Product design

Rapid prototyping and testing – Logic model builder


Sametrica had a first version logic model builder that was not meeting expectations, and was built in older technology.  The team had been getting feedback from customers and brainstorming on their own about what would make a great logic model builder.  We did a lot of whiteboard work, synthesis of ideas and epic planning.  Design was underway, but the pace quickened when we won a round of innovation funding from the Government of Canada to build out this app and the AI we expected it would utilize. 


Since we now had immediate technical deliverable dates, testing our ideas had to be expedient and as “real” as possible.


  • Logic model practitioners
  • Program managers
  • Program staff

Getting paper to wireframe

Economists and mathematicians are really fascinating stakeholders, but they think very differently from product people.  I find the best way to get started is to take ideas out of heads and put them on paper.  Once concepts are translated from sketch to wireframe, then the testing of assumptions can begin in earnest.  We can’t design together until diverse people on the team have processed ideas and we are on the same page about the definition of what we are designing.

Getting stakeholder ideas on paper into a wireframe
Wireframing stakeholder ideas for initial hallway testing

Prototype 1 – CodePen

The construction of a logic model can vary tremendously, there is no standard formula.  However, they usually have a similar set of components.  We needed to find out if different practitioners could use our system to construct a logic model?  At the same time, we needed to test our hunches about where we should insert AI.  

Logic model practitioners are few and far between with huge geographic dispersion.  Since we could not afford in person visits, an online prototype seemed like the best way to garner feedback.  Using CodePen we got a prototype up and running.

Initial CodePen prototype

Talking with people about this prototype underscored some major issues around how complex our product needed to be.  We quickly added a CodePen of our calculations modal, and listened as testers described how they construct calculations. 

Initial CodePen of Calculations modal

Prototype 2 – CodePen

The second prototype incorporated the feedback we got from the first one.  Plus we added more throughput so the tester could actually add realistic data.  As people structured real logic models in the prototype, it was getting very cluttered.  The need for AI in multiple seen and unforeseen touchpoints was becoming very evident.

Second CodePen prototype modified based on user feedback
Cluttered UI driving home the need for an AI recommender system with a simpler interface

Experience Map – Narrowing the focus

Time was getting tight on our Government of Canada code base deliverable.  To help us narrow the focus and get a usable MVP from our prototype research we had testers map out their experience.  We broke the system down into essential touchpoints.  For each touchpoint the testers noted their emotions, peak/valley, thoughts and ideas.

An experience map helped us narrow our prototype research and define an MVP

Invision – Working out the kinks in the UI

The experience map showed us specific spots in the UI where testers were getting stuck, or in some cases had latent needs we had not foreseen, but were essential for MVP.  We used Sketch and Invision to make a very realistic prototype of UI fixes gleaned from the experience mapping feedback.  We needed to make sure that our UI was going to be intuitive to use before we committed it to code.  This exercise allayed our concerns and we were ready to write the MVP stories.

Invision prototype to test proposed fixes in the UI
Stories based on feedback

Results from prototype testing

Configurable stakeholders
Helper prompts
Elegant un cluttered AI based UI
Contextual learning for users who don’t often construct logic models
The need for this admin table was discovered during the testing process