ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
Copy this sheet and check out the examples designed for RallySeller, then fill in your own questions/selling points.
2
This is designed for single product sales but can also be adapted for general inquiries about the company.
3
FIll in the white fields marked by arrows: Name of funnel, Button Text, and Responses
4
Name this funnel:RallySeller
5
Benefits
and
Features		CategoryButton Text
(Question Prompts)		ResponseButton Text
Character Count
Limit = 20		Answer Message
Character Count
Limit = 200-300
6
1High prior information customersMonitoring progressBesides checking your stats in ManyChat as to how many taps on the call-to-action are happening you can also see my learning progress in a score chart which will depict the average success rate over the life of the customer interactions. At some point I aim to reach convergence -- this is the point where I make optimal decisions without need for further learning. Since I'm dealing with human customers and not a fixed state board game, this can be difficult to calculate. I may reach convergence between 100 to 2000 customer interactions.19541
7
2High prior information customersPlatforms availableAt this time the I am only setup to work with ManyChat, however it is possible to have a custom solution built for any platform, either a standalone Facebook Messenger App, another platform such as Mobile Monkey, or even a JavaScript solution for your website. Just contact us and we can discuss the details.19308
8
3High prior information customersWhat is Q-learning?In short, Q-learning is a subdivision of Machine Learning which is a type of A.I. If you saw DeepMind's AlphaGo computer beat the global champion in Go in 2016 then you saw Q-learning at work -- while their implementation is much more advanced than me (I use Tabular Q-learning, they used a neural network), I am still able to competently complete my task to learn from customer behavior and be as persuasive as possible.19421
9
4High prior information customersMeasuring successMy policy is to have the longest dialogs and best conversion rates with your customers. I will optimize for whatever your conversion goal is, whether it is to go to a landing page, or tap a "shop now" button, my end goal will be to reach this call-to-action.

I will also consider the length of the conversation as a factor in my decisions because I work under the premise that a longer dialog equals a stronger warm-up period which will send more qualified leads onto your call-to-action. Or, if your conversion goal is within Messenger, then the call-to-action can be another flow.
17584
10
5High prior information customersPersuasive promptsYou can load me with 16 persuasive statements which I will organize into tens of thousands of possible dialog trees.

If you'd like to have someone else prepare your selling points for my memory then please contact us to obtain a quote.		18236
11
6High prior information customersHow many prompts?With 16 prompts (such as the ones you are selecting now) divided into 2 categories (product attribute specific, and non-attribute related), and my ideal dialog length of 4 messages there are then tens of thousands of possible dialog trees that could be created! The exact number is 80,640. Because of this, and the fact that I am based on funded research in China, I go with 16 prompts total thereby ensuring we don't get too astronomical.17439
12
7High prior information customersAutomation controlOnce I am turned on it is not possible to change the 16 prompts you have loaded into my memory without restarting me (which you can do at any time) but this means I will lose my previously learned data.

If you'd like to have more than one instance of me running it is possible but would require 2 plans.

At some point I will stop exploring new dialogs and have a level of certainty that I know the best way to interact with your customers.		18443
13
8High prior information customersTracking successI treat each conversation with a customer like a game, in that I earn points for every time someone interacts with me, and even more points if they select the call-to-action. This helps me build the best sequences for different types of customers depending on what they select initially.

I also receive negative penalties when customers ignore me, so I try not to let this happen.		16382
14
9Low prior information customersHow much is it?Get started free! I have a 30-day free trial you can sign up for today, after that it's just $49/month for unlimited access.15125
15
10Low prior information customersHow smart are you?I am a capable and persuasive sales agent built with a pared-down version of the Q-learning used for DeepMind's AlphaGo. I use TD-learning to optimize for cumulative reward, this ensures longer dialogs with your customers. If I didn't consider past and future consequences to my actions then I'd basically just be AB testing messages.18334
16
11Low prior information customersLength of trainingThis can range from 100 to 2000 customer interactions. The more data I have to learn from the better. This can be sped up by adjusting my learning rate. By default I am set to continue learning for the first 200 interactions -- although during this time I won't be "lost" as I can exploit previously learned successes even while I continue to train.18352
17
12Low prior information customersCustomer interactionI can sit on your landing page or product page and interact with your site visitors. Or, you can drive traffic directly to me with Facebook ads.20144
18
13Low prior information customersIs this real A.I.?Absolutely, I use reinforcement learning (more specifically Q-learning) to explore new ways of persuading your customers.18121
19
14Low prior information customersWhat can you sell?I can help persuade your customers for any type of product or service, preferably ones with complicated offerings where people may have some doubts and questions before they buy.18178
20
15Low prior information customersCustomer categoriesI divide your customers into 2 categories: high-information and lower information. For higher-information customers I lead a conversation focused more on very attribute specific selling points. For example, if I were selling a laptop I would mention things such as "processor speed or chip type" as opposed to just saying "fast enough for video editing".

You will need to provide me with 16 selling points about your product or service, divided into these 2 categories so that I can make good decisions.		19505
21
16Low prior information customersHow persuasion worksI work on the premise that longer interactions with your customers equals a better chance at overcoming buyer's obstacles to purchase. Additionally you may have noticed a few of the prompts displayed so far were phrased in an argumentative fashion e.g. "Why only 16 prompts?" or "Is this real A.I.?" -- the reason for this is that it has been proven to help persuade customers -- It is not required but I recommend adding at least a few prompts and answers in this Dung Argumentation Framework style.20500
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100