Management Communication

Setting the Stage, Choosing Your "Weapons", Shaping Your Message, Applying This as a Data Professional,

If you work in a Data role, you will communicate with the C-Suite on a bi-weekly basis. If you are a Quant, you’ll communicate with Portfolio Managers on the regular. Highly recommend reading this, so you know how to actually convey your message, and sell your model to them.

If no one buys your model, or your message, you don’t get a nice bonus.

By BowTiedBear42. You can find his twitter here:

In my experience most data scientists communicate in great detail and precision - and are often completely misunderstood by everybody but their direct peers.

Sounds familiar? Come along for a ride if you want to improve.

Table of Contents:

1. Introduction

2. Setting The Stage

3. Choosing Your Weapons

4. Shaping Your Message

5. Applying This as a Data Professional

6. Wrap Up

1 - Introduction

There are three key takeaways that I want you to keep in mind:

1. Communication styles across every organization can be very different.  You have to tailor messages for your target audience if you want to highest chance of success

2. Find your supporters before a key meeting. The decision may be formally made in the meeting, but the right supporters make victory so much easier

3. Practice & rehearse. The Rolling Stones have been playing live shows since 1962 and they still rehearse for the greatest show possible every single time. For good reason. Good communication is a performance and practice is the only way to become really great at it.

(some grandfather, not to bad for guys pushing 80. )

You may have some disdain for the “strategy consultants”. They create nothing but PowerPoint slides for 12+ hours a day. But, understand that they are there to craft persuasive messages for high-stake situations. A lot of what they do is testing & refining their presentations over and over again. I would recommend that you do the same, when you prepare for important meetings.

I assume that most readers here are in university and entry level tech positions. You have to realize that the communication style you use with your peers daily is unusual. Most people will either feel dumb or label you as nerds if they encounter it. Both are negative reactions that you want to avoid at all costs.

Emotions should have nothing to do with it, but humans are emotional creatures. You may or may not like this fact, but you have to work with it if you want to communicate with a broader audience.

If you call this dumbing down, you use the wrong frame to start with. You should probably follow Mr. Frame

BowTiedSalesGuy

anyways.

2 - Setting The Stage 

This discussion is most relevant for presentations in high-stakes situations. This could mean:

1. You asking for budget or a project approval far up

2. A sales pitch

3. A public presentation with important stakeholders around

4. The aftermath of a major fuckup

There is no one-size-fits all communication for those situations. 

I would recommend that you start with two questions:

1. What do I want to achieve?

2. Who do I need to persuade?

On a most basic level that means identifying the decision maker. It is usually worth your time to analyze the wider stakeholder landscape. One approach I really like is the power map. You basically arrange relevant stakeholders along two axis: power level and interest in the topic:

Powermap Sketch

I´ll repeat it here because it is so important:

Important decisions are very rarely made in the actual meeting.

You need to make the right friends beforehand and prepare them for the decision.

This may or may not work. In some cases you can have individual meetings with key decision makers before. In other cases, you may have to use indirect channels. For busy executives, your best shot is usually somebody like an assistant/advisor.

You should especially pay attention to relevant influencers.

In a very classic example secretaries are not the formal decision makers. But, mess up with them and you are done without knowing what hit you. You'd be surprised how much can be gained in many companies by treating people with common courtesy.

3 - Choosing Your Weapons

Once you have determined what you want to achieve and who needs to be persuaded. It is time to pick your weapons - or to choose how you want to communicate in this case.

Weapons may still be useful, but let us agree to try words first.

The very first question you must ask is:
What are their incentives?
This is usually understood from corporate strategy and some questions around the organization. Even better if you know what their bonus is based on.

What do they care about?
People usually also have personal goals too, those are a bit harder to understand. Does the person like to be on a stage? Is the person a control freak? Try to give them some of that to make sure they enjoy the interaction.

Personality type analysis can help here. This will also give you some hints for your communication.

If you are not familiar with personality type classifications, I recommend that you learn this. Here are two examples:

• Myers Briggs Personality Testing

• The Disc Model

All of them are of course a gross oversimplification, but they are useful. I especially find the “what they want from others” section helpful if you want to target communication:

What They Want From Others

Flip this around and you get a “what drives them nuts” overview. Should be a decent starting point If you have nothing else. Generally I would use the following sources of information in this very order:

1. Briefing by people who know the target person well. 

2. Analysis of communication by the target person. How do they communicate? I would include some web search, have they given any interviews?

3. Analysis of past successful presentations to the target person. How did presentations that the person liked look like?

4. Personality type - you usually have to take an informed guess here.

4 - Shaping Your Message

By now you have a good idea of what you want to achieve and who you need to target. You also have a rough idea what kind of message would work.

Now is the time to leave behind writing as you know it from university. Get used to starting with the key conclusion if you want to make sure most people understand you.

Let me introduce you to the Minto pyramid.

This style of communication is taught in the consulting world and I became a big fan of it. You should always put the key point out first and dive as you need to with the respective counterpart.

Note: Quants also use the Minto Pyramid when creating reports, or presenting.

You should always prepare all levels for important meetings. A way I would approach this in a presentation is :

• I would put levels 1-2 in polished slides

• I would put the detailed arguments in the backup as far as possible

• I'd be prepared in case someone asks for this huge level of detail.

5 - Applying This as a Data Professional

Let us assume you work for a Christmas cookie manufacturer. Let’s say you have developed a great model to predict which cookies certain customers are most likely to buy. 

Tasty Cookies

You already know from some test runs that you can increase revenue by targeting customers with their favorite cookies. It would cost the firm some investment but you ran the numbers and know it would pay off handsomely.

How would you sell this internally?

Let us assume that the firm primarily focuses on revenue. I will spare you the slides and focus on the arguments and presentation here:

Level 1: Direct Team & Team Leads

Conclusion:

I developed a great model to predict cookie preferences. Already tested out of sample with a precision of 90%. Let us bring this up with the department head.

Key arguments:

• Probably a detailed discussion of the model design and KPIs

• A projection of the model's predictive power in the context of the use case. That would usually mean to create a sales scenario with realistic figures. Should give you a clear idea that using this model will increase revenue. 

Detailed information:

• Model code review

• Detailed review of model testing approach and various KPIs

• Discussion of alternative models

• Detailed discussion of the underlying assumptions 

• Detailed discussion of the 

( example for illustration and not related )

More details on ANOVA below:

Anova Regression

Level 2: Department Head

Conclusion:
We could generate X in more revenue this year with the new machine learning model we developed. Customers buy more if they see the cookies they like and our new model can predict this quite well. Implementation would cost us Y, which is less than 10% of the expected revenue.

Bonus point:
High share of budget Y going to the department head you are talking to.

Key arguments:

• High level revenue projection with and without the model to show impact

• High level explanation how the model works

• Your team lead saying: “We checked the model in every possible way, kid did a good job”

Detailed information:

• Detailed revenue model

• Likely a discussion into the testing and your confidence in the model

• Likely a detailed discussion on what is needed to make the whole thing work

Level 3: Executive 

Conclusion:
We can gain market share and generate X in additional revenue with investment of Y.

Key arguments:

• We developed a well working ML model that predicts customer preferences and can be used for improved targeting

• Revenue scenarios 

Detailed arguments:
Maybe anything from the department head level.

Summary:

Note how we built a little pyramid for each level and shifted focus of the argument from model quality to commercial performance. This is of course done because we expect that this is the thing they care about.

6 - Wrap Up

By now you should certainly understand why this post is structured this way.

As I mentioned initially: The big key here is really that you practice your communication skills. This is one of those skills that you should develop further for the rest of your life.

I want to leave you with a few good videos to take it further: