail Media
So why is this relevant to media? Well, as Wikipedia would have it:

The idea of sunk costs is often employed when analyzing business decisions. A common example of a sunk cost for a business is the promotion of a brand name. This type of marketing incurs costs that cannot normally be recovered. It is not typically possible to later "demote" one's brand names in exchange for cash.

The idea of being able to demote your brand to get the money back if your promotion was not as successful as you had hoped is lovely but is unlikely at the moment. There have been cases of agencies who can measure their success linking their profits to success, which is nearly the same thing. As this is not the norm we have to think that we will not be getting our money back, and so in order to prevent throwing good money after bad we need to get as much from our marketing as we can.

In many promotions there is an ongoing commitment based on the assumption that the promotion will be a useful spend. There are very few single cost campaigns. Even the term campaign implies use of multiple forces. Your in-store spend, your online spend and your external spend should all be part of the campaign. This leads to a time when you should terminate the spend for a campaign that is just not working.

The sunk cost dilemma
This has general applications to projects of all types and so the sunk cost dilemma is worth a mention here. The way the dilemma works is based on the same reasoning I use when skiing: What I have already spent is irrelevant, it is what will get the best return for what I do now that should determine what I do now. This falls down over long projects because it fails to take account of the additional information about potential success contained in the history.

Let us say there is a project that is projected to make $100m with an initial outlay of $1m. The project is launched and the £1m spent. The project hits problems and an additional 500k is required and the expected return is now going to be $500m. The odds still look good, we still stand to make a hundred fold profit. Then the project hits problems and an additional 100k is required and the expected return is now going to be $10m.The odds are still looking good and if we do not take the history into account we will go for it. This can keep on with the spend building up until someone takes a look at the project history and realises that while the odds look good they are not actually good. Each individual decision is a good decision as the future return always out weighs the required expenditure to a sufficient degree to make it a good project to invest in.

So what can we do? Measure!
Coping with this kind of situation means continually measuring all campaigns so that we can use the ongoing information to produce better promotions in future and terminate campaigns that should not be happening. It also means ensuring that the person who was responsible for the initial spend is not the one to make the decision about putting further funding into the same promotion.

There are two different ways this can be managed practically:

1) Have a peer review panel examine the promotions effects to make the go-no go decision.

2) Ensure that success and failure criteria are set up before hand with rigorous evaluation standards. These can be set up by the same person as long as it is done before hand while there is no sense of responsibility for the sunk cost.

The requirement for decision makers to have no sense of responsibility for a past event is supported by a fair amount of research that shows that people are biased to support things they feel they have started. This is true whether it is belief in a horse you have bet on or belief in a project you funded. Interestingly people are biased even if they did not actually agree to in the first place but have been since told that they were the one responsible.

In the current situation the ability to measure retail media is one of the arguments for allocating spend there: you can stop wasteful spending sooner and so stretch your budget further.

Rufus Evison

Black hat hypothesis testing and retail media

If you are a CMO and you want to keep your job then Christmas can become an opportunity to shine. In the current climate it is important to make every penny count and so you want to target your spend appropriately.

If you are an unscrupulous CMO then it can be an opportunity to appear to shine. In the current climate you want to show what you did as having worked whether it did or not. Today we are going to look (but not in too much detail) at ways that people can, and often do, use numbers to show things in a better than reasonable light.

Why is this worth doing? Not because we actually want to use any of these techniques. Quite the reverse in fact. Each of these is based on a fallacy that people use unconsciously when they are looking at testing a hypothesis by looking back. With the holiday season approaching I thought it would be interesting and amusing to look at the whole problem as if we were actually trying to get the figures wrong. I tried originally to do it the other way round, creating examples that were wrong and going to look at why. I found that the examples were too convincing. People who I discussed them with started to believe the false results and arguing them round was difficult. This way you can see from the start that it is wrong and we can save a discussion of how to avoid it for the New Year.

When I was at college I was advised that before throwing a Frisbee to perform a trick the best thing to shout was "watch this". If you managed the trick you would get the applause. If you failed the trick spectacularly you would get applause. If something went strange people would appreciate it. You were safe because you had not claimed a particular target. Whatever happened you could claim success. This is probably the most common way of making yourself look good as a black hat CMO.

Imagine you manage to create a 1% uplift where you were hoping for a 5% lift. How do you make this look good post factum? Well first you examine the 1% and determine how it is made up. let us say it falls in to 10 segments (if it doesn't segment it differently until it does). Now look at the uplift amongst each of these segments. One of them will have done better than the others. check the uplift for this one and claim they were the people you were trying to influence. If they are only a small proportion of the people reached then you have clearly found the right way of influencing them, but the wrong channel to reach them. You have learned something that you can now propagate on all channels to find the right channel.

If instead they are a high proportion then you have used a good method for reaching them and your message was better for them than for the average recipient. As a first attempt you have done well, you have a first stab message that isn't bad but that you can improve and you have clearly shown insight into targeting in the first place.

This is not just spin, it is spin with numbers! Numbers cannot lie (yeah right) so it must be true.

Now, what is the next way of making the results look good? Well you can do the same sort of thing to your competitors. If you have outperformed the industry then great, shout about it. If on the other hand you have done worse than average find the people who have done worse than you and say that they are the competitors who are relevant. Choose particular similarities to show why they are relevant. Once you have done that you can compare your results to theirs and you are starting to shine. Define your company in terms that restrict comparisons to the set of underperformers. Phrases like “the best in the North East (SW, wherever). The best Offline-only agency, direct mail agency, again whatever will restrict comparisons. Once you have a small enough set for comparison you can say poor performance is the economic climate or some other environmental factor and ay that it has affected everyone. When people compare (using your supplied figures) they will see you have done best despite the problems.

Next examine what figures you are comparing. Say you have that 1% uplift, what can we compare that with? Well clearly we could compare with the 5% we wanted, so you achieved 20% of optimum without even looking further. Is there something smaller than the 5% we could look at? Has anyone else produced a 4% we can compare with? This one is particularly common in the online industry.

Equally where did you get that 1%? Was it from last years figures? Well why not compare it with last months figures instead? The run up to Christmas usually produces an uplift all by itself, so why not claim credit for that? You may find that you have an 11% uplift (10 from the season and 1 from your work). Well there is a lot of news about poor Christmases so we can assume that this year it was all from our work.

So to summarise, when testing a hypothesis you need to:

*) Compare like with like (this year with last might be good).
*) Compare like with like (you vs. all competitors not you vs. a selection).
*) Set your targets clearly in advance, don’t just paint a bull’s-eye around where you happen to end up.

An do not let a black hat explain what he has done first as he will sound convincing and you will have more rubbish to remove before you can evaluate. Instead check first and let him explain afterwards. If he can convince you in retrospect he may even be a white hat (or a red hat with a white bobble).

Best wishes to you all,
Rufus Evison

Retail Media Association

Thanks in part to the economic downturn retail media are beginning to be accepted as a real force both in terms of driving trade and improving branding. This acceptance is clearly shown in terms of spending patterns amongst many of the big CPG/FMCG companies and in the use by some of the big retailers. That said, while they are starting to be accepted by those who rely on measurement and data to decide how to apportion their budgets there is still a lot of resistance amongst those who are uncertain about trying what they view as an unproven set of media.

TV may be only measures on the say so of a few thousand households out of millions, but it has been around for long enough that people know where they stand with it. Pioneers may have a great deal of proof of efficacy but it is not generally published. What is needed is a plan. The plan should involve education, PR and innovation.

This begs the question of who should own such a plan and who benefits? The answer to who benefits should determine who will own it. Here are a few thoughts on possible benefits:
1) Retailers will benefit
Retailers are the owners of the media and so improvements in acceptance will lead to more sales for them, which can only be a benefit in the current climate.

2) The supplier base will benefit
As retail media keep the customer better informed at point of sale they inevitably lead to more sales, which will benefit the supplier base.

3) The customer will benefit
The retailer will have an interest in seeing that only advertising useful to their customers is displayed, thus benefiting the customers.

4) The innovators will benefit as their innovations will have more attention and so they will be able to make money out of moving the whole area of retail media forward.

So where are we with respect to benefits? Well it seems that everyone benefits, so no one is a clear candidate to own the necessary work. This suggests the need for an industry body funded by all players in the industry and open to everyone for membership. So far at least two of the major players in the area have expressed an interest in seeing that such a body come in to existence, so watch this space for an official announcement in the near future...

Oh, and if you have any interest in becoming involved in the nascent Retail Media Association send an email to blog at retailmedia.org and we will see it reaches the relevant ears.

P.S. The first two points of benefit are fairly standard for any effective advertising, but the third is different. Because there is benefit to the retailers in long term customer satisfaction this promotes and incentive for the media owners to act as a watch dog for the consumer. A retailer can become a trusted source for advertising and so can curb the excesses in a standard "advertising war" model. A retailer is not interested in their customers spending more over all as that will just lower their revenues down the line as the customer gets hit with interest charges and spend less with the retailer and more on paying for the money they have already spent. No, the retailer's interest is in getting the customer to spend more in their store. This means that rather than shouting ever louder in an escalating battle for space, the suppliers are forced to provide relevant offers and to help the customer as much as they can. This is a much more sustainable model and may well benefit the supplier long term, but...

But even if it doesn't, it will disadvantage the supplier not to get involved if it means that their competitors are the only ones talking to their customers.