Reading Sequence Flow Results: Consensus Routes vs. Averaged Rank¶

Ten respondents each build a five-item sequence. Average each item's position independently across all ten, and you'll get back a single ranked list - one number per item, sorted low to high. It looks like an answer. The problem is that it can be an answer nobody actually gave.

Picture half the group building A, B, C, D and the other half building B, A, D, C. Average each item's position independently and item A comes out on top (it was picked first by half the group, second by the other half - a strong average position), with B a close second. The result reads as a single, confident sequence: A, B, C, D. But not one respondent in the study actually built that sequence. Half of them paired A right after B; the other half paired B right after A. Averaging threw away the one thing the question was actually asking - not which item is individually strongest, but what order things happen in, together.

Sequence Flow doesn't average positions independently. It analyzes item-to-item transitions - how often each item was placed immediately after another, across every respondent - and gives you two dedicated views built on that instead.

Table of Contents¶

• Why Averaging Independent Positions Breaks Down
• How Sequence Flow Reads Results Instead
• Route Flow: The Single Consensus View
• Route Network: Seeing the Full Picture
• When the Two Views Disagree
• Common Mistakes Reading Sequence Results
• A Worked Example
• FAQ

Why Averaging Independent Positions Breaks Down¶

The failure mode isn't unique to sequencing exercises - it's the same reason a straight importance-rating scale clusters everything near the top, and the same reason MaxDiff exists to force real trade-offs instead of independent scores (see our MaxDiff and TURF guide). What's specific to a sequencing question is that the thing being asked for is relational by definition: not "how do you feel about item A," but "what comes right after item A." A scoring method that treats each item's position as its own independent number can't represent that relationship at all - it can only ever report on items, never on the connections between them, which is exactly the part of the answer a sequencing question was built to capture in the first place.

This is also why a straight average can produce a sequence that's internally contradictory in ways that aren't obvious just from looking at the final ranked list - two adjacent items in the averaged output might never have actually been placed next to each other by a meaningful share of respondents, and the averaged list has no way to flag that it's happened.

How Sequence Flow Reads Results Instead¶

Every completed Sequence Flow section produces a full sequence per respondent, not just a set of independent item scores. Sequence Flow's analytics look at every adjacent pair in that sequence - which item came immediately after which, for every respondent - and build a picture of the transitions that actually occurred across the whole group, not just which items were individually popular.

From that transition data, two things fall out: a single dominant path (the specific sequence of transitions that best represents the group as a whole), and the full graph of every transition anyone actually made, including the ones that lost. Route Flow shows you the first. Route Network shows you both, together.

Route Flow: The Single Consensus View¶

Route Flow renders the dominant path as a straightforward step-by-step diagram - one card per item, in winning order, connected by simple lines. Alongside each transition, it shows a per-step reading of how much the group actually agreed on that specific placement, and it flags a transition where the group was close to a coin flip rather than letting a weak, barely-there consensus look identical to a strong, decisive one on the diagram.

That distinction matters in practice: a Route Flow diagram where every transition shows strong agreement is telling you something genuinely different from one where the overall sequence "wins" but several of its individual transitions were nearly split. The first is a sequence you can act on directly. The second is a sequence that's still your best single answer, but one where a specific step deserves a closer look - a follow-up question, a second round of testing, or just healthy skepticism about that one part of the order - rather than blanket confidence in the whole thing.

Route Network: Seeing the Full Picture¶

Route Flow answers "what's the winning sequence." Route Network answers a broader question: "how much genuine agreement or disagreement surrounds it." It renders every transition respondents actually built as a connected network - not just the winning path - with the dominant route from Route Flow highlighted inside that wider network in its own distinct color, so you can see the winning sequence and the full spread of what people actually did in one view.

This is the view to reach for when a Route Flow diagram looks clean but you want to know what else was going on underneath it - whether the losing alternatives were close runners-up clustered right around the winning path, or scattered and genuinely inconsistent with each other, which would suggest the group didn't share much of a mental model for the task at all.

Very rare or one-off transitions are already filtered out of the network before it's drawn, so what you're looking at is the meaningful structure in what people actually did, not every idiosyncratic one-respondent sequence diluting the picture.

When the Two Views Disagree¶

They won't disagree on what the dominant path is - Route Network highlights the exact same route Route Flow computes. What they can disagree on is how confident that shared answer deserves to be. A Route Flow diagram can show strong agreement on most transitions while Route Network reveals that the network around it is genuinely tangled - lots of near-equal alternative transitions that didn't quite win, rather than a clearly second-place runner-up. When that happens, trust Route Network's broader picture over Route Flow's clean-looking diagram: a dominant path surrounded by a dense, competitive network is a weaker, more contested answer than the same dominant path sitting alone with clearly weaker alternatives around it, even when both cases render an identical-looking Route Flow diagram on their own.

Common Mistakes Reading Sequence Results¶

Treating the dominant path as unanimous. It's the single best representation of the group, not a claim that everyone agreed. Check the per-step agreement reading in Route Flow, and the surrounding network in Route Network, before treating any individual transition as settled.

Ignoring a flagged weak link. A transition Route Flow flags as close to a coin flip is a genuine signal, not noise to scroll past - it's telling you the crowd didn't actually converge on that specific step, even though the sequence won overall.

Reading Route Network without checking Route Flow first. The network view is denser and easier to misread in isolation. Start with Route Flow to get the headline answer, then use Route Network to stress-test how confident that answer deserves to be - not the other way around.

Comparing results across two Sequence Flow sections with different item counts or different full-vs-partial settings as if they're directly comparable. A dominant path computed from a full 12-item sequence and one computed from a partial top-4 selection out of 12 answer meaningfully different questions, even when several of the same items appear in both.

A Worked Example¶

A subscription software company is deciding the order of five setup steps for a new admin console: connecting a data source, inviting teammates, setting permissions, configuring a first dashboard, and enabling alerts. They run a full-sequence Sequence Flow section with fifty respondents from their existing customer base.

Route Flow returns a dominant path: connect a data source, configure a first dashboard, invite teammates, set permissions, enable alerts - with strong agreement on every transition except one, flagged as a near-coin-flip: whether inviting teammates should come before or after setting permissions.

Rather than guessing, the team opens Route Network and finds that the network around that one flagged transition is genuinely split into two roughly equal clusters - one meaningful group of respondents consistently inviting teammates before setting permissions, another consistently doing the reverse - while every other part of the sequence shows one dominant pattern with only scattered, minor alternatives. That's a materially different finding than "the whole order is roughly this, with mild noise everywhere" - it tells the team specifically that two real, different mental models exist for one part of the flow, which they can now design around directly (for instance, by supporting both orders rather than forcing one), instead of shipping a single sequence and hoping the flagged uncertainty was noise.

FAQ¶

Do I need both views, or does one replace the other?
They answer related but different questions - Route Flow gives you the headline sequence and where it's shaky; Route Network shows you what's happening in the network around it. Use Route Flow first for the quick read, Route Network when you need to understand why a transition is uncertain.

Is the per-step agreement reading in Route Flow the same thing as a simple "percent who placed these two items adjacently"?
It's built from the same underlying adjacent-transition data described above - the exact figure shown reflects how consistently that specific step-to-step transition showed up relative to the alternatives at that point in the sequence, not a raw share of the full respondent pool.

What happens with a partial-ranking Sequence Flow section, where not every respondent placed the same number of items?
The transition analysis only uses the placements a respondent actually made - a respondent who placed three of twelve possible items only contributes the transitions between those three, not guesses about the other nine.

Why doesn't Sequence Flow just show me an average rank per item, the way some other tools do?
Because it throws away the relationship between items, which is the part of the answer a sequencing question is actually asking for - see Why Averaging Independent Positions Breaks Down above.

For more on forced-choice and relational methods that beat independent scoring generally, see our MaxDiff and TURF guide and Beyond Averages: The Professional's Guide to Survey Analysis.