Knowledge Improvement Loop

A knowledge base is only useful if it is complete, accurate, and current. Simpli's KB, Reply, and Macro services form a continuous improvement loop that identifies gaps, fills them, keeps content fresh, and measures the impact. This page walks you through each part of the loop.

The gap-to-article pipeline

The most powerful feature of this loop is automatic gap detection. Here is how it works:

1. Customers ask questions. Tickets arrive about topics your KB does not cover.
2. KB /gaps identifies the missing topics. The gap analysis endpoint clusters unanswered queries and surfaces the topics that customers ask about most frequently with no matching article.
3. Your team creates articles. Using the gap report as a prioritized backlog, writers create new KB articles via the /articles endpoint.
4. Reply uses the new articles. When Reply generates drafts, it calls KB /search to find semantically relevant articles and uses them to ground its responses.
5. Quality improves. Drafts backed by KB content are more accurate and more consistent. Agent acceptance rates go up.
6. Repeat. As new gaps emerge (new products, new issues, evolving customer needs), the cycle continues.

Setting up gap tracking

Call the KB /gaps endpoint on a regular schedule -- weekly is a good starting cadence. The response includes:

• Topic clusters with query counts
• Sample queries for each cluster
• Suggested article titles

Create a simple workflow: review the gap report each week, assign the top gaps to writers, and track progress. As you close gaps, the report shrinks and your KB coverage grows.

Keeping content fresh

Articles that were accurate six months ago may not be accurate today. Products change, policies change, and processes evolve.

Stale article detection

The KB /stale endpoint flags articles that:

• Have not been updated within a configurable threshold (default: 90 days)
• Reference product features or policies that may have changed
• Have declining search relevance scores (customers search for the topic but do not engage with the article)

Review cadence

Set up a monthly review cycle:

1. Pull the stale article list from KB /stale
2. Assign articles to subject matter experts for review
3. Update, archive, or rewrite as needed
4. Track the stale article count over time -- it should trend downward as you build the habit

For fast-moving products, consider a biweekly cadence. For stable products, monthly or even quarterly may be enough.

From ticket to article

Some of the best KB content comes directly from resolved tickets. When an agent solves a novel problem, that resolution is a candidate for a new article.

The workflow

1. Agent resolves a novel issue. The resolution required research, troubleshooting, or creative problem-solving that is not captured in existing KB content.
2. Agent flags the ticket. Using a tag, internal note, or dedicated workflow, the agent marks the ticket as a potential KB article.
3. Extract key information. Pull the problem description, root cause, and resolution steps from the ticket.
4. Create the article. Use KB /articles to create a new article with proper tags, categories, and metadata.
5. Review and publish. A KB owner reviews the article for accuracy and clarity before publishing.

This workflow turns your support team into a content creation engine. The agents who solve the hardest problems are also the ones best positioned to document the solutions.

Macro analytics for content strategy

Macros (templates) and KB articles serve related but different purposes. Macros are for agents to use in responses. KB articles are for customers to find on their own. The Macro /analytics endpoint reveals which templates your agents use most frequently.

What high-usage macros tell you

• A macro used 50 times per week means customers are asking about that topic constantly. That topic should almost certainly be a KB article so customers can self-serve.
• A macro that is never used might be outdated, poorly written, or covering a topic that no longer comes up. Consider archiving it.
• A macro that is frequently edited before sending suggests the template needs updating or the topic is too nuanced for a one-size-fits-all response.

From macro to article

When a macro is heavily used:

1. Check if a KB article already covers the same topic
2. If not, create one -- the macro content is a solid starting draft
3. Add context, screenshots, and self-service steps that go beyond what the macro provides
4. Track whether ticket volume for that topic decreases after the article is published

This is how you shift volume from agent-assisted to self-service, which is better for customers (instant answers) and better for your team (lower volume).

Reply + KB integration

Reply generates draft responses by calling KB /search to find semantically relevant articles. The quality of Reply's drafts is directly tied to the quality and completeness of your KB.

How it works

1. Reply receives a ticket or conversation
2. Reply extracts the key topic and intent
3. Reply calls KB /search with a semantic query
4. KB returns the most relevant articles with relevance scores
5. Reply uses those articles as grounding context when generating the draft
6. The draft references accurate, up-to-date information from your KB

The virtuous cycle

• Better KB content leads to better Reply drafts
• Better drafts lead to higher agent acceptance rates
• Higher acceptance leads to faster handle times
• Faster handle times free up agent capacity to write more KB articles

This is the core loop. Every article you add makes Reply smarter, which makes agents more productive, which creates capacity to add more articles.

When Reply drafts are wrong

If Reply is generating inaccurate drafts about a topic, check the KB first:

• Is there an article covering that topic? If not, the gap-to-article pipeline should catch it.
• Is the existing article accurate and up-to-date? If not, update it and Reply's drafts will improve immediately.
• Is the article well-structured with clear answers? Ambiguous articles produce ambiguous drafts.

Measuring KB health

Track these metrics to understand whether your KB is improving over time:

Review these metrics monthly. The gap count and stale article count are your leading indicators. If gaps are growing or stale articles are piling up, the loop is not running fast enough.

Next steps

• KB service overview -- full API reference and configuration
• Macro service overview -- template management and analytics
• Reply service overview -- draft generation and feedback