All guides

Capture & deduplication

Deduplicate leads by name and company

Review that two identical people match while two people with the same name at different organizations stay separate.

Start with a small representative sample

An event program lists speaker names and employers but lacks profile links. Staff records from adjacent tracks overlap and use inconsistent corporate suffixes.

Choose a source page you are authorized to use. Make a named list for the job so you can compare new rows with earlier entries from the same source. Include a typical row, a row with a missing or ambiguous value, and a row with formatting that tends to break the next step. A sample can be just a few rows as long as it gives you a way to verify the process before using a whole campaign.

Start with two entries that vary only by corporate suffix, then add one same-name contact at a different company and one potentially different person with the same employer.

Check the source before export

Start in the ordinary page or view containing the records and choose an appropriately named list. Click the toolbar capture button and inspect the added count. For a multi-page job, move to the next page yourself, wait for it to render, and capture into the same list. Compare the first few added records with the visible source. If the page is a virtualized grid, work from the segments you can actually see after changing the view.

Names can be transliterated more than one way. Keep profile evidence when available and manually review possible namesakes.

Write down the number of source records you expect to review. If only part of a source table renders at once, make sure your count describes the portion you actually captured. The evidence can be simple: a distinct first row, a last visible row, and a field you can find again later. On a repeated run, these reference points help explain the difference between a bad mapping and a changed source page.

Work through an example

The table shows what a useful before-and-after check looks like for Full name and company. Use it as a model for your own field map. Inspect the visible input first. Look at the output in its intended form, then consider what an operator would do if the source value turned out to be unavailable or different. When the field has more than one legitimate interpretation, include a second example with a different expected result.

FieldVisible exampleExpected result
Full name and companyDr. Ada Rao, MBA / Cedar Systems LLCAda Rao / Cedar Systems

Review that two identical people match while two people with the same name at different organizations stay separate.

Review capitalization and accents together. Case normalization can be useful without making spelling and transliteration judgments.

Try three entries for Jordan Kim at one company, including an honorific and two company suffix spellings. Then add Jordan Kim at a second company to make sure the fallback can make the intended distinction.

Review the cases that can slip through

Deduplicating by surname or first name creates large numbers of false matches.

A company can employ multiple people named Alex Chen. If your sources provide clear evidence that two rows are different people, use separate, explicitly named review lists so you can reason about the collision. For partner directories, keep the publication date or source context for roles that can change.

Check the sample after filtering and editing. Search should return both the ordinary case you expect and the exception you selected deliberately. If the exception is hard to find, add a review note or place it in a separate list. Keep an original source reference where it is useful. The goal of this pass is a repeatable decision about those few ambiguous rows before the selection takes on a life of its own.

Prepare the destination

Open the Workbench in its own tab and inspect raw source columns as well as derived ones. Use the search box to find representative names, employers, or terms. Update a cell if the visible source supports the change. A skipped duplicate may be an expected overlap from two pages, or an ambiguous match you need to look at. Keep the context of a match available before relying on the final total as a measure of what was captured.

Record why the fallback used the name and employer combination and keep variant spellings available for human review.

Label the first output a trial and record the list name, selection, target format, and the field that determines whether the row can be used. Compare at least one expected blank and an ordinary populated row after export. Keep test changes easy to roll back. For larger jobs, measure the number of rows that need manual handling so you know where the remaining review will take time.

Validate what the recipient will use

Compare the final presentation to the human decision that uses it. A CSV is suited to an editable follow-up, TSV works for a quick paste, a PDF supports reading in a meeting, and a ZIP keeps campaigns apart. Capture and export are separate decisions, especially when a list contains contact details that only one reviewer needs. If the downstream task needs just a narrow set of rows, apply a filter in the Workbench before the CSV, TSV, or PDF step.

Send the lead reviewer an example of an intended duplicate and a row kept because its company context differs. Explain how exceptions will be handled.

If you find two real people at the same company with the same normalized name, document the stable profile URL or other allowed identifying evidence available in that directory. Tell the next reviewer why a forced merge would hide useful records and keep the exceptions accessible.

Review one result where the converted value looked straightforward and one where your team had to choose between plausible interpretations. Give the recipient the updated row count and a practical next action. If review exposed a source inconsistency, note the observable issue and include enough context for the recipient to reproduce it. Close the loop by storing or removing the trial material according to how long your team needs it.

Record the minimum needed for a repeat run

Save a simple checklist with your source path, list name, review filter, format, and destination fields. Keep the choices about Full name and company near the top so the next operator can pick a meaningful test row. A useful recurring process records the fields that affected the outcome and the points where someone needed judgment. It can be rerun when the source layout changes without forcing every old decision into a script.

LeadSheet Clean keeps captured rows and derived values in the browser. Start with Standard Clean CSV or copy TSV when you need a quick review sample. Lifetime Pro adds unlimited export rows, specialized presets, PDF briefs, and multi-list ZIPs. Select the path that produces the format and selection required for your handoff, and verify the license or Free allowance before repeating a large export.