Lifecycle Part 3: Building a Durable KOL Library

Parts 1 and 2 of this series turned a keyword into 32 vetted creators ready to contact. If you stop there — copy them into a Google Sheet, run the launch, archive the sheet — the next launch starts from zero. The whole funnel rebuilds itself in every launch, every quarter, forever. That is the loss this chapter is about.

The third stage of the lifecycle is the one most teams quietly skip. Discovery and vetting feel like work because they happen inside the campaign window. Library-building feels like it can wait until next quarter. By the time next quarter arrives, the only thing that compounded is the spreadsheet rot.

The third-launch problem

Ad-hoc workflows survive the first launch. They survive the second. They break on the third — predictably, on a pattern that every influencer-marketing team eventually hits.

The shape: by launch three you have three or four Google Sheets with different column orders and different definitions of "contacted". Half the duplicates are subtle — the same creator appearing as @handle in one sheet and @Handle in another. The new hire cannot find the perfect skincare creator from Q1 because nobody remembers which sheet they were in. Two creators have churned (followers dropped 40%, last post 8 weeks ago) and nobody noticed because nobody re-checked. The whole operation is one departure away from losing the institutional memory.

The reframe that fixes this: stop treating creators like contacts on a list and start treating them like accounts in a CRM. The unit of work is not "creator" but "creator + stage + history". When you have that, the library compounds.

KOL Lists are pipelines, not bookmarks

A KOL List in KOLens is not a saved-search bookmark. It is a first-class pipeline object: every creator in it has a stage, every action is logged, and the whole thing exports as a snapshot of the funnel at any moment.

The six pipeline stages baked into KOLens lists, in the order creators move through them:

1. Pending. Creator is in the list but not contacted yet. This is the post-vetting holding tank. Often where 40-60% of a list lives at any one time.
2. Contacted. Outreach has been sent. No reply yet. After 14 days here, the creator either advances to replied or moves to a dormant bucket for a re-touch later.
3. Replied. The creator has responded. This is the stage where the marketer takes over from the automation — pricing conversation, deliverables, brief.
4. Negotiating. Active deal conversation. The stage with the highest stakes per minute of attention.
5. Signed. Contract or PO is in. This stage feeds Part 4 (monitoring) — once a creator is here, you care about their performance, not just their existence.
6. Rejected. Explicit no, ghost, or declined-after-conversation. Keep them in the list with a rejection note so you do not re-approach in three months and embarrass yourself.

The discipline is moving creators through the stages in real time, not in batches at the end of the week. Every minute a replied creator sits in the contacted stage is a minute the rest of the team thinks that contact is still pending. Use the list at /lists as the live source of truth, not a weekly report.

Discovery Plans — the library that fills itself

Pipeline stages handle the creators you already know about. The harder problem is the creators you do not know about yet. The keyword you ran two months ago is now stale — new creators have emerged in the niche, the old top creators have moved on, the algorithm has surfaced different content. A library that does not refresh is a library that is dying.

Discovery Plans are the fix. A plan is a subscription-style automated scrape: you configure a keyword, the platforms (TikTok plus optionally Instagram, Douyin, X, Xiaohongshu), and the filters you would normally apply manually (follower range, country, niche). KOLens then re-runs the discovery on a daily schedule and adds the net-new creators to your workspace, deduplicated against everything you already have.

The math is friendly. A single Discovery Plan on a healthy keyword adds 400-800 net-new creators per month — not because TikTok produces that many new creators every month, but because the keyword surface is wide enough that every daily run pulls a slightly different slice. After 90 days the compounded total from a single plan is typically 1,200-2,000 creators, all deduplicated, all ranked, all with audience country and engagement already computed.

Plans live at /discovery-plans. The full playbook is in automated TikTok KOL Discovery Plans and the compounding-library piece in Build your own TikTok creator library.

find_similar_creators — lateral expansion

Discovery Plans answer the question "who matches my keyword today". They do not answer "who is adjacent to my best creators and would never have shown up under my keyword in the first place".

That is what the find-similar-creators tool is for. You pick a creator you already trust — someone who responded well in a past launch, someone whose audience country lines up perfectly, someone whose content style is exactly your brand voice — and KOLens returns others with overlapping audience country, content shape, and engagement profile.

This is lateral expansion. It pulls in creators who never used your keyword but share the qualities of your best anchor. It matters most for narrow niches where keyword search saturates fast — "minimalist skincare for sensitive skin in the US" has a small candidate pool, but "audience-similar to a known-good anchor" has a much wider one.

Run lateral expansion on your top 10-20 creators per month and the library widens in ways keyword discovery alone never will.

A 90-day playbook from 200 to 2,000

1. 1

   Week 1: Seed with three to five keyword searches.

   Run discovery from Part 1, vet with Part 2, save the survivors into a list called something like cohort-q3-raw. Expect 600-1,000 raw entries collapsing to 100-200 vetting-ready creators. This is the only manual day in the workflow.
2. 2

   Weeks 2-4: Turn the winning searches into Discovery Plans.

   For each keyword that produced a strong cohort, create a Discovery Plan with the same filters at /discovery-plans. Set the schedule to daily. By the end of week four you typically have three to five active plans adding 100-200 net-new creators per week each.
3. 3

   Months 2-3: Lateral expansion and monitoring.

   Identify the strongest 20-30 creators from the library so far. Run find-similar-creators on each. Push the promising results into Watchlist (Part 4) rather than straight into the contact pipeline — let them accumulate 30 days of signal before reaching out. By day 90 the library is typically 2,000-2,500 creators with 200-400 actively monitored and 50-150 in active outreach.

The compounding shape is what matters. Week one feels like the same one-off-search workflow you had before. Week four feels like an automated pipeline. Month three feels like an asset on the balance sheet of the marketing team.

The CRM and the library are different things

A common confusion at this stage is whether KOLens is replacing your CRM. It is not. The library is the discovery + intelligence layer; the CRM is where the actual conversation lives.

KOLens at /crm ships a lightweight conversations layer for teams that want the whole loop in one product, but it is designed to coexist with Notion, HubSpot, Airtable, or whatever you already use. The library handles "who exists, what do they look like, are they still active". The CRM handles "what did we say, when, and what was the reply".

For the snapshot-vs-live distinction (which library data is cached vs which is computed in real time), see the snapshot vs live workflow post.

Where KOLens enters the workflow

• /lists — the library itself. Every saved cohort, every pipeline stage, every export. The home base for the library workflow.
• /discovery-plans — automated daily scrapes. The lever that turns the library from a snapshot into a compounding asset.
• /crm — conversations and outreach state. Lightweight by design; coexists with your existing CRM.
• /dashboard — the workspace overview. Library size, active plans, pipeline-stage distribution, last-week additions.

A worked example — six launches in

The phonecase brand from Parts 1 and 2 ran the library workflow for two quarters. The trajectory:

Launch one ended with a 32-creator vetted list saved as launch-1-q1. Launch two started by pulling that list, removing the 12 already-signed creators, and merging in 47 new candidates from a Discovery Plan that had been running for six weeks. Vetting time per launch dropped from 90 minutes (launch one) to 20 minutes (launch two), because most candidates were already enriched with audience and cadence data.

By launch four, the library held 1,840 creators across three sub-niches. The marketing lead opened /lists and could pivot a new launch from "we need 15 creators for a magsafe accessory drop" to "here are 47 candidates already enriched and pre-ranked" in under five minutes. By launch six, the new hire who joined that quarter could run the whole library workflow without ever talking to the founder, because the pipeline stages and Discovery Plan configs told the story end to end.

The total cost of running this workflow for two quarters was about $80 in KOLens credits. The institutional memory it replaced was worth at least one full-time hire.

vs the alternatives

In the next chapter

A library you own is only valuable if you watch it. Part 4 — Monitoring is about the second half of the lifecycle: how do you know when a creator you signed three months ago has gone dormant, when a watchlist candidate has just gone viral, when a competitor has just signed your top underused creator. The Watchlist surface, alerts settings, and weekly digest workflow are the tools.

Continue to Part 4 — Monitoring TikTok Performance →

The full lifecycle series

1. Part 1 — Discovery: Finding TikTok candidates at scale
2. Part 2 — Vetting: Separating signal from noise inside the funnel
3. Part 3 — Library (you are here): Building a durable creator asset
4. Part 4 — Monitoring: Tracking the picks you made
5. Part 5 — AI / MCP: Claude integration for natural-language research