Algolia Cost Tuning

Algolia Cost Tuning

Overview

Algolia pricing is based on search requests and records. A search request is one API call (which may contain multiple queries via search({ requests: [...] })). Records are counted across all indices including replicas.

Prerequisites

• Access to current Algolia usage, billing, and index statistics for the application being reviewed.
• An Admin or monitoring credential used only from a trusted backend environment.
• A reporting window that distinguishes production traffic from load tests and bots.

Pricing Structure (2025)

What Counts as Records

• Every object in every index = 1 record
• Standard replicas duplicate records (multiply your count)
• Virtual replicas share records (no extra cost)
• Synonyms and rules do NOT count as records

What Counts as Search Requests

• searchSingleIndex() = 1 request
• search({ requests: [q1, q2, q3] }) = 1 request (multi-query)
• browse() = 1 request per page
• saveObjects() = NOT a search request (indexing operations are free)

Instructions

Examples

The audit and analytics calls below show how to measure records, requests, and no-result searches before changing replicas, caching, or query design. Measure a baseline first so savings claims are testable.

Step 1: Audit Current Usage

import { algoliasearch } from 'algoliasearch';

const client = algoliasearch(process.env.ALGOLIA_APP_ID!, process.env.ALGOLIA_ADMIN_KEY!);

// Check total records across all indices
const { items } = await client.listIndices();
let totalRecords = 0;
let replicaRecords = 0;

items.forEach(idx => {
  const records = idx.entries || 0;
  console.log(`${idx.name}: ${records.toLocaleString()} records, ${(idx.dataSize || 0 / 1024).toFixed(0)}KB`);
  if (idx.name.includes('_replica') || idx.primary) {
    replicaRecords += records;
  }
  totalRecords += records;
});

console.log(`\nTotal: ${totalRecords.toLocaleString()} records (${replicaRecords.toLocaleString()} in replicas)`);

Step 2: Replace Standard Replicas with Virtual Replicas

// Standard replicas: duplicate all records (doubles cost)
// Virtual replicas: share records with primary (no extra cost)

// BEFORE: 3 standard replicas = 4x record count
await client.setSettings({
  indexName: 'products',
  indexSettings: {
    replicas: [
      // 'products_price_asc',      // Standard: costs records
      // 'products_price_desc',     // Standard: costs records
      'virtual(products_price_asc)',  // Virtual: FREE
      'virtual(products_price_desc)', // Virtual: FREE
    ],
  },
});

// Virtual replica limitation: can only customize ranking and customRanking
// If you need different searchableAttributes or attributesForFaceting, use standard

Step 3: Use Multi-Query to Reduce Request Count

// BAD: 3 separate requests = 3 search operations billed
const results1 = await client.searchSingleIndex({ indexName: 'products', searchParams: { query: 'laptop' } });
const results2 = await client.searchSingleIndex({ indexName: 'articles', searchParams: { query: 'laptop' } });
const results3 = await client.searchSingleIndex({ indexName: 'faq', searchParams: { query: 'laptop' } });

// GOOD: 1 multi-query request = 1 search operation billed
const { results } = await client.search({
  requests: [
    { indexName: 'products', query: 'laptop', hitsPerPage: 5 },
    { indexName: 'articles', query: 'laptop', hitsPerPage: 3 },
    { indexName: 'faq', query: 'laptop', hitsPerPage: 3 },
  ],
});

Step 4: Cache Frequent Searches

import { LRUCache } from 'lru-cache';

// Cache popular searches — Algolia's own CDN caches are limited
const searchCache = new LRUCache<string, any>({
  max: 1000,
  ttl: 5 * 60 * 1000, // 5 minutes for product search
});

async function cachedSearch(query: string, filters?: string) {
  const key = JSON.stringify({ query, filters });
  const cached = searchCache.get(key);
  if (cached) {
    console.log('Cache hit — saved 1 search request');
    return cached;
  }
  const result = await client.searchSingleIndex({
    indexName: 'products',
    searchParams: { query, filters },
  });
  searchCache.set(key, result);
  return result;
}

Step 5: Delete Unused Indices

// Audit and clean up test/development indices
const { items } = await client.listIndices();
const devIndices = items.filter(i =>
  i.name.startsWith('test_') ||
  i.name.startsWith('dev_') ||
  i.name.startsWith('ci_')
);

for (const idx of devIndices) {
  console.log(`Deleting unused index: ${idx.name} (${idx.entries} records)`);
  await client.deleteIndex({ indexName: idx.name });
}

Step 6: Monitor Usage with Analytics API

// Track search volume trends
const { count: searchCount } = await client.getSearchesCount({
  index: 'products',
  startDate: '2025-01-01',
  endDate: '2025-01-31',
});
console.log(`Search requests this month: ${searchCount.toLocaleString()}`);

// Identify no-result queries (wasted searches users will retry)
const { searches } = await client.getSearchesNoResults({
  index: 'products',
  startDate: '2025-01-01',
  endDate: '2025-01-31',
});
console.log('Top no-result searches (fix these to reduce retries):');
searches.slice(0, 10).forEach(s => console.log(`  "${s.search}" — ${s.count} times`));

Cost Reduction Summary

Output

The review identifies the record and request drivers of the current bill, produces a prioritized reduction plan, and gives measurable before/after signals for each change. It does not alter billing settings by itself.

Error Handling

Resources

• Algolia Pricing
• How Algolia Counts Records
• Virtual Replicas
• Analytics API

Next Steps

For architecture patterns, see algolia-reference-architecture.