Combine results from many tables without flattening nuance

This guide helps you turn many parallel discussions into one output leaders can act on. Use it when ten tables produce fifty "themes" and you need a synthesis that is both concise and honest. In a hurry: collect table outputs separately, merge only true duplicates, keep contradictions visible, and rank themes by decision impact.

Fast path (2-6 steps)

1. Export or collect raw outputs table by table before merging anything.
2. Label each point with source table and evidence strength.
3. Classify similarity as duplicate, near duplicate, related, or conflict.
4. Merge only duplicates and near duplicates.
5. Rank themes by "changes a real decision in 30 days".
6. Publish synthesis with short rationale for major merges.

When to use this workflow

Use this workflow when:

• you ran multiple tables in parallel
• stakeholders need one summary, not ten separate transcripts
• disagreement is expected and should remain visible

If your session is still live and decisions are not yet stable, first run [Capture decisions during discussion, not after](/guides/workflows/capturing-decisions-during-discussion).

Step-by-step synthesis flow

1. Start from evidence, not from theme names

Create a sheet with these columns:

• point text
• source table
• evidence type (example, opinion, repeated observation)
• decision relevance (high/medium/low)
• relationship to other points

Do not merge in the first pass. Early merging is where nuance usually gets lost.

2. Use a merge ladder instead of yes/no merging

For each similar point, assign one relationship:

• duplicate: same claim, same intent
• near duplicate: same claim, different wording
• related: connected but not equivalent
• conflict: opposite recommendation or assumption

Only merge the first two automatically. Keep related and conflict separate until you add rationale.

3. Surface contradictions as decision material

A common synthesis mistake is to "smooth out" disagreement so the report looks tidy. That usually creates bad downstream decisions because tradeoffs disappear.

When two tables conflict:

1. keep both viewpoints in the synthesis
2. add one line on the underlying difference (context, assumptions, constraints)
3. assign who decides next

If conflicts seem caused by different definitions, run a short alignment loop next session using [Aligning tables on shared definitions](/guides/facilitation/aligning-tables-on-definitions).

4. Rank by decision impact, not by frequency alone

Frequency matters, but it is not enough. A low-frequency risk can be more important than a high-frequency complaint.

Ask one question per candidate theme:

"If we act on this in the next 30 days, does a real decision change?"

Use this question to split outputs into:

• decision-driving themes
• contextual signals
• backlog observations

5. Make synthesis traceable for stakeholders

Before publishing, include a short method note:

• what you merged and why
• what you kept separate and why
• what still requires owner decision

This traceability is essential when people compare your synthesis against raw transcript excerpts or table summaries.

Practical failure modes

Failure mode: one loud table dominates synthesis

Fix: require at least two table sources for "shared pattern" claims, unless the claim is explicitly marked as an outlier.

Failure mode: everything becomes one umbrella theme

Fix: split umbrella themes into "same cause" vs "same symptom". Symptoms may look similar while root causes differ.

Failure mode: output is tidy but unusable

Fix: add one "decision implication" line to each top theme.

Failure mode: missing data from one table

If one table was disconnected, do not silently ignore it. Note the data gap and check session timeline. If needed, backfill from facilitator notes and mark confidence lower.

This guide is for...

Use this guide when your challenge is synthesis quality across many tables.

If your challenge is validating wording and ownership before people leave, use [Review session results before people leave the room](/guides/workflows/reviewing-session-results).

Related guides

• [Capture decisions during discussion, not after](/guides/workflows/capturing-decisions-during-discussion)
• [Review session results before people leave the room](/guides/workflows/reviewing-session-results)
• [Comparing themes between tables](/guides/analysis/comparing-themes-between-tables)
• [Building a workshop report](/guides/analysis/building-a-workshop-report)
• [Turning summaries into reports stakeholders can use](/guides/workflows/turning-summaries-into-reports)