How to search and filter your LinkedIn connections

Import it without mangling it

The file opens with a Notes: line and a paragraph of prose, then a blank line, then the real header row. Import it as-is and your spreadsheet treats that prose as the header, producing one column of nonsense.

Delete the first three lines first. Then check your dates: most exports use 21 Nov 2015, but US accounts emit 11/21/15, and if sorting by Connected On produces a random-looking order, some rows have imported as text.

Normalise employers before you count anything

This is the step that decides whether your numbers mean anything, and it is the one everyone skips.

Acme Ltd, ACME Limited, Acme Ltd. and Acme are one employer and four strings. So are Société Générale and Societe Generale. A pivot table counts them separately, so the company where you actually know five people shows up four times with one or two each, and never appears near the top of your list.

On a real 1,601-connection export, this collapsed roughly 1,400 apparent employers into meaningfully fewer real ones. Whether it is worth the twenty minutes depends entirely on whether you are about to make decisions from the counts.

Filtering by role, without missing people

Job titles are free text and nobody agrees on them. VP Ops, V.P. Operations, Vice President, Operations and VP of Ops (EMEA) are four strings for one job. Searching for any single one of them finds a quarter of the people you want.

The technique that works is to filter twice and intersect:

• Function words, for what they do: ops, operations, revenue, finance, marketing, engineering, product.
• Seniority words, for whether they can decide: head, director, vp, vice president, chief, founder, owner, partner, managing.
• Exclusions, for titles that sound right and cannot sign: coordinator, assistant, associate, intern, student.

Someone matching a function word and a seniority word is a candidate. It is crude, it over-includes, and it is still far better than guessing at exact titles.

Three sorts worth doing

By connection year, oldest first

Counter-intuitive and the most useful. Old connections are the ones you have forgotten, and forgetting is not the same as the relationship being weak. You met when you both had less to lose, and a message after a decade reads as genuine in a way that one after six months does not.

By employer, grouped

After normalising, a pivot on employer shows where you have depth. Three people at one company is a different kind of opportunity from one person each at three companies: you can ask one of them who owns the problem.

By what is missing

Filter for rows with no employer or no title. They are usually people who deleted their account or never filled the profile in, and knowing how many there are stops you wondering later why the totals do not add up. On the export we measured it was about one percent.

Where a spreadsheet runs out

Filtering answers *who matches this pattern*. It cannot answer *who is worth contacting first*, because that requires weighing people against each other rather than testing each against a rule.

It also cannot tell you how big any of these companies are, and the size band is usually what decides whether a title is your buyer. A spreadsheet will happily hand you a VP of Operations at a twelve-person company and one at a twelve-thousand-person company in the same filtered list, with nothing to distinguish them.

For a few hundred rows you can carry that in your head. For a few thousand you cannot, which is the point at which people either give up or reach for something else.