Affinity for Ideation

Affinity Clustering and Ideation.LATERAL THINKING

  1. Evaluative OVOC
  2. Generative OVOC
    1. Affinity Clustering
    2. Using Affinity Clustering with OVOC Data
    3. The Ideation Session

1. Evaluative OVOC

For an evaluative OVOC, it can be straightforward to integrate your findings into your next prototype iteration or develop design ideas for offering improvements. Follow the following steps:

  • Print out your prototype or current solution: If your evaluation is of a prototype, print it in large format so you can hang it on the wall of your ideation room. For websites or apps, print each screen. If it’s a business process like order-to-cash use the as-is process map instead. If it’s a physical product, print a large format picture of it.
  • Print out your Roses, Thorns and Buds: Print out the output of your debrief sessions on post-its. For a small project you can hand write them, but for larger projects, purchase printable post-its and print them as outlined in the executing OVOC section.
  • Group the Roses, Thorns and Buds on the prototype: Stick each Rose, Thorn and Bud on the prototype printout next to the section each pertain to. Overlap duplicates to give yourself more space on the wall.
  • Create a separate affinity with the system-wide Roses, Thorns and Buds: As described in the last section, you likely have a set of Roses, Thorns and Buds that aren’t specific to a particular part of the offering, but pertain to system-wide or work practice observations. Use these to create a separate small affinity (se next section for details).
  • Address design changes: Gather your design team and have them read through the system-wide affinity first. Then, address each section of the prototype or process in turn, deciding on design changes that you will make. Record each design change. Since each Rose, Thorn and Bud is uniquely numbered via the template, you can optionally track exactly which customer data points motivated each design change.

 . Generative OVOC

Like OVOC interviewing itself, there are a variety of ways to analyze the data that is produced. And like OVOC, for our implementation is PMT, we are simplifying the analysis process for standardization and ease of teaching. We will adopt the following progression of tools:

  • Affinity Clustering
  • Optionally,Personas and Experience Diagramming
  • Concept Ideation

 

2.1       Affinity Clustering

 

Affinity Clustering (or Affinity Analysis) is one of the most versatile and useful data analysis tools in the HUE toolbox. Affinity is a process to meaningfully cluster observations and insights from qualitative research and draw out common themes. The affinity diagram organizes all of the key issues and visually shows the scope of the problems and opportunities.

The original idea harks back to the Japanese quality movement of the 1960’s and 1970’s – you may hear it referred to as “KJ Analysis” after Jiro Kawakita, who first developed it. Affinity features prominently in both the Six Sigma and Lean UX movements.

A sample section from a small affinity is shown here. Each of the post-its represents data points – in your case they will be individual Roses, Buds and Thorns from your OVOCs.

Note that the individual data points are grouped together and labeled by some aspect of their meaning that is important for this team: “Organization,” “Error Handling and Prevention,” “Language,” etc.  These are the threads of meaning that emerged in this analysis – as will become clearer below, this team did not start the analysis with these groups, the groups emerged from the affinity analysis process itself.

For small affinities having fewer than about 50 data points, this kind of single-level analysis works well. For larger analyses—especially analyzing OVOC data, the affinity needs extra layers of hierarchy.

An example portion of one of these multi-level affinities is shown here, taken from a UOP project on contractor use of UOP’s Schedule A engineering package. Here, each yellow post-it came from an OVOC debrief session. Note that each data point is uniquely identified with a user number and a note number within the debrief session (e.g. “U14-9” in the middle column is the 9th note captured in the debrief from User 14’s OVOC). Your notes will look the same, except each post it will be prefaced with “Rose”, “Bud” or “Thorn.” In this affinity, the individual first-level clusters are labeled by the blue post-it text. Likewise, the blue post-its are also clustered and are labeled by the second-level pink post-its. Typically, a third level is also included in these larger affinities.

You can think of the resulting affinity as a giant upside-down tree, with the third-level labels representing the trunk, the second-level pink labels representing large thick branches, the first-level blue post-its representing the small branches, and the individual data points as the leaves. The main point of the analysis is the threads and patterns that make up the first-, second- and third-level labels. These explicitly show the issues and opportunities contained in the OVOC data you collected and how they’re tied to individual groups of direct customer observations.

Note that the label post-its are written in first person. For example, as shown here, “We use the ‘live’ aspect of Smart Plant over the life of our project” is written as if it were the users speaking to us off of the affinity. Writing the cluster labels in the voice of the customer increases the impact for your stakeholders who will use the affinity for ideation later.

Using Affinity Clustering has a number of advantages. When constructed in teams, like the debrief meeting, it is a social innovation mechanism. The process of going through the analysis in a team setting often leads to innovative insights that are a product of both immersion in the data that’s been collected and the team interaction around it.

In addition, creating the affinity is inductive reasoning process that drives systemic thinking and helps the team identify opportunities that are not obvious. The affinity diagram is created from the bottom-up, and the threads embodied in the first-, second- and third-level labels emerge as part of the analysis process. Affinity is an exercise in recognizing patterns from individual points – inductive thinking. It is not a deductive exercise in sorting individual observations into predetermined categories. Encouraging inductive thinking is a key tool for helping teams identify and explore new opportunities.

Lastly, the affinity is a powerful tool for idea creation and communication. It plays a key part in grounding your later brainstorming in the actual customer observations as described in the section on ideation. It can also be used as an easily digestible summary of your team’s observations and findings for stakeholders. Many teams hang their affinity in their work areas or conference rooms as a handy reference to go back to over the life of their project.

2.2       Using Affinity Clustering with OVOC Data

Whether you are conducting a quick, small affinity or a multi-level affinity from OVOC data, the process for constructing the affinity is similar.

Follow these steps to create the affinity:

  • Reserve a room with enough space: You’ll need a conference room-sized space with plenty of wall space. Make sure you are allowed to tape or affix paper to the walls in this room. Take into consideration the number of team members and make sure the space is large enough to accommodate.
  • Creating an affinity with more than about 250-300 notes will take more than a day, so make sure you can leave your work in progress in place overnight.
  • Print out your Roses, Thorns and Buds: Print out the output of your debrief sessions on post-its. For a small project you can hand write them, but for larger projects, purchase printable post-its and print them as outlined in the section on Executing OVOC
  • Collect materials and prepare room: Make sure you have the following materials:
    • 1 roll of 36” wide white butcher paper
    • 1 roll of blue painter’s masking tape
    • 5-10 packages of 3×3 blue post-its for first-level affinity labels
    • 3-5 packages of pink post-its for second-level affinity labels
    • 1 package of green post-its for third-level affinity labels
    • 1 box medium blue Sharpies

To prepare the room, cover the walls with the butcher paper, tacking each sheet vertically with the masking tape. Overlap the paper an inch or two but don’t tape the individual sheets together.

Build the affinity: A three-level affinity is built in three stages. First, you will get all of the notes on the wall and grouped. After all of the notes are grouped, you will write first level labels. Then, you’ll cluster these first level labels and write second and third level

Stage 1: Get the notes up and grouped

 

Get the team together and hand each participant three sheets of post-its. Have each team member familiarize themselves with the sheets they have.

The first part of the affinity is done out loud. Start with any random note – a team member puts it on the wall and reads it off. Then, all team members look at their post-its to identify if they have another post-it that seems to be similar. When a team member finds one, s/he reads it off and sticks it directly beneath the first one. Repeat until no other post-its seem related. Then start a new column and repeat. Do this as a team and start slowly until the team gets the hang of it – it might take 20 or 30 post-its in 5 or 6 columns or so. During this start up period, make sure you’re placing one note at a time and the team member placing it reads it aloud.

At this point, do not worry about whether the post-it is a Rose, Thorn or Bud – just group them without regard to the type of note it is. You can look at the patterns of where the different types of notes end up later.

 

For small single-level affinities, this “out loud” mode can continue until all of the notes are placed. In this case, after all notes are placed, you can rearrange the groups until the team is satisfied, then simply label the groups. For fewer than about 50 or 60 notes, this is probably all you need to complete the affinity.

For larger affinities, you’ll continue the grouping portion, but once the team is comfortable, switch to more of an individual approach. Notes no longer need to be read out loud, and everyone can put notes up in tandem.

Some rules and guidelines for this stage:

  • Put up post-its in groups without too much justification or rationalization. This is where the word “affinity” comes in – it’s just like when you say you have an affinity with someone. You don’t know why, you just get along. We’re doing the same thing with the post-it notes at this stage.
  • No one owns a post-it or group, and anyone can move any post-it for any reason. In this way, groups constantly form and re-form. The team’s collective understanding gets baked in this way, and the clustering becomes truly owned by the team.
  • If two team members are moving a Post-it back and forth, stop and talk about it. If moving the Post-it helps to create a new group, then move it. If not, it doesn’t matter where the post-it goes, because ultimately the labels are what capture the new understanding anyway. There is no “right” place for any one post-it.
  • Hold off labeling the columns as long as you can – as soon as labels go up, our brains move too easily into category-sorting, deductive mode. For very large affinities, if you have to write temporary labels before you get all of the labels on the wall, go ahead and do so.

Stage 2: Write first-level labels

 

Labels tell the story of the data in the columns – they concisely capture the distinction that ties a group of notes together. And as described above, we write them in the first person, as if the customer or end-user were speaking to us from the affinity.  The next step is to write these labels.

At this point, you should have all the post-its on the wall, some in very long columns, some in shorter ones. Start by breaking very long columns up into groups of no more than 4-6 notes. Usually the longer columns that form in the first stage have several thoughts or distinctions buried in them. Break these out first, then write labels for the resulting groups.

Here are rules and guidelines for writing labels:

  • Common practice is to use blue post-its for first level labels.
  • Write labels in the first person, as if the customer or end-user were speaking.
  • Good labels create a story relevant to design.  The affinity is created to support design thinking. So the labels should be written from that point of view. Capture issues that are important to the user and that have design significance—in other words, that change the way you think about designing your offering.
  • Good first-level labels capture the issue that ties the individual notes beneath them together with enough detail so you don’t need to read the individual notes themselves. So don’t write “About” labels, like “This is about Smart Plant.” These labels force you to read the individual notes to find out anything interesting. Instead, move the meaningful distinction up into the label itself, like “We all use Smart Plant because we think it’s the industry standard.”
  • Aim for a maximum of 4-6 data points per labeled group. Singles are allowable if they carry a very significant point that is design relevant that you don’t want lost. But try to limit these “singletons” as much as possible.

It sometimes helps teams that are new to affinity to break apart the groups and write labels in pairs, but the usual practice is to do this individually. If you notice a teammate really struggling with a part of the affinity, you can switch sections with them.

Stage 3: Second- and third-level organization and labels

 

The next stage is simply a repeat of the clustering-labeling process, but this time with the blue first-level labels as the things to be clustered. In practice, this is much easier than the first level labels because many clusters that are similar are already near each other on the wall.

However, there is always the possibility that similar issues have arisen in separate areas of the emerging diagram. So your first step should be to have the team read through the first-level labels looking for clusters of blue first level labels that should go together but might be far apart on the wall.

After this, write second level labels that cluster and call out distinctions in groups of blue first level labels.

And lastly, write third-level labels that group entire sections of second level issues.

Many of the rules from first-level labels apply to second- and third-level labels, with just a few differences:

  • Common practice is to use pink post-its for second-level labels, and green post-its for third-level labels.
  • Aim for no more than 8 blue first-level groups per pink second-level label, and no more than 8 pink second-level groups per green third-level label.
  • “About” labels are okay at the second and third levels, as your constituents will already have read the first level labels.

Once your affinity is completed, you can capture the entire diagram in a Microsoft Word document. The biggest manual task involved is typing in the handwritten labels – cutting and pasting the individual notes goes surprisingly quickly. Once typed in, using Outline View, Word will allow the user to expand and collapse specific labels, and also allow display of the entire affinity at either first, second or third level.

2.2.3       Personas and More Advanced Models

Personas can also be constructed from your OVOC debrief notes and the affinity. There is a separate PMT Workshop and Handbook for Personas.

You should also use your OVOC data and affinity diagram to update the Touchpoint, Stakeholder and Experience Maps you created during the Kickoff Workshop. By this point, you have likely validated many of your predictions, but likely also discovered new touchpoints, stakeholders and details about both.

There are also a set of separate Contextual Design models that can be used, but we are not recommending them for PMT at this time.

 

2.3       The Ideation Session

This section gives a quick overview of Ideation – the process of creating and prioritizing ideas from your collected data. It’s a quick summary; for more information, there is a separate PMT Workshop for Ideation.

There are many ways to run an Ideation session, but this section summarizes a few important best practices.

Whatever method you use for Ideation, careful planning is crucial for a successful session.

Selecting a stakeholder team:

 

Make sure you have the right stakeholders in the room for the session. “The right stakeholders” means several things.

First, make sure you get the right expertise. This means having the right mix of development, engineering, finance, marketing, etc. to cover the kinds of ideas you’re looking to generate in your session. This mix will be project dependent. For example, an iterative project on a website will require design, development, marketing and IT, while an order to cash project might need credit, collections, sales, customer service, etc.

You also need the right personalities. Be realistic here. Ideation requires suspension of disbelief, an ability to withhold criticism until the proper time, and thinking outside the norms of current convention. There are those in your organization who can do this, and there are those who simply cannot. You know the difference. Take this into consideration.

Finally, you need the right political mix. Again depending on the project, make sure you have decision-makers who can commit to taking the next step, and make sure you don’t exclude people who can kill forward progress with a word.

Aim for no more than 8 or so people in an Ideation session. More people means it’s usually harder to manage and less creative overall. Resist the tendency for people to just forward the meeting notification to everyone they think might be mildly interested.

These constraints can all conflict, so realize that there is no perfect mix. But it definitely pays dividends to think about your invitations ahead of time.

Preparing the Room:

 

You want all of your OVOC data in the room, so like in the OVOC case, make sure you have a room with enough room for the participants to move around, enough wall space to contain your data, and that you can tape things to the walls.

At minimum, hang your affinity and personas (if you have them) on the walls of the room. Optionally, you can hang your stakeholder map, touchpoint map and/or experience maps as well, but usually affinity and personas work well by themselves.

Gather the following materials for your session:

  • 1 package of 3 x 5 yellow post-its per participant
  • 1 box blue medium Sharpies
  • Sticky flags or adhesive dots
  • A flipchart with flipchart paper

 

“Walking” the OVOC data:

 

The first step in Ideation is to immerse all of the participants in the OVOC data that your team collected. Do this by “walking the wall,” giving the entire team a chance to review the affinity and personas.

We take the time to read the data before brainstorming because we want the ideas that come out of the session to be grounded in the reality of what we found. Some people will still come into your session with their own pre-conceived ideas about solution direction, but at the very least you want to sensitize participants to what was found.

Give each stakeholder a pack of 3 x 5 post-its and a Sharpie to record design ideas, insights and questions while they read the affinity and personas. Write ideas, one per post-it, and stick them to the affinity and personas next to the part of the data that spawned the idea.

It should take your stakeholders about an hour to read through your affinity and personas, but this will obviously depend on the scope and number of your OVOCs.

These are best practices for “walking the wall” and immersing the team in the collected observations:

 

 

  • Try to create ideas that address higher-level labels in the affinity. It’s relatively easy to come up with one-off ideas that can “fix” any particular Thorn, but these low-level ideas are usually not high value, breakthrough innovations. Instead, try to think more systemically and drive your design ideas to address larger issues in your collected data. The affinity supports this holistic thinking explicitly – by driving up the label hierarchy, you’re forcing yourself to be more systemic with your ideas.
  • Do your reading in silence – think of this as an “art gallery” like experience. This respects the diversity of thinking types that you may have in your group. Some people think by talking, but many more need to concentrate to weave the threads together in their minds about what they’re reading and responding to. The brainstorming step that is coming next is where we discuss – right now it’s more of an individual exercise.
  • After the wall walk, summarize your team’s impression. Have a moderator stand in front of the flipchart and have team members offer the big issues they saw in the affinity and the personas.
  • In addition to summarizing the big issues in the customer data, you may also want to remind the team about the technological and business model capabilities that you can bring to bear as well. Technologies can be business-specific (e.g. “modular equipment,” “Aclar”) or generically available (e.g. “voice interface,” “Bluetooth”). Do this immediately after the customer data summary.

Brainstorming solutions – The diverge step:

 

Once your team is immersed in the data, you’re ready for brainstorming. Good brainstorming is a distinct two step process – the first is to create as many ideas as possible and the second is to evaluate and prioritize these solutions and narrow them down to the best and most actionable ideas.

The easiest and most straightforward way to brainstorm is to assign a moderator to record the team’s ideas. The moderator stands in front of the flipchart and records ideas team has. Standard brainstorming rules apply here – add to others’ ideas, don’t criticize or evaluate at this point, etc.

 

The PMT Workshop on Ideation also sometimes makes use of two other LUMA tools, Round Robin and Creative Matrix.

Prioritizing solutions – The converge step:

The simplest and most straightforward prioritization method is multivoting – LUMA calls this Visualize the Vote.

To run multivoting, follow these steps:

  • Give each stakeholder a small number of post-it flags or dots (typically 3 but can vary)
  • Tell them to decide on the ideas they think best meet these criteria: fit with the organization’s capabilities, fit with the organization’s business model, and fit with the organization’s cultural safe zone.
  • Once everyone has decided in their heads, each person places the flags on the flipchart next to the ideas they want to vote for.
  • Tally the votes.
  • Optionally, do successive run-offs.

Importance/Difficulty Matrix:

You can also plot your ideas on a flipchart against two axes. The horizontal axis represents how large an impact the solution would have on the customer experience, and the vertical axis represents how difficult that solution is to implement.

Exploring relationships between the ideas on this landscape can help the team prioritize solutions as well as create a roadmap for which ideas to address in which order.

More detail on Round Robin, Creative Matrix and Importance/Difficulty Matrix are given in the PMT Ideation Workshop.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s