Getting Agreement on âTHE PATH TO DEPLOYâ yourâ¦ predictive model, big data application, machine learning based algorithm, segmentation, insights, optimization modelâ¦
Battle Scars along the journey to the truthâ¦.
When you took your first class in Data Science, R or Linear Regression either at your university or at Coursera did you ever imagine you would struggle with getting your recommendations or analysis implemented?Â Did you know you were also joining the diplomatic Corp?
As a good Data Scientist you go through the process of:
- Defining the problem and objectives for your model or insights project.
- You gather your information/data for your analysis
- And you do all of the steps that go into any best practices analytics methods approach. Data Cleansing, Reduction, Validation, Scoring and more.Â
When you go to present your analysis or data science results to your clients you end up getting push back in terms of what we call âthe path to deploy.â
Frustrating though it may be, hitting this first wall in your data science career, should be viewed, as a learning experience and your greatest teacher, not only in your current role but also for how to assess the organizations analytical maturity (including top management support) when searching for your next role.Â Our greatest problems are our greatest teachers.
I have a few painful memories in this area as well, but since then the diplomatic corps has been after me to enroll, one was when I was working with my data science team on behalf of the CMO in a card business to build out a Credit Card Customer Card Acquisition model. The purpose of the predictive model was to make cross selling credit cards to a retail customer easier, more efficient and more profitable.Â All Noble and correct goals by all appearances.Â Â The CMO wanted to ensure we acquired customers that not only responded to our offer but use and activate their card.Â This strategy made sense not only from a profitability standpoint but also from a customer centric point of view.Â The CMO who had agreed with the Decision Science team on the modeling approach and the âpath to deployâ was stymied by his boss the CEO who didnât want to implement the model as the CEO was under pressure to grow the number of card customers and was afraid of not hitting his account/widget goal.Â The CMO was very frustrated as was the entire team.Â So after much debate we scaled back our full âpath to deployâ approach, by offering to set the model up to score and be applied to a smaller netted down number of cross sell leads, so we could prove the value of the model by showing that we may get slightly less customers but they were 5 x more profitable then random.Â I call this deferring for collaborative agreement.Â
Some Key Learnings
- So if you think that Data Science is only about science, and trust me we wish it was, then think again. Decision makers are biased by their educations, judgements and what they know about a subject and the more you can show up as helpful and supportive in terms of educating the executives about the statistics and how the modeling works the more supportive they will be.Â Think back to when you were hired and someone said that you had âgood energyâ well what does that mean?Â It was a value judgement about how well they liked you, your physical presence during the interview and your ability to articulate your value story.
- View an initial âNoâ as a deferral and regroup with the team to come up with an alternative test and learn approach to deploying your analysis or model. Sometimes you need to crawl before you can walk and then run.Â
- Ensure that you ask this question up front of the client: Are you the ultimate decision maker?
- When you present your analysis results consider presenting a variety of options for executives with more detail around how each option impacts the business. Use data visualization to make your case.Â
- Try and anticipate both positive and negative analysis outcomes upfront and list them as potential challenges in the initial project definition: this way decision makers can decide if they would even go forward with the project.Â
- Use influence maps to determine who can help with pre-socializing your analysis results to convince the decision maker that the outcome is worth testing.
- Try and understand who are the data science and analytics champions within the company and request that they be your mentors or better yet invite them to key meetings if possible. Â Â Â Â
In Conclusion:Â Â Â Become known as the Data Scientist who can drive collaborative decisions through both horizontal and vertical thinking through:
- Listening to your clients and or corporate culture is key.
- Getting agreement on the âpath to deploy.â
- Using Influence management techniques to persuade the client or executives to test the analysis in some way. Donât consider it rejection just consider it deferring.
- Demonstrate how the analysis directly aligns to the mission in the organization. For example, many firms are talking about customer centricity and your very analysis is key.Â You need to remind people that test and learn and innovation and risk taking is the key to driving toward this vision. Â Â
- Know and cultivate executive sponsors, evangelists and champions of analytics. Â The best firms will have an executive sponsor.
We would love your thoughts and if you find this article helpful please feel free to reach out to our advisors on call or our coaches if you need a sounding board before you react to one of these situations.