You have the dashboard, the customer interviews, two analyst notes and a gut feeling, and they don't quite agree. The temptation is to pick the one that confirms what you already wanted to do and call it evidence. The better move is to treat the disagreement as information.

The quick version

  • Triangulation = look at the same question through genuinely different lenses (sources, methods, people, theories) so their errors don't line up.
  • Synthesis = combine those lenses into one coherent account that honours where they agree and explains where they clash.
  • Sense-making = the human act of turning that account into a plausible story you can act on, then updating it as reality answers back.
  • The payoff isn't certainty. It's a decision whose reasoning you can show your work for, and the honesty to notice when the lenses are really just one lens repeated.

Triangulation: don't trust one angle

The word is borrowed from surveying and navigation, where you fix an unknown point by taking bearings from two known ones. In research, the sociologist Norman Denzin made it a discipline. In The Research Act (1978) he set out four kinds of triangulation: data (different sources or times), investigator (different people looking), theory (different explanatory frames) and methodological (different methods entirely). The logic is simple and powerful: every method has a built-in bias, but different methods have different biases. Line them up on the same question and the biases that aren't shared start to cancel out.

Todd Jick gave this its most-quoted business framing in Administrative Science Quarterly in 1979, describing how mixing qualitative and quantitative methods lets each compensate for the other's blind spots, a survey tells you how many, the interviews tell you why, and where they converge you can believe the finding more. (This is the same instinct behind mixed-methods research: numbers and stories are stronger together than either alone.)

So the move is: before you act on a finding, ask "what's my second, independent angle?" Not a second slice of the same dataset, a different kind of evidence. If your churn theory rests only on the analytics, go read ten cancellation emails. If it rests only on anecdotes, go pull the numbers. One lens used twice is not triangulation; it's confirmation with extra steps.

flowchart LR
  Q(["One question:
why are renewals slipping?"])
  A("Lens A:
usage analytics")
  B("Lens B:
cancellation interviews")
  C("Lens C:
front-line sales notes")
  S(["Synthesised view"])
  Q --> A --> S
  Q --> B --> S
  Q --> C --> S
Triangulation: the same question approached through independent lenses, then combined. Leaders Loop

Synthesis: combining without flattening

Gathering angles is the easy half. The hard half is putting them together without quietly throwing away the parts that don't fit. The most rigorous version of this lives in evidence-based medicine: a systematic review is, in Cochrane's own words, "combining information from multiple studies investigating the same topic to comprehensively understand their findings." The craft is in the word comprehensively, you don't get to cite the three studies that agree with you and ignore the two that don't.

Most leaders will never run a formal systematic review, and shouldn't pretend to. But the underlying habit transfers: weight your inputs by quality, surface the ones that disagree, and write down the reasoning that gets you from "here are five sources" to "so here is what I think is true." Synthesis is not averaging. Averaging a strong signal and a weak one just contaminates the strong one.

Synthesis isn't taking the mean of your sources. It's deciding which ones have earned your trust, and showing your working.

Here is the honest limitation, and it's a big one. We tell ourselves triangulation works because the angles will converge on the truth. Sandra Mathison demolished that comfort in Educational Researcher in 1988 ("Why Triangulate?"). In practice, she showed, triangulation yields three outcomes, not one: convergence (the angles agree), inconsistency (they partly agree) and contradiction (they flatly disagree). Convergence is the rare, comfortable case. Most of the time you get inconsistency or contradiction, and that is not a defect in your data, it's the actual texture of reality. The job of synthesis is not to make the contradiction disappear. It's to explain it: which source is measuring something the others aren't, which is simply wrong, which clash is the most interesting thing you've learned all week.

So the move is: when sources clash, resist the urge to average or to pick a winner. Instead ask, "what would have to be true for both of these to be right?" Often the answer reveals a hidden segment, a timing effect, or a definition mismatch, a real insight the convergent story would have buried.

Sense-making: the story you can act on

Triangulation and synthesis are about the evidence. Sense-making is about the human in the middle of it. The organisational theorist Karl Weick built a career on this. In Sensemaking in Organizations (1995) he argued that people in confusing situations don't first analyse and then act, they act, watch what happens, and build a plausible story from the wreckage. Sense-making, in his framing, is retrospective (we make sense looking backward), social (we do it in conversation), ongoing and driven by plausibility, not accuracy. We don't need the true map; we need a map good enough to take the next step.

That last point sounds like a licence to wing it. Weick's own work shows the opposite stakes. In his study of the 1949 Mann Gulch wildfire (Administrative Science Quarterly, 1993), thirteen of the sixteen firefighters died, twelve smokejumpers and a ground-based fire guard, when their shared understanding of the situation collapsed: the fire stopped matching their map, the role structure dissolved, and the group came apart. His chilling lesson is that when sense-making breaks down, organisations don't just make a wrong decision; they lose the ability to function as an organisation at all. Plausible-enough maps are powerful precisely because we cling to them, which is also why we must be willing to drop them.

Deborah Ancona, at MIT Sloan, turned this into a leadership practice. In her chapter "Sensemaking: Framing and Acting in the Unknown" (2012), she describes a loop: build a plausible map of a shifting situation, test it through data, action and conversation, then refine or abandon it depending on how well it holds. The failure modes she names are worth pinning to the wall: rigidity (clinging to the map after it stops fitting), leader-dependence (waiting for one person to declare the meaning) and erratic behaviour (changing the map so often that nobody can act).

flowchart TD
  M(["Draw a plausible map
of the situation"])
  T("Test it: data, a small action,
and conversation with others")
  D{"Does reality
confirm the map?"}
  R(["Refine and act
with more confidence"])
  A(["Abandon it and
draw a new map"])
  M --> T --> D
  D -->|"mostly yes"| R
  D -->|"no, it's breaking"| A
  R --> T
  A --> M
Sense-making as a loop, after Ancona: map, test, then refine or abandon, never freeze. Leaders Loop

So the move is: treat your read of any messy situation as a draft map, not a verdict. Say it out loud to your team as a hypothesis ("my current read is X, what would change it?"), set a cheap test, and put a calendar marker on revisiting it. The willingness to abandon a map you're invested in is the whole skill. It also pairs naturally with knowing which decisions you can reverse, the more reversible the call, the cheaper it is to act on a rough map and learn.

A worked example

A product lead, "Priya," sees support tickets climbing after a redesign. Illustrative figures follow. The dashboard shows complaints up around 30% week-on-week, her first map is "we broke the checkout flow," and the room is ready to roll back.

Instead she triangulates. Lens one (data): the tickets are concentrated in one browser. Lens two (interviews): five quick calls reveal users aren't confused by the new flow, they're hitting a rendering bug on that browser. Lens three (sales notes): the front-line team reports the same accounts are otherwise happier with the redesign. Now she synthesises. The three lenses don't fully agree, and that's the gift: the "redesign is bad" story and the "users are happy" story are both true once you see the hidden split, a technical defect for a slice of users, layered on top of a redesign that's actually working. Averaging the signals ("net sentiment is flat, do nothing") would have buried both facts.

Her revised map: ship a hotfix for the browser bug, keep the redesign. She says it to the team as a hypothesis, sets a 48-hour check on ticket volume, and is ready to abandon it if the numbers don't fall. They do. A rollback, confident motion in the wrong direction, was avoided not by having better data, but by refusing to trust the first angle.

Frequently asked questions

Is triangulation just collecting more data?

No. It means looking through genuinely different lenses, different sources, methods, people or theories (Denzin's four types), so their independent errors don't line up. Five exports from the same dashboard is one lens five times, not triangulation.

What if my sources disagree?

Disagreement is a result, not a failure. Mathison (1988) showed triangulation produces convergence, inconsistency and contradiction, and the clashes are usually where the real learning hides. Ask what would have to be true for both to be right.

How is synthesis different from triangulation?

Triangulation gathers independent views of the same thing; synthesis combines them into one account that respects where they agree and explains where they don't. You triangulate to get the inputs; you synthesise to get the answer.

What is sense-making in plain terms?

Weick's idea: the act of turning a confusing situation into a story plausible enough to act on. It isn't about the single correct answer, it's about a workable map you test, then refine or abandon as reality responds.

Doesn't this slow decisions down?

Scale the effort to the stakes. For a reversible call, two quick angles and a short conversation are plenty. For an expensive, hard-to-undo decision, an extra hour triangulating is cheap insurance.

Related in the Toolkit

Where to go next