This guide explains what FANlines does, what to gather before you start, how to work through the tool’s four steps, how to read the report it produces, and where the tool’s assumptions bend. It is the companion reference for the FANlines tool.

What FANlines Does

Every researcher knows the best practice: don’t just find your ancestor in the census — read the whole page. Then read the page before it and the page after it. The enumerator walked a route, and the households recorded around your ancestor or research subject were, in most cases, the people who lived nearby. These are people who witnessed deeds, administered estates, married their daughters, and moved west with them. This collateral or cluster research is often known as FAN analysis (Friends, Associates, and Neighbors), a term defined by Elizabeth Shown Mills. In many cases the “F” can represent Friends or Family, as a reminder to also consider extended family. Seeking out your subject’s FAN is one of the most productive techniques in a genealogist’s toolkit, especially for brick walls where direct evidence has run out.

U.S. census data is a rich dataset for FAN analysis, yet is arguably tedious to do well. Reading the surrounding households in one census year is easy. Doing it across six census years — tracking which neighbor families persist, which ones appear and vanish, which ones followed your subject across a state line, and which unfamiliar surname keeps turning up two dwellings away — that is a correlation problem. Tracking several households across multiple decades is exactly the kind of clerical work that swallows research time when done by hand.

FANlines does the clerical work. You paste census page data, decade by decade, from the table view on major subscription record sites (such as Ancestry). The tool then structures each page into households, follows your research subject across the decades you supply, and surfaces the patterns that a careful and persistent researcher would find by hand:

The output is a standalone HTML report you can save, print to PDF, or file with your research notes. The report includes the data inventory, source citations, the analyses you selected, recommended next steps, documented negative findings, and a full methodology statement. Alongside it, your work saves as a local JSON research file that belongs to you — it represents the census data you loaded into the tool, and you can re-load it later to adjust or re-perform the analysis, or to add additional census pages.

What FANlines Is Not

The current version of FANlines does not search any database, fetch any image, or connect to any service. It analyzes only what you paste into it. It draws no conclusions: every analytical output is framed as a hypothesis or research lead to be tested against original records. It will never tell you two households are the same family; instead it shows you the evidence and leaves it to you to draw conclusions, as you would in a proof argument.

Your Data Never Leaves Your Browser

FANlines runs entirely on your own computer. Pages you paste are processed in the browser tab; nothing is transmitted or uploaded anywhere. Your research exists in three places local to the computer you are using: the current browser session, a local autosave that protects your current session against an unexpected browser crash, and the JSON file you download. It is recommended to use the “Save JSON” feature often, to ensure your raw data is saved in a way the tool can use now and in the future. FANlines is free and requires no account.

Before You Start: What to Gather

FANlines works from the table view of census pages on major subscription record sites (Ancestry). This is the spreadsheet-style index view that lists every person on the page in rows and columns, not the image itself. Supported censuses are the U.S. federal population schedules, 1830 through 1950 (there is no 1890 option).

For each decade you want to analyze, plan to capture, at minimum, three pages: the page before your subject’s page, the subject’s page, and the page after. The enumerator’s walking path means the adjacent pages hold the nearest neighbors — and a subject at the top of a page has most of their close neighbors on the previous one. Pasting more pages widens the net, and you can paste non-adjacent runs from the same district when both matter to your research (the tool handles the gap — more on that below).

Have the census images available as you work, for two reasons: the FANlines intake preview lets you correct index transcription errors on the spot, and the sheet number you enter should be the one printed on the census image itself.

Using the Tool

FANlines works in four steps, shown at the top of the tool: Start → Enter census pages → Link households → Analysis & report.

Step 1 — Start

Begin a new analysis, or resume from a saved JSON research file. If you were working recently in the same browser, an autosave banner offers to restore your session. (If your browser blocks local storage — common in private browsing — the tool tells you, and reminds you to download your JSON file often.)

Step 2 — Enter Census Pages

This is where most of your time goes, and where care pays off downstream.

Page metadata. Only the census year is required — everything else improves your citations. State, county, and civil division carry forward between pages so you aren’t retyping them (unless you need to); the sheet number clears for each new page, because it changes. The tool reports citation completeness per page, and you can choose to be as complete or incomplete as you wish on the citations. A partial citation is still a citation; the report simply marks it partial and lists what’s missing.

Pasting. On the record site (Ancestry), select the table view from the very top — including the column titles — and paste into the area FANlines indicates. The header row is not optional: FANlines identifies each census year’s format by its column headers, and a paste without them is rejected to avoid the tool providing erroneous information.

Validation. The moment you paste, a validation panel reports what it found: junk rows removed, the census format detected (and a warning if the paste matches a different year than you entered), column counts, and location consistency with the other pages in the decade. Warnings are presented clearly but do not block you from proceeding, so you retain the choice of making corrections or moving forward.

Preview and correction. Before a page is committed, you see every parsed row. This preview functions as a correction workbench: with the census image still open or in hand, this is the best moment to fix an index error (a mistranscribed age, a mangled surname, and so on) before it propagates into downstream analysis. Corrections never overwrite the record — the tool stores your corrected value alongside the recorded one, and the final report discloses each correction (“index reads ‘Lanks’; corrected by researcher against the image”), with an optional note recording your rationale. Click “Looks right — add this page” to commit.

The page list. Committed pages appear in a per-decade list where you can reorder them, edit their metadata, re-open them to make further corrections, or delete them. Pages are ordered automatically by the enumeration sequence where the census data includes one; for the decades that don’t (1830, 1840, and 1940), the tool asks for the printed sheet numbers so it can order pages correctly. FANlines falls back to your manual ordering if sheet numbers aren’t provided.

Subject household. After committing a decade’s pages, designate exactly one household as the research subject’s for that decade. This is the anchor for the analysis — every proximity measurement and FAN analysis for that decade is computed from it. You can optionally tag the subject person within the household (it defaults to the head), and you can change the designation at any time without losing any other work. The same pasted data can be re-analyzed from a different family’s perspective just by re-designating the subject — one of several ways the up-front pasting work keeps paying off.

The watchlist. Optionally, add surnames from your prior research. The analysis highlights them wherever they appear in surname recurrence, and calls out when they do not recur in the report’s negative findings — a record toward a reasonably exhaustive search.

Step 3 — Link Households

To follow a neighbor family from 1850 to 1880, the tool has to know that the Quisenberry household in 1850 and the Quisenberry household in 1880 are the same family. FANlines proposes these cross-decade matches from names, ages, and birthplaces — and you decide every one.

Each candidate is presented as a card with its evidence: surname agreement, a member whose age advances consistently across the gap, birthplaces that match. No match is confirmed automatically, at any strength, and no numeric score is ever shown. Instead you confirm against evidence the way you would in a proof argument. Confirm, reject, or leave a candidate undecided; undecided candidates are simply excluded from analysis and counted in the report. Confirmed links chain naturally (1850↔1860 plus 1860↔1870 becomes one family group across those three decades), and the tool blocks contradictions — you cannot confirm two different 1870 households as the same 1860 family without resolving it.

The matcher is built for the realities of the records. It compares across all decade pairs, not just adjacent ones, so a family missing from one enumeration is still matchable across the gap. Location plays no role in matching, which is consistent with the reality of migrating families and neighbors. Head succession (father in one decade, son as head in the next) surfaces as a normal candidate with the succession spelled out in the evidence. And a household that passed to a son-in-law can surface through a “possible continuation by marriage” candidate keyed on a daughter who reappears as a wife.

Step 4 — Analysis & Report

Click Run analysis and the panels appear: the correlation grid, neighbor stability, household evolution, surname recurrence, and the two marriage-origin scans, plus recommended next steps and negative findings.

A surname-variant grouping toggle controls how recurrence groups spellings: exact matching, Soundex (Kowalski/Kowalsky grouped, every member spelling listed), or Soundex ignoring the first letter — which catches first-letter transcription errors like Hanks/Lanks, at the acknowledged cost of sometimes grouping genuinely distinct surnames. Grouped rows always list every member spelling, with a caution to verify before acting.

Making changes. If you change anything after a run — add a page, correct a cell, confirm a link, re-designate a subject — the results are marked stale, a banner makes this visible, and the report cannot be generated until you re-run.

Building the report. Choose the sections to include; the data inventory, sources, and methodology are always included. The report downloads as a single self-contained HTML file — open it anywhere, share it, or print it to PDF.

Reading the Report

A brief orientation to the sections, in their fixed order:

Data Inventory and Sources. What you entered, per decade: pages, sheets, household and person counts, citation completeness, and any warnings accepted at entry. Sources render one citation per contiguous page run, with corrections, realignments, and any page-continuity assertions you made disclosed as footnotes. The report errs on being direct about its own inputs.

Cross-Decade Correlation Grid. The subject family’s recorded attributes (location, head, spouse, children present, occupation, birthplaces, and more) laid out decades-across, attributes-down, in the manner of a proof argument. Consistent progressions render plainly; expected changes (a succession, a migration, a birthplace shift explained by a historical boundary change) carry a labeled marker; genuine conflicts are flagged unmissably. A flagged conflict is a lead for resolution against the original images.

Neighbor Stability. Confirmed family groups ranked by persistence near your subject, with a proximity tier for each covered decade: T1 within 5 dwellings, T2 within 15, T3 in the pasted window but farther, and “same structure” for households sharing a building. The strip distinguishes a decade where the group is absent from your data (“—”) from a decade you never entered, which simply doesn’t appear — a small distinction that matters enormously when you cite the analysis later. Migration flags call out the strongest FAN signals: a family present on both sides of your subject’s move, or one that reappears in the new location after a gap, consistent with chain migration.

Household Evolution. A timeline of observed changes in the subject household between enumerations: departures, arrivals, head successions, occupation changes, migrations, and derived facts such as child deaths implied by children-born/children-living figures. Every event is an observed change in the records.

Surname Recurrence. Which surnames appear in how many of your covered decades, with person and household counts and a presence strip. Recurrence counts surname presence only — FANlines does not assert that same-surname households are one family.

Maiden-Surname and Daughter Marriage Hypotheses. The backward scan looks for each wife in your subject family as a daughter in a nearby different-surname household in an earlier decade; the forward scan looks for each departed daughter as a wife under a different surname in a later one. Candidates are ranked by age-band fit, proximity, and how rare the given name is in your pasted data — a matching “Permelia,” for example, outranks a matching “Mary” based on commonality. The report distinguishes three outcomes carefully: “candidates found,” “searched with no candidates,” and “search could not run” because your pasted window contains no decade before the wife’s first appearance — a coverage gap.

Recommended Next Steps and Negative Findings. Next steps translate the strongest patterns into record-specific research suggestions. Negative findings document what was searched and not found, what could not be searched and why, and the standing evidentiary gaps (the 1890 destruction, decades outside your window, etc.).

Methodology. A full statement of how the tool works and where its assumptions bend, suitable for filing with the report.

What FANlines Can and Cannot Do

Enumeration order approximates geography — it is not a map. The tool’s central assumption is that the enumerator’s route puts neighbors near each other on the page. This assumption is strong but imperfect: routes occasionally backtracked, multi-unit buildings compress dwelling counts, and district boundaries can split near-neighbors onto pages you’d never think to paste. Proximity tiers are lead-ranking heuristics built on this assumption, and the report says so.

Proximity is computed within each decade only. A family’s move between decades never affects any proximity figure; each enumeration is an independent window. This is a feature, not a limitation — it is what lets migrating families be a primary use case rather than a corruption of the math. Within one decade, however, the tool supports one enumeration sequence (one district, one place); if a family genuinely appears in two places in one census year, analyze them as separate research files. The tool warns if pages in one decade carry conflicting locations.

The tool works from the index, not the image. Everything FANlines knows, it knows from the transcribed table view you pasted (from Ancestry). Index transcription error is possible throughout; the census images remain the authoritative source. This is why the correction workbench at intake exists, why corrections carry provenance, and why the report is transparent about changes you’ve made.

Early and awkward decades are handled to the extent possible. The 1830 and 1840 censuses name only heads of household, so no wives or daughters exist in that data to scan; the marriage-origin features exclude those decades with a stated notice, and proximity there falls back to household order. The 1850–1870 censuses name everyone but record no relationships, so “wife” and “head” are positional inferences and are flagged as such. 1940 indexes no dwelling numbers, so page ordering leans on the sheet numbers you enter. And 1890 is simply gone; the report records the 1880→1900 interval as a twenty-year evidentiary gap wherever it applies.

Gaps are never bridged by invention. If you paste two non-adjacent page runs from the same district, the tool treats them as separate segments: households never auto-merge across the gap, and positional proximity is never computed across it. A cross-segment pair reads “true proximity undetermined.” If you have verified against the images that seemingly gapped pages are in fact consecutive enumeration (viewer numbering quirks cause this regularly), you can assert that explicitly within the natural flow of using FANlines, and the assertion is disclosed as a footnote in the report. The control is worded as an evidence assertion.

Ages are soft evidence and are treated that way. Persons align automatically across decades only when their recorded ages advance within three years of the gap. Discrepancies of four to ten years are presented to you with the recorded values shown; census ages are known to drift, and plausibility is your judgment, not the tool’s. A discrepancy of exactly ten years is flagged as consistent with a single-digit transcription slip. Beyond ten years, persons are not matched. Your decisions are recorded and carried, flagged, into the timeline, the grid, and the report.

FANlines’ birthplace comparison knows some history, not all of it. A curated equivalence table keeps known historical boundary and name changes (Virginia/West Virginia, Prussia/Germany) from being flagged as contradictions. The table is necessarily incomplete; an unrecognized pair is flagged for your judgment.

Common names produce coincidental fits. The marriage-origin scans cap at the ten strongest candidates per woman and rank by given-name rarity, precisely because a common name matches many people. The forward (daughter) scan is particularly noisy for common names, since young girls’ common given names match many later wives. Approach it this way: the rankings are the noise control, and the researcher is the filter.

Absence in your window is not absence in life. A daughter not found as a wife may have married outside your pasted pages, died, never married, or any of several other possibilities — the report is intentional about what is claimed and not claimed when a result is absent.

No probabilities, anywhere. FANlines asserts no statistical significance for any pattern. Persistence of neighboring families across decades is common in stable communities; co-location counts are descriptive lead-ranking heuristics, and the report’s methodology section states this in as many words. A number that looks like rigor but isn’t would be worse than no number at all — the sensible design choice for this particular tool.

Scope. U.S. federal population schedules only, 1830–1950. No state censuses, no non-population schedules, no non-U.S. records. Scope expansion is among the items on the roadmap.

Where FANlines Fits in Your Workflow

FANlines sits between record collection and proof argument. It does not replace reading the images — it gives you a reason to read more of them, and a structure for what you find. A productive pattern:

  1. When you locate your subject in a census, capture the surrounding pages into FANlines while the images are open — correcting the index against the image as you go.
  2. Repeat for every census in the subject’s lifespan, following them through migrations.
  3. Work the linkage queue with the same evidentiary care you’d apply on paper.
  4. Run the analysis and read the report as a lead sheet: the correlation grid’s conflicts, the strongest-stability neighbors, the maiden-surname candidates, the co-migration flags.
  5. Take the recommended next steps to the record groups they point at — marriage, deed, probate, tax — and file the report and its negative findings with your research log.

The JSON research file preserves all of it: your pages, your corrections with their provenance, your linkage decisions, and your analysis runs. Re-open it next year, re-designate the subject to a different family in the same neighborhood, and the same data yields a new perspective without an hour of re-entry.

Frequently Asked Questions

Is my research data uploaded anywhere?

No. All processing happens in your browser. Your data exists only in the browser session, the optional local autosave, and the JSON file you download.

Why won’t the tool accept my paste?

The one hard rejection is a paste with no detectable header row. Re-copy from the very top of the table view, including the column titles. Every other issue produces a warning you can read and proceed past.

Why don’t I see a match score or percentage on link candidates?

Deliberately. The internal ranking only orders the queue; the defensible artifact is the evidence list, and that is what you see and what you decide on. A displayed score would invite exactly the shortcut — “87%, good enough” — that the Genealogical Proof Standard exists to prevent.

The tool flagged a birthplace conflict that I know is a boundary change. Is that a bug?

No — it means the pair isn’t in the equivalence table yet. The flag asks for your judgment rather than suppressing the discrepancy; resolve it against the images and note it in your research log.

Can I analyze a family that appears in two counties in the same census year?

Not in one research file. FANlines stitches one enumeration sequence per decade; analyze the two locations as separate files.

My pages show gapped sheet numbers but I’ve verified they’re consecutive on the images. What do I do?

Use the consecutive-pages control in the page list — it asks you to assert that you’ve verified continuity against the images, merges the runs, and discloses the assertion in the report.

Does FANlines work with state censuses or records outside the U.S.?

Not currently. The parser is built on the column formats of the U.S. federal population schedules, 1830–1950.

FANlines is free, browser-based, and uploads nothing. Every output is a research lead under the Genealogical Proof Standard — never a conclusion.

About the Author

Nathan is an avocational genealogist and the founder of Evidence Toolbox. His research practice is grounded in the Genealogical Proof Standard, with primary-source work conducted at major repositories including the Library of Congress. He builds the tools on this site to solve problems encountered in his own research, and field-tests each one against real family lines before release. You can reach him at contact@evidencetoolbox.com.