Modelling customer behavior-cohort system

Use this forum to discuss any issues relating to Systems Dynamics and Systems Thinking.
Post Reply
dale74
Junior Member
Posts: 2
Joined: Thu Jun 16, 2016 3:16 am
Vensim version: PLE

Modelling customer behavior-cohort system

Post by dale74 »

Hello Ladies and Gentlemen,

Hope this message finds you all well.

I am modelling a 6 stage cohort system to figure out customer behavior in my online business to then make sales projections.

Customers order on this business via a mobile app and they start their life in this model as a rate variable called "Downloads of app" (constant at 500/month) feeding a stock variable that I call "Registered on app". At this stage I have not received any order from people "Registered on App". Some of these "Registered on App" (my first stock variable) then go to place an order and trough another rate variable I move them to a second stock variable called "Customers 1 Order" and so on till a stock variable "Customers 6 orders" that is the last stage of the model and represents customers that have placed 6 orders. There is another rate variable that comes out of each stock variable that is a loss rate, representing customers that after the n-th order drop out of the system without placing the n-th+1 order.

I based this model on an age cohort model (population modelling), but instead of age I used number of orders.

My question is about rate variables and I will explain it using the real numbers I am working with:

I have had so far 5,600 downloads that is current level of the "Registered on the app" stock variable and of these 1,400 have placed one order on my mobile app (becoming customers); and this is the current level of the next stock variable in the cohort system "Customers 1 order". 750 have placed a second order and that is the level of the third stock variable in the cohort ("Customers 2 orders").

I assume that my success rate from downloads to first order is 1,400/5,600=25%. I also assume that my success rate for customers placing a second order is 1,400 (total customers that placed a first order)/750 (total customers that placed a second order)= 54%. The more I progress the more the success rate increases with 85% of customers placing a 5th order placing also a 6th order.
Am I correct or should I use for customers that place a second order a ratio of 750/5,600 ("Registered in the app")=13.39% (I don't think so as it does not sound logical, but not 100% sure)?
The success rate and churn rate for each cohort sums up to 100% so I am not adding or subtracting customers going from one stage to the next.

The behavior of the model after simulation shows that since my success rates are getting higher as customers progress in the system they would accumulate in the last stage of 6 orders (for which I assume then a 35% churn rate, but this is still unknown as the business is young).

Thank you for reading and for any comment. Uploaded a sketch of the system.

Best regards,
Dale
Attachments
sketch.docx
(20.05 KiB) Downloaded 361 times
LAUJJL
Senior Member
Posts: 1421
Joined: Fri May 23, 2003 10:09 am
Vensim version: DSS

Re: Modelling customer behavior-cohort system

Post by LAUJJL »

Hi

It is better if you can upload a model.

Regards.

JJ
dale74
Junior Member
Posts: 2
Joined: Thu Jun 16, 2016 3:16 am
Vensim version: PLE

Re: Modelling customer behavior-cohort system

Post by dale74 »

Model included.

Thank you
Attachments
6 buckets customers model.mdl
(9.61 KiB) Downloaded 407 times
johnjohnson4
Member
Posts: 47
Joined: Mon Mar 14, 2016 4:58 pm
Vensim version: PLE+

Re: Modelling customer behavior-cohort system

Post by johnjohnson4 »

Is your formula for the Customer 6 order correct?

Customers 6 orders=Drop out sixth order+Sixth order rate

Seem like it should be

Customers 6 orders = Sixth order rate - Drop out sixth order
LAUJJL
Senior Member
Posts: 1421
Joined: Fri May 23, 2003 10:09 am
Vensim version: DSS

Re: Modelling customer behavior-cohort system

Post by LAUJJL »

Hi

Your model can generate negative levels in extreme conditions.

To avoid that you must change all drop out and order rate, especially the order rate to take into consideration what has been already leaving with the first drop out.

Look at the joined model. I corrected only the two first equations, but you may correct the next ones.

In your first model it is easy using the synthesim and varying the drop out no order and the first order rate, to make the Downloads going negative, that has no sense, because you can draw more than exists out of the level.

Besides this the model looks complex. Do you really need to know the amount of each levels?
How will you calculate the drop out no order. Is is possible to say after one month which client will leave definitively?

Did you study other diffusion models, like the Bass one? It might be simpler to start with this one.

What is really the problem and the purpose of the model. What do you expect? It seems that you just try to model the reality, hoping that it will tell you something, the best way to failure.

Regards.

JJ
Attachments
6 buckets customers model_2.mdl
(9.89 KiB) Downloaded 360 times
Post Reply