Maybe this is ill-timed, not being long after my ‘weight loss’ blog post, but here’s another story from the land of “Is that large?” In my defence, I was on a long journey and was so hungry when I stopped at some services in the middle of nowhere that I’d have eaten my own toenails. Anyway, feel free to dispense advice on health and nutrition if you want, but take comfort from the fact that when I’m at home I usually cook from scratch; neither do I make a habit of hanging about in fast food joints.
Being thoroughly British, one of my pet hates is queue-jumping. There are many variants of this despicable pastime, but one I find particularly annoying is when people send their kids to ‘reserve’ tables in busy food places whilst they queue. This was endemic at the burger joint I was at today and it drives me potty. It was busy and the queue was massive, but looking around, there were actually more people at tables without food than there were people actually eating.
This meant that suckers like me who wait our turn get to the front of the queue can’t find anywhere to sit and eat, because some bright spark who has come in after me has despatched their offspring to ‘hold’ any table that suddenly becomes vacant. Even adults were doing it, which I found really irritating.
Anyway, as ever, my uncontrollable rants usually come with a systems-based moral to the story, so here it is:
When people behave in a selfish, rational, manner such as this, they only ever score individual victories at the expense of the overall system (and the others within it). Okay, so it was busy, but my guess is that the turnover of people eating at tables is pretty constant, so if everyone just took a table after purchasing food, tables would become available at roughly the rate they are required. (I appreciate this is contingent on queuing theory, but there’s no way I was going to do the maths). What happens when people grab tables well before they purchase their food is that this natural flow is interrupted and blockages occur.
The purpose of this system (i.e. to allow people to eat their food at a table if they wish) is therefore more difficult to attain because of the behaviour of those people who elevate their own individualist aims over the collective good. It may at first seem like a good idea to guarantee your party will have somewhere to sit in 10 minutes time after you’ve finished queuing, but the fact is that this behaviour has a negative effect on the overall system. If everyone does it, the whole thing grinds to a halt. (I have a feeling there are examples in economic theory that demonstrate how rational choice behaviour damages the overall system – price wars perhaps? I’d be grateful for the reference).
Such unhealthy competition is commonplace in public and private sector organisations alike. Sales teams are pitted against each other, departments are compared and ranked, schools and hospitals are placed in league tables, and so on. Of course, the classics in policing include such things as performance documents that count numbers of arrests, or individuals being held to account for the crime rate on their beat / sector / division /force (delete as applicable) as compared to somewhere else. The difference though, is that unlike the bunch of hungry, yet selfish, travelling strangers who abused the fast food system, organisations actually build the triggers that cause this type of behaviour into the system design, as if in some kind of bizarre foot-shooting exercise.
Traditional modes of management in the Western World are based on assumptions such as ‘competition is a good thing’, ‘comparing people’s performance and ranking them is a good way of presenting data’, and ‘if we didn’t do this, people would stop working’. I argue that this is not the case; indeed that it makes performance worse, not to mention relationships. Where one work unit focuses on out-performing others regardless of the consequences to the system, this is called sub-optimization, i.e. one part of the system is optimized at the expense of other parts of the system.
The effect of this is some parts of the system ‘win’ at the expense of others that ‘lose’. Furthermore, the overall system always loses, as individual or departmental selfishness results in capacity and ingenuity becoming focused towards introspective survival activity, instead of achieving purpose from the customer’s or service user’s perspective. In addition to this, cooperation between individuals, departments and organisations is always adversely affected when people become caught up in internalised competition. Targets, ranking, singling out the ‘bottom 10% of performers’ or ‘naming and shaming’ this week’s victim at the bottom of whatever league table, causes dysfunctional behaviour and makes it incredibly difficult for the system to attain what it is there for.
The type of behaviour associated with sub-optimization is so destructive because it ignores the interrelationships and interdependencies that are so crucial to the fabric of an effective system. Flattening a bump in the carpet here might ‘solve’ that problem, but it is likely to cause another bump in the carpet elsewhere. Snatching a table in advance so you can scoff your burger sitting down might solve your problem, but it will certainly put someone else out. “But, hey, that’s not my problem right?” Well it should be, if you care about the system. If the bunch of random strangers in an isolated burger joint acted in the best interests of all those present, then everyone would get a table.
Sub-optimization and unhealthy competition in the burger setting had the relatively innocuous effect of irritating me (and giving me the idea for this blog post) – in organisations, the consequences are much more severe. What’s more, the staff in the burger joint probably could not have influenced their customers’ behaviour much (I suppose being ordered off vacant tables in the interests of fairness would not go down that well with some people) – this is totally different to the examples of sub-optimization that are endemic in organisations. Managers have it within their gift to design and adjust systems so as to remove the perverse incentives that cause this very type of self-destructive behaviour.
There are studies in organizational psychology and organizational identity where the researchers found that, more or less, competing teams within the same organization (i.e. the same global goal) would go as far as to reduce their own production and to spend resources hindering the production of the other teams as opposed to increasing their own production, therefore reducing the global output massively. So your observation on fries is totally correct. Now where are my fries?
Not sure this is ever beatable, have a look at ‘The Prisoner’s Dilemma’
Not buying this ‘don’t hang about in fast food joints’, first a blog about burgers, and then a tweet from a kebab house! I might be daft, but I’m not stupid. 😉
In the context of how customers behave in a fast food restaurant, I think you’re right. Prisoner’s dilemma is a good example – you know your stuff, my man (or woman)!
Are you on Twitter then?
Actually, I think you fell into your own trap and overlooked some important points.
Where I come from, burger joints usually have 3 kinds of customers: people coming in to eat their food at the place, people taking their food away, and people at the drive through.
Now, if everybody (and their noisy kids) would wait at the counter, the waiting area would soon clog and take-away people could not get to the counter to order. Therefore, to send the noisy kids ahead to the table and to have only one person standing in line actually speeds up the ordering process, as then only one operator at the counter is occupied for each group, freeing the other operators to cope with other customers, e.g. at the drive through.
To put it in terms of queueing theory:
Every customer “job” is composed of wait time (t_w), serve time (t_s), and eat time (t_e), where I include the waiting at a table into the eat time.
Now, following your model, where everybody has to get his own food, t_w and t_s are the same for all – everybody has to wait the same time in the queue and at the tables.
In the group ordering model, I would expect the mean value of t_w to be shorter (groups are handeled as 1 job, therefore less jobs in the queue, and groups block only 1 server, therefore higher throughput for smaller jobs, i.e. single persons). The mean value of the serving time t_s would increase (same amount of things ordered, but distributed amoung less jobs – but the higher serving time affects only the group orderings, t_s for single persons is the same as in your model).
Now let’s consider t_e, what you ranted about. In your model, t_e is ok for single persons (your argument), but long for groups – as a group has to wait for a whole table to clear after getting their food. In the group order model, t_e is less for groups (achieved through clever work distribution), and a little bit longer for single persons (it is easier to get a single place at another table than to find space for many). So actually, the overall mean value of t_e is LESS in the group order model.
To conclude: for single persons, both process models amount roughly to the same time – in one model you wait longer in the queue, in the other longer at the table. For groups, the group order model yields better performance, since ordering amoung the group is bundeled and wait time at the table is reduced. So overall, the group ordering model has the better overall throughput, enhances order in the waiting area and reduces stress for kids and staff.
I would therefore prefer the group ordering model – and rant much more about single persons hogging whole tables to themselves.
Btw: in computer science, when considering memory usage, there is also evidence saying that placing large blocks first yields better usage of memory. When you transfer this to the table management in the burger joint, it also favors the group ordering model – placing large groups first in the seating area amounts to better usage of the tables. Another advantage that you overlooked…
Thanks for your insight and the thought you have put into your response. It’s not always easy to get the balance right between covering all bases one has considered when writing short blogs, against overloading readers with too much information. To respond to your points, firstly let me agree with you on the typology of demand these burger places face (i.e. the three types of customer).
However, from what I could tell in my example, drive through customers were dealt with on a separate till, meaning that the queue inside the establishment was unaffected. (I appreciate this was not clear in the post). Next, I think its important to note that there is no real distinction between groups and individuals who order food (save for the time it takes to ask for their order) – what usually happens is that one person orders the food per group, regardless of whether the other group members are standing with them at the till or already waiting at a table. Therefore the queuing and ordering time is unaffected for anyone else in the queue, whether they intend to eat in or take food away.
I agree that worthy cause of a rant are those individuals who block entire tables when there are opportunities to sit elsewhere, but the point of my example was to show how table-blocking disrupts flow and restricts available capacity (i.e. tables) – I don’t agree that sending kids to the table speeds the process up as claimed. Perhaps someone with a bit of spare time could time the process under experimental conditions and see who’s right.
All the best.
Yes, that thought about counting customers at my local burger joint occured to me – at my university, we once had students counting in the cafeteria to enhance the serving process there 😉
My guess would be that the real factor is the size of the seating area – if this is planned too small, blocking happens regardless of the queuing process.
This also conforms to queuing theory – job priority (= claiming tables for groups) and job size have less impact than queue capacity and operator throughput. And since you cannot force people to eat faster (would be very unhealthy), the number of tables would have the highest influence on the overall process performance.
Might be fun to do a real model for that 😀
(As you can guess, my background is in computer science – and queuing theory was a favourite subject of mine.)
Moving from computer science to psychology, I think it is unrealistic to hope for all people to see the big picture. For me, an ideal system is designed so that local optimisation (i.e. everybody for himself) automatically amounts to global optimisation. It is a question of setting the right incentives. Hard to achieve, though.
Keep up the writing – even though I disagree in this case, I like reading your thoughts!
Hi again. It’s a shame the university experiment never happened. Maybe there’s a PhD opportunity for someone one day…? Apart from that, I think you make a couple of important points here – firstly, you’re right about the size of the seating area being a major influencing factor. As I tell anyone who’ll listen, “A process is only as good as its narrowest point”. If the amount of seating has not been determined by analysing predictable demand then it will result in a system that exhibits insufficient capacity, resulting in the blockages and bottlenecks you describe. It’s true that this will occur even if people don’t block tables. Likewise, I agree that in this setting it would be impossible to restrict the individualist behaviour of paying customers (who are free agents after all) – I argue however, that in an organisational setting, there is scope for management to design the system so as to prevent internal competition which damages the overall system.
Just coming back briefly to how I envisage the act of blocking a table slows the whole burger system down (and maybe I should have included this in the post, although I was conscious of word count), think of it like this…
If at the starting point, all tables are occupied and the average wait time (t_w) is 9 minutes, the average serving time (t_s) is 3 minutes, and the average eat time (t_e) is 15 minutes, then if (and it’s a big ‘if’, I appreciate) the restaurant has designed its system against demand, the rate that tables become available such be constant enough to accommodate each new customer (or group) clearing t_s. If, at the point I walked in, the level of demand was toward the upper limit of capacity (i.e. it was extremely busy), then the system is at its most fragile, but should still be within the boundaries of being able to cope – there would be enough tables to accommodate a throughput of customers at a rate of t_w + t_s + t_e (27 minutes) per customer/group.
Therefore, whilst the end-to-end time per customer/group involves 15 minutes at the table during the t_e phase, then assuming that the t_w stage is constant (that’s a whole other problem – perhaps insufficient till staff etc) then as long as people allow natural turnover to occur with only customers at the t_e stage occupying tables, the system is able to cope. The problems occur when someone enters the restaurant (Customer A) and immediately occupies a vacant table that has just become free, which by rights is likely to be required by another customer or group about to complete the t_s stage (Customer B). This results in a situation where at least one member of the Customer A group still has to undergo the t_w and t_s stages (12 minutes), before using their ‘reserved’ table for what it was intended (i.e. scoffing the food at the t_e stage for another 15 minutes). Meanwhile Customer B has to wait a few more minutes for another table to become available.
The result of this behaviour is that the table is occupied for a total of 27 minutes, instead of the 15 minutes actually required for t_e activity. This represents the introduction of idle time (waste) into this part of the system of 12 minutes – almost a whole t_e opportunity for someone who requires it. The result is that as more and more people do it, the cumulative effect is that a ‘queue’ of customers waiting to enter the t_e stage begins to form, which, if fresh customers grab tables as they enter the restaurant, will grow exponentially, as per queuing theory. If the system is almost at capacity, then whereas it could cope with tables being occupied in line with t_e ( i.e. every 15 minutes), it will collapse if people block tables from the outset, as their behaviour artificially limits the availability of tables at the point they become necessary to permit flow to continue at the rate of natural turnover.
Nothing about table grabbing actually speeds up t_w or t_s for the customers who do it, either. It’s a false economy that adversely affects the overall system and the other people within it.
Goodness, that was almost a blog post in itself. Hope you managed to stay awake throughout it. Anyway, it’s good to discuss this stuff. Watch this space for more posts very soon…
Well, no PhD opportunity, I think – there is no new theoretical insight to be gained, as it is all application of existing methods (twice the shame, if you ask me – to have all that methodology available and NOT use it is something I am really challenged to accept). It’s a bachelor’s project, or, as it was at my University, an exercise accompanying a lecture.
(I was tempted to to something similar at the University where I did my PhD, but the topic unfortunately did not fit. The redesign of the cafeteria there, called “free flow”, but amounting to blocking queues in the serving area, is one of my favorite pet rants.)
As to your analysis of your burger joint, I agree. In CS, the described phenomenon is called “job starving”. In computers, this usually happens when lots of high priority jobs displace long, low priority jobs, so that the low priority jobs actually never execute. There are, however, scenarios where this is intended behavior, e.g. when losing the high priority jobs is critical for safety issues. Also, prioritizing systems usually have a better overall performance, even if the cost is the starving of some jobs. It amounts to a design decision of performance vs. fairness vs. fast reaction in critical cases. As a restaurant customer, I would wish for fairness – however, from the viewpoint of the owner, optimizing for performance might yield more revenue.
I think your point about fairness sums it up. Whereas a system should always be designed to achieve purpose from the customer’s point of view, there exists the risk that, as in this case, ‘true’ purpose can be disregarded in favour of the system owner’s aims, i.e. to maximise profit, even if this is at the expense of disgruntled customers. Ultimately, this may we achieve short term gains for the company, but at what unseen long term cost? I for one, won’t be stopping by again…
Cheers for your comments, see you next time!
Long ago I noticed how annoying it is if other customers hog the table, and a long time ago I worked out my response to it – I would sit at the empty seats, and when the “hogger” said “This is for my partner in the queue” or similar I would respond by saying that by the time they got here there would be another seat for them somewhere. I would beat them at being anti-social!
But here is the interesting thing – I have only had to do this once ever. And I formulated the plan when we had two young boys, which expanded to four, so we were often after the biggest tables (6 people). Now there are just two of us (sigh) and still there is always a table when we need one after queueing nicely.
So my quiet observation over time (20+ years) tells me it doesn’t really matter very much. And so the killer question; after queuing nicely did you have to wait for a table, or was there one when you needed it?
What a great idea… I must try that! In answer to your question, I had to wait… pah!