Coder Legion - Recent posts tagged per

SEO Rates Philippines 2026: Costs & Packages | FloxoLab

Sun, 18 Oct 2026 15:50:00 +0000

Direct answer: published 2026 Philippine rate guides put limited freelance retainers around ₱5,000 to ₱15,000 a month, agency starter packages around ₱25,000 a month, and broader agency campaigns from about ₱50,000 to ₱150,000 or more. Published hourly benchmarks run around ₱1,500 to ₱5,000, while defined projects and audits vary with scope.

Those numbers are useful benchmarks, not an official Philippine market average. Many providers do not publish prices, and two quotes at the same rate can include very different work. This guide helps you compare rates and packages. It does not rank agencies, estimate your return without business data, or promise a ranking outcome.

The short answer, by pricing model

The table combines current published Philippine rate guides with a live public package page. It shows what is visible in the market, not a guaranteed quote for every provider.

Engagement	Published 2026 benchmark	Useful when
Limited freelance retainer	₱5,000 to ₱15,000 per month	You need a narrow scope and can handle some implementation in-house
Agency starter package	About ₱25,000 per month on one current public package page	You need defined on-page, technical, local, and reporting work
Growth or mid-market campaign	About ₱50,000 to ₱150,000 per month	You need ongoing content, links, technical support, or several markets
Enterprise or multi-market campaign	₱150,000+ per month	The site, competition, content operation, and reporting needs are large
Hourly consulting	₱1,500 to ₱5,000 per hour	You need advice, troubleshooting, training, or a bounded review
Defined project	About ₱10,000 to ₱80,000 in one current local rate guide; complex migrations can cost more	You have a clear deliverable and acceptance criteria
One-time SEO audit	About ₱10,000 to ₱50,000 in one current local cost guide	You need diagnosis and priorities before implementation

Evidence and exchange-rate note: SEO.com.ph is a market participant, so its published guides and package page are transparent examples rather than an independent census. The global Ahrefs survey of 439 providers also shows large differences by provider type, experience, region, and pricing model. Dollar comparisons on this page use the Bangko Sentral ng Pilipinas August 2026 average through August 20 of ₱61.202 per US dollar. Prices and exchange rates may change.

What changes an SEO quote

A price becomes meaningful only after you can see the scope behind it. Ask what changes when each of these dimensions grows:

Market and geography. One local service area is different from a national ecommerce category or several locations.
Site condition. A stable ten-page site and a migration with duplicate URLs, rendering problems, and thousands of products need different work.
Implementation ownership. Recommendations cost less than a provider who also edits templates, content, internal links, redirects, and structured data.
Content scope. Research, expert input, writing, editing, media, publishing, and updates should not be collapsed into one unexplained line item.
Authority work. Digital PR and legitimate outreach take research and relationship work. A raw backlink count does not describe quality.
Team and communication. A senior consultant, specialist team, account manager, and reporting process have different cost structures.

How the main pricing models differ

Monthly retainer

A retainer fits work that genuinely repeats: content research and production, digital PR, multi-location maintenance, ecommerce changes, monitoring, testing, and implementation on a site that keeps evolving. Ask for the work planned each month, who performs it, what access is required, and how unfinished work carries over. The separate guide to what a typical SEO retainer actually does gives a fuller scope check.

Defined project

A fixed project fits a migration, technical cleanup, content consolidation, template improvement, or initial optimization sprint. The proposal should identify the URLs or templates in scope, deliverables, implementation responsibility, dependencies, acceptance checks, and what happens when new problems appear.

Hourly consulting

Hourly work fits training, troubleshooting, review calls, and tasks whose duration is easier to control than their final output. Set a cap and decide what evidence or handoff you receive. An hourly rate without a time estimate is not yet a budget.

One-time audit

An audit should explain what is wrong, why it matters, the evidence behind the finding, the exact affected URLs, and how to verify a fix. Clarify whether implementation is included. A tool export by itself is not the same as a prioritized audit.

Performance-based pricing

Define the measured outcome before accepting a performance fee. Rankings, traffic, qualified leads, and revenue are not interchangeable. Document the data source, baseline, attribution, brand-query treatment, payment window, and what happens when demand or conversion changes outside the provider's control.

When one-time work is enough, and when it is not

Some businesses first need a finite project: fix indexability conflicts, repair a migration, improve core service pages, clean up duplicate intent, or establish measurement. If the site and offer will remain stable afterward, a project plus periodic review may be enough.

Other businesses have recurring work. A store adds products, a publisher competes on new topics, a multi-location company opens branches, or a service firm needs continuous content and authority building. Those cases can justify a retainer because the underlying workload does not end after the first technical fixes.

Use your own data to decide which situation applies. The Search Console pre-audit shows whether important pages are indexed, receiving impressions, and assigned to the intended queries. The SEO break-even guide then connects qualified visits, lead rate, close rate, and contribution profit to the budget decision.

How to compare a low-cost SEO package

A low price is not proof of bad work. It usually means the scope, time, or senior involvement must be narrow. Ask the provider to make that boundary explicit:

Which pages, locations, products, or queries are included?
Does the provider implement changes or only recommend them?
How many hours or named deliverables are included?
Who creates and approves content?
What link work is performed, and can every placement be inspected?
Which technical changes require your developer?
What is excluded, and what triggers an additional quote?
Can you cancel, export the work record, and retain ownership of every account?

If a provider can answer those questions clearly, a narrow package may be useful. If the proposal promises comprehensive SEO but cannot name the work, the mismatch matters more than the price.

A practical way to set the budget

Define the business outcome. Choose qualified leads, bookings, sales, or another result that can be checked. Do not begin with a keyword count.
Record the baseline. Use Search Console, sales records, and current landing pages before paying for a change.
Separate finite and recurring work. Put migrations, audits, and initial fixes in a project scope. Put ongoing content, outreach, monitoring, and repeated implementation in a recurring scope.
Compare the same scope. Normalize proposals by deliverable, implementation owner, content volume, link method, reporting, exclusions, and contract length.
Calculate the break-even point. Estimate how many additional qualified customers must cover the fee, then test whether the required search demand is realistic.
Choose a review date. Agree on leading indicators and a decision point before the engagement begins.

What not to do when comparing SEO rates

Do not choose by price alone. A cheaper narrow scope and an expensive full-service scope are not competing products.
Do not buy a keyword or backlink count without inspecting the method. Volume does not show relevance, quality, or business value.
Do not accept guaranteed rankings. Google states that no one can guarantee a number-one ranking.
Do not grant broad account access before the work requires it. Match permissions to a named task and retain company-controlled ownership.
Do not treat reporting as implementation. A report should show decisions, completed work, evidence, and the next action.
Do not assume a high fee guarantees a return. The proposal still needs a plausible path from search demand to qualified business outcomes.

SEO proposal verification checklist

The proposal names the pages, locations, products, or query groups in scope.
Recommendations and implementation are priced separately when different people own them.
Content includes research, review, publishing, and update responsibilities.
Link work is described by method and reviewability, not only quantity.
Technical dependencies and required access are explicit.
The baseline, reporting source, review interval, and decision metrics are agreed.
Exclusions, extra-work rates, contract length, cancellation, and account ownership are clear.
No ranking, traffic, or revenue outcome is presented as guaranteed.

The bottom line

Published SEO rates in the Philippines cover a wide market. Limited freelance retainers can start around ₱5,000 to ₱15,000 a month, public agency packages can start around ₱25,000, and broader campaigns can reach ₱50,000 to ₱150,000 or more. The useful comparison is not the number alone. It is the work, implementation ownership, evidence, exclusions, and business case behind that number.

Frequently asked questions

How much does SEO cost per month in the Philippines?

Current published benchmarks span roughly ₱5,000 to ₱15,000 for limited freelance retainers, about ₱25,000 for one public agency starter package, and ₱50,000 to ₱150,000 or more for broader agency campaigns. Treat these as comparison points, not an official market average.

Is an SEO audit included in the monthly price?

Sometimes. A provider may include an initial audit in a retainer, sell it separately, or provide only automated checks. Ask whether the audit includes manual diagnosis, affected URLs, priorities, implementation instructions, and a verification step.

Should a small business hire a freelancer or an agency?

Choose based on the required scope and ownership. A freelancer can fit a narrow project, consulting, or a business with internal implementation capacity. An agency can fit work that needs several specialties and continuous production. Check the actual named team and deliverables rather than assuming the provider label guarantees a capability.

How much is $150 in Philippine pesos?

At the BSP August 2026 average through August 20 of ₱61.202 per US dollar, $150 is about ₱9,180. The live amount changes with the exchange rate and payment provider.

Sources checked

Rate benchmarks, public package prices, the exchange rate, global context, and Google's hiring warning were verified on August 21, 2026.

Need a bounded diagnosis before choosing a package?

FloxoLab's $150 audit is about ₱9,180 at the exchange rate used above. It reviews the site and available Search Console evidence, identifies the affected URLs, and returns a prioritized fix list without a monthly commitment or ranking promise.

Explore the SEO audit

Originally published on FloxoLab.

AI Email Classifier Cost per 1,000 Messages: Real Math

Thu, 08 Oct 2026 15:50:00 +0000

For 1,000 typical text emails, the model API can cost roughly $0.05 to $1.08, or about PHP 3 to PHP 66 at a rounded PHP 61.55 per US dollar. That estimate assumes 900 input tokens and 30 output tokens per message, no attachments, no retries, and one classification call per email.

Partner disclosure. This guide includes a partner link to n8n Cloud. If you sign up through it, I may earn a commission at no extra cost to you.

The low token bill is real, but it is not the whole operating cost. Your automation platform, monitoring, human review, and the consequences of a wrong route can cost much more than inference. Prices and exchange rates may change, so keep the formula and replace the rates instead of trusting a static total.

Practical answer: budget about $0.10 to $0.25 per 1,000 ordinary emails for a small classification model, then evaluate accuracy on your own inbox. A model that costs five cents but misroutes valuable leads is not the cheaper option.

The workload behind the estimate

A useful comparison needs the same input and output size for every model. This article uses a typical business email workload rather than each provider's most flattering example.

Component per email	Assumed tokens	What it includes
Classification instructions	250 input	Category names, definitions, edge-case rules, and output instructions
Email content	650 input	Subject, cleaned body, and a small amount of sender context
Classification result	30 output	One label plus compact structured fields such as reason and review flag
Total for 1,000	900,000 input 30,000 output	No attachment extraction, second model call, retry, or human-written reply

Token counts vary by provider and tokenizer. Treat 900 and 30 as planning inputs, then replace them with usage data from the API response or provider dashboard. Long quoted threads, HTML, signatures, disclaimers, and pasted documents can multiply the input without improving the decision.

The cost formula

Provider pages normally quote text prices per one million tokens. The calculation is:

cost = (input_tokens / 1,000,000 × input_rate)
     + (output_tokens / 1,000,000 × output_rate)

Typical 1,000-email workload:
cost = (900,000 / 1,000,000 × input_rate)
     + (30,000 / 1,000,000 × output_rate)

Example for OpenAI GPT-5.4 nano at the rates checked on July 20, 2026:

input  = 0.9 × $0.20 = $0.1800
output = 0.03 × $1.25 = $0.0375
total  = $0.2175, rounded to $0.22 per 1,000 emails

Current model API comparison

These totals use standard synchronous API pricing, without free tiers, batch discounts, cached-input discounts, regional uplifts, or taxes. The selected models are small or cost-focused options that can plausibly handle classification. Model quality is not equivalent, so the last column is a cost comparison, not a ranking.

Rate check: every provider price in this table was checked against the linked official source on July 20, 2026. The PHP estimates use a BSP rate of PHP 61.548 per US dollar checked on the same date. To update any peso estimate, use PHP estimate = USD total × current PHP per USD. For example, $0.2175 × 61.548 = PHP 13.39.

Model	Input / 1M	Output / 1M	Typical 1,000 emails
Groq Llama 3.1 8B Instant	$0.05	$0.08	$0.047 / about PHP 3
Google Gemini 2.5 Flash-Lite	$0.10	$0.40	$0.102 / about PHP 6
OpenAI GPT-5.4 nano	$0.20	$1.25	$0.218 / about PHP 13
Anthropic Claude Haiku 4.5	$1.00	$5.00	$1.05 / about PHP 65
OpenAI GPT-5.6 Luna	$1.00	$6.00	$1.08 / about PHP 66

OpenAI describes GPT-5.4 nano as suitable for classification and other simple high-volume tasks. Its newer GPT-5.6 guidance points cost-sensitive new workloads toward Luna, but Luna costs more in this narrow token calculation. Test both only if the quality gain matters for your categories. Google calls Gemini 2.5 Flash-Lite its smallest and most cost-effective model for at-scale use. Anthropic's current Haiku tier costs more per token, while Groq's Llama 3.1 8B rate is the lowest in this selected set.

Email length changes the bill, but not equally

Output is tiny for a classifier, so input length usually drives the total. This sensitivity table keeps output compact while changing the amount of text sent to the model.

1,000-email workload	Input / output	Gemini 2.5 Flash-Lite	GPT-5.4 nano	Claude Haiku 4.5
Short	400k / 25k	$0.05	$0.11	$0.53
Typical	900k / 30k	$0.10	$0.22	$1.05
Long	2M / 40k	$0.22	$0.45	$2.20

The long scenario is still inexpensive as a token bill. It is also a warning: paying to resend an entire quoted conversation may expose unnecessary customer data and make the classification noisier. Clean the message before the model call.

What the token estimate leaves out

Automation platform cost

n8n Cloud is priced by workflow executions, not by node or step. If one incoming email starts one production workflow, 1,000 emails normally consume 1,000 executions. The model API remains a separate provider charge when you connect your own credential. If you already have unused execution capacity, the incremental platform cost may be zero. If the classifier forces a plan upgrade, the subscription difference can dwarf a 22-cent model bill.

One n8n execution can contain several items, but the Text Classifier documentation says each item is passed to the model. Grouping emails into one scheduled execution can reduce execution usage, while the token charge still follows the number and size of classified items.

Retries and output repair

A failed request, provider retry, or second-pass validation adds another call. n8n's Text Classifier has an Enable Auto-Fixing option that sends a schema parsing error back to the model and asks it to repair the output. That is useful, but it means the estimate above is no longer one call per email when repair is triggered. Measure the repair rate before assuming it is free.

Human review and mistakes

Suppose a reviewer spends 30 seconds on 10% of 1,000 emails. That is 50 minutes of work. Even at a modest internal hourly rate, review costs far more than the API tokens. The review may still be a good investment if it catches invoices routed to sales, urgent complaints marked as routine, or valuable leads sent to the wrong queue.

The correct metric is not only cost per 1,000 messages. Track cost per correctly classified message, manual-review rate, false-negative rate for urgent categories, and the operational cost of each error.

A simple n8n cost-measurement workflow

n8n already has a Text Classifier root node. It accepts an input prompt, category names and descriptions, optional multiple classes, an Other branch for unclear matches, a custom system prompt, and optional auto-fixing. For cost measurement, keep the workflow smaller than a full email router:

Email trigger or scheduled inbox fetch
  → Edit Fields: subject + cleaned body + minimal sender context
  → Text Classifier: named categories and descriptions
     ↳ connected small chat model
  → Log: provider, model, input tokens, output tokens, category
  → Route Other and selected high-risk categories to review

The separate AI email routing guide covers the downstream workflow. This article stays focused on cost measurement and model choice.

Use categories that are testable

Define categories by the action they trigger. For example: sales_lead, support, billing, urgent_complaint, and other_review. Give each category a short description and two or three edge rules. Do not ask the model for a paragraph when one label and a compact reason are enough.

Keep Allow Multiple Classes To Be True off when the business requires one owner. Use the documented Other branch instead of discarding an unclear item. Discarding may look efficient, but it can hide the messages most in need of review.

Log actual usage

Provider response fields differ, so inspect the output from your selected n8n chat-model node or the provider dashboard. Record input tokens, output tokens, model ID, retry count, latency, chosen category, and the reviewer's corrected category. Recalculate weekly:

observed cost per 1,000 =
  total provider cost / classified messages × 1,000

cost per correct classification =
  (provider + platform allocation + review labor) / correct decisions

Reduce cost without weakening the classifier

Use deterministic rules first. Known senders, exact aliases, and clear invoice mailbox rules do not need an LLM.
Remove quoted history. Classify the newest message plus only the minimum thread context required for meaning.
Strip HTML and signatures. Boilerplate consumes input and can introduce misleading terms.
Return a compact schema. A label, short reason, and review flag are enough for routing.
Do not add search or tools. An inbox classifier should not browse the web or call unrelated tools.
Measure auto-fixes and retries. A low base price can double if malformed output repeatedly triggers a second call.
Test a small model first. Upgrade only when a measured accuracy gap is worth the higher rate.

Choose the model by evaluation, not the price table

Start with at least 100 to 200 representative emails that a person has already labeled. Include short replies, forwarded threads, vague subjects, multilingual messages, promotions, complaints, and categories that are easy to confuse. This is an operational starting point, not a universal statistical guarantee.

Test	What to measure	Decision
Overall accuracy	Correct labels divided by all tested messages	Useful for orientation, not enough for rare urgent classes
Per-category recall	How many real messages in a category the model found	Prioritize billing, complaints, security, and other costly misses
Other or review rate	Share of messages that need a person	Convert the review share into minutes and labor cost
Correction stability	Whether prompt or model changes improve the same held-out set	Use a versioned test set before changing production

If two models are close, choose the simpler operational fit: supported credential, stable model ID, acceptable latency, data-processing requirements, and clear usage reporting. A few extra cents per 1,000 emails can be a sensible trade for fewer manual reviews.

Questions small teams usually ask

Can the classifier really cost less than one dollar?

Yes. A classification output is short, and small models have low token rates. The estimate stops being valid when you send long histories, extract attachments, ask for generated replies, use a larger model, or trigger repair and retry calls.

Does n8n include the model tokens?

Do not confuse n8n's AI Assistant credits with the third-party model used inside your production workflow. With your own OpenAI, Gemini, Anthropic, or Groq credential, the model provider meters the API usage. n8n Cloud separately counts workflow executions according to its current plan rules.

Should I use the cheapest model?

Use the cheapest model that passes your labeled evaluation and operational requirements. Price alone cannot tell you whether a model understands your category boundary, language mix, or unusual messages.

Should low-cost classifications send automatic replies?

Classification and outbound sending are separate risk decisions. Route first. If AI also drafts a customer-facing response, use a duplicate-safe human approval gate until measured evidence supports narrower exceptions.

Sources checked

n8n Cloud for small teams

Measure the classifier before you scale it

n8n Cloud is the simplest starting point if you want to connect an inbox, test categories, and log real usage without maintaining a server. Affiliate disclosure: I may earn a commission if you sign up through this link, at no extra cost to you.

<a href="https://n8n.partnerlinks.io/6mg2xng7s1m6" target="_blank" rel="sponsored noopener" class="btn b

Apollo Pricing: Plans, Credits, and Real Small-Team Cost

Sat, 29 Aug 2026 15:50:00 +0000

Direct answer: choose an Apollo plan by the paid action you must prove, not by the longest feature list. Free can be enough to define a market and inspect the interface. Basic adds a much larger recurring credit allowance and several plan-gated targeting and enrichment options. Professional becomes relevant when a team needs larger calling, reporting, testing, or workflow limits. Organization starts with a multi-seat annual commitment. The real monthly cost is seats plus consumed data actions, add-ons, any missing verification or sending tools, and the time spent rejecting stale or irrelevant records.

Partner disclosure. This guide includes a partner link to Apollo. If you create an account or purchase through it, I may earn a commission at no extra cost to you.

This page has one job: explain Apollo pricing for a solo operator or small B2B team. It does not replace the free-account and trial checklist, teach a complete CSV enrichment workflow, or provide a full product review. Prices, credits, and limits below were checked on August 27, 2026 and may change.

Apollo plan prices and included credits

Plan	Monthly billing screen	Annual billing screen	Credits shown	Small-team boundary
Free	US$0	US$0	75 per seat per month	Start the ICP and interface test. Confirm enrichment access in the actual workspace.
Basic	US$65 per seat per month	US$49 per seat per month, US$588 billed yearly	2,500 per month or 30,000 per year upfront	Larger credit allowance, advanced filters, and more enrichment options.
Professional	US$99 per seat per month	US$79 per seat per month, US$948 billed yearly	4,000 per month or 48,000 per year upfront	Higher workflow, calling, reporting, and testing allowances.
Organization	US$149 monthly equivalent, annual commitment	US$119 per seat per month, minimum three seats	6,000 per month or 72,000 per year upfront	Enterprise controls, larger limits, and a multi-seat commitment.

The annual prices and credit grants match Apollo's official pricing page. Monthly prices, add-on quotes, and plan details were also captured from the FloxoLab account on August 26, 2026. Taxes, promotions, workspace configuration, and rollout differences can change the checkout.

Annual billing snapshot captured on August 26, 2026: Basic US$49, Professional US$79, and Organization US$119 per seat per month, billed annually. Organization showed a three-seat minimum.

Monthly billing snapshot from the same account: Basic US$65 and Professional US$99 per seat per month. Organization still displayed annual billing only at US$149 per seat per month.

Trial-credit conflict: Apollo's public pricing FAQ currently says a trial includes 50 credits plus 5 mobile credits. A separate Apollo Help article says 100 free credits. The Professional trial activated in the FloxoLab account displayed 75 credits. Do not use a trial allowance to estimate normal paid capacity. Record the exact confirmation and Plan Overview screens for your workspace.

What the long feature list actually buys

The comparison screen groups many different jobs under one subscription. A checkmark only proves that a plan matrix listed the feature. It does not prove that the workflow is useful, available in every workspace, or free of additional credit and add-on costs.

Feature label	What it does	Why a small team may care
Contact & Account Data	Provides person and company records that other Apollo workflows use.	Useful as a discovery layer, but the records still need identity, role, recency, and relevance checks.
Waterfall Enrichment	Can try Apollo and external data sources in sequence when a requested field is missing.	May improve coverage, but the source path and variable credit cost must be recorded before scaling.
CSV, CRM & API Enrichment	Enriches an uploaded file, connected CRM, or custom application through different entry points.	Choose the route where the record already lives. More integration is not automatically better data.
Intent Topics & Filters	Lets the operator filter or prioritize accounts by selected intent topics.	Treat it as timing evidence for research, not proof that a company is ready to buy.
Domain & Mailbox Purchasing	Adds sending infrastructure inside the Apollo workflow.	The matrix showed separate credit costs, so these are not simply included because sequences are available.
Website Visitors and Form Enrichment	Identifies eligible site visitors and enriches inbound form records.	Relevant only when the site has enough qualified traffic and a privacy-reviewed follow-up process.
Inbound Routers and Meetings	Routes inbound records and schedules meetings under configured rules.	Useful after ownership, qualification, calendars, and fallback handling are defined.

The plan matrix separates basic filters from advanced filters and signals. It also showed one intent topic on Free, six on Basic and Professional, and twelve on Organization.

The data-enrichment category lists several entry points. This screenshot records the comparison claims, not a successful hands-on enrichment run.

Domain, mailbox, dialer, workflow, and meeting rows expose costs and limits that are easy to miss when comparing only seat prices.

Inbound and Advanced Dialer are separate US$149 team add-ons

The current checkout displayed Inbound and Advanced Dialer at US$149 per team per month each. The Inbound trial card showed visitor, domain tracking, form enrichment, form builder, and US contact identification limits. That full add-on allowance should not be confused with a smaller base-plan allowance that may appear elsewhere in the account.

Account checkout on August 26, 2026. Inbound was active as a fourteen-day trial; both add-ons were quoted at US$149 per team per month.

Free is a preparation plan, not a complete cost test

Free is useful when the team still needs to define its ICP, exclusions, role set, and required fields. It can also reveal whether the interface and basic search route make sense. It cannot prove a paid feature that the account does not expose. In the checked workspace, opening Data enrichment redirected to the upgrade screen because CSV enrichment was not included on Free.

Stay on Free while any of these are unresolved:

the target market is still described as everyone who might buy;
no one can name the required contact and company fields;
the team has no duplicate, suppression, or manual-review rule;
the paid feature has no pass or fail criterion;
the operator cannot estimate how many accepted records are needed per month.

Credits turn usage into the variable cost

The checked account screen separated several actions instead of presenting one universal credit price. That makes a record count alone misleading.

Apollo's account explanation shows why a credit is not a universal lead unit. Different actions consume different amounts.

Action shown	Draft credit rate	Cost-control question
Verified email	1 credit	How many net-new verified emails are actually required?
Phone number	8 credits	Does the sales process use calling enough to justify the higher rate?
Enrich data	1 to 8 credits	Which fields and data sources are selected, and what will a usable result cost?
AI Research	1 credit per run	Which decision improves because this research was run?
US Dialer	1 credit per 30 seconds	What connected-call volume belongs in the monthly estimate?

Do not multiply the plan allowance by one favorite rate and call that the number of leads. A workflow may use several credit categories on the same accepted record. It may also spend credits on a record that later fails the role, company, geography, or legal-use check.

Observed add-on credit prices

The active Professional checkout allowed the team to add shared monthly credits on top of the included 4,000. The following slider positions were captured on August 26, 2026. They are account evidence, not a promise that every region, plan, or future checkout will show the same quote.

Extra credits per month	Displayed monthly price	Displayed price per extra credit
400	US$10	US$0.025
1,000	US$25	US$0.025
2,500	US$50	US$0.020
10,000	US$180	US$0.018
250,000	US$3,750	US$0.015

One captured slider state: 10,000 extra shared credits for US$180 per month on top of the current plan allowance.

Apollo's plan-management documentation confirms that add-on credits are recurring, take effect immediately, are not prorated, and expire at the end of the billing cycle. Buying midway through a cycle can therefore make the effective cost higher than the slider alone suggests.

Calculate cost per usable record, not cost per reveal

Use a small worksheet before choosing a tier:

Set the number of active seats, not the number of people who might log in someday.
Estimate monthly candidate records for one target segment.
Split required actions into email, phone, enrichment, AI research, export, and dialer categories.
Apply the current account-specific credit rates and included allowances.
Estimate the percentage that will pass company, role, recency, and manual-verification checks.
Add every external tool and the operator time required to finish the workflow.

Working formula: real monthly cost = Apollo seats + add-on credits + required external tools + review labor. Cost per usable record = real monthly cost divided by the records that pass the complete acceptance check.

Three small-team cost scenarios

These examples use the checked monthly prices and credit rates. They are planning models, not recommended volumes. Taxes, external tools, and labor are shown separately because they vary by team.

Scenario	Credit actions	Apollo software cost	Usable-record check
Free preparation	40 verified emails = 40 credits	US$0	If 20 records pass review, software cost is US$0 per usable record. Research labor still counts.
Solo Basic test	300 emails + 25 phones + 100 AI runs = 600 credits	US$65 monthly seat, within the displayed 2,500-credit allowance	At 75 usable records, seat cost is about US$0.87 each before labor and other tools.
Basic with extra capacity	Up to 5,000 available credits after adding 2,500 to the displayed plan allowance	US$65 seat + observed US$50 add-on = US$115	At 250 usable records, platform cost is US$0.46 each before labor, tax, and required infrastructure.

The lowest unit price is not automatically the best purchase. A larger allowance saves money only when the team can use it before expiry and the additional records still pass company, role, recency, suppression, and lawful-use checks.

Compare Apollo with the separate stack it may replace

Apollo can combine several jobs, but a combined platform is cheaper only when the team uses those jobs and the output fits its market. Compare the complete workflow:

company and people discovery;
business email and phone finding;
verification and enrichment;
CSV or CRM handoff;
sequence and task execution;
mailbox, domain, deliverability, and suppression operations;
manual research and quality control.

Do not insert competitor prices until they are checked on the same date, currency, billing period, and usage assumption. The comparison should show which jobs disappear, which remain, and where another tool is still required.

Choose the tier by the first hard requirement

Situation	Likely starting route	Evidence required before paying
Solo operator def

Why can't we just use one storage technology for everything?

Sun, 05 Jul 2026 22:41:00 +0000

Hello, I'm Ganesh. I'm building git-lrc, an AI code reviewer that runs on every commit. It is free, unlimited, and source-available on Github. Star git-lrc on GitHub to help more developers discover the project. Do give it a try and share your feedback for improving the product.

Modern computers use multiple storage and memory technologies because no single medium can simultaneously provide massive capacity, ultra-low latency, high bandwidth, and low cost.

Every level of the memory hierarchy represents a tradeoff between these characteristics.

Why is there a memory hierarchy?

This is because each storage technology has its own advantages and disadvantages. We can't use one storage technology for all the purposes.

The main characteristics to consider are:

Capacity
Latency
Bandwidth
Cost

Each Storage technology is optimized for a specific purpose.

For example:
Imagine you have to travel from one place to another.
You can go with walking, bicycle, car, train, or plane.

Each mode of transport has its own advantages and disadvantages.

Another better example is think of using LLM model.

Depending on task complexity and size of data we can choose the model.

If we mess up with choosing model we will not get the desired output wheather cost will rise, or response will be very low quality.

Memory Hierarchy

Now I have listed down the memory hierarchy From top to bottom.

CPU Registers

This is the fastest storage available inside a processor.

Role:
Stores values currently being operated on by the CPU.

Characteristic	Value
Capacity	Extremely small (typically 32 × 64-bit registers per core)
Latency	~0.3 ns (about one CPU cycle)
Bandwidth	Highest in the system
Cost	Highest cost per bit (16 transistors per bit)

CPU Cache (L1, L2, L3)

A small, ultra-fast memory layer designed to keep frequently used data close to the processor.

Role: Reduces the need to access slower main memory.

Characteristic	Value
Capacity	Small (tens of MB total)
Latency	~7.5 ns (L3 cache)
Bandwidth	Extremely high
Cost	Very expensive (SRAM, 6 transistors per bit)

DRAM (Main Memory)

The working memory used by applications and operating systems.

Role: Holds active programs and data currently being used.

Characteristic	Value
Capacity	Typically 32 GB
Latency	~45 ns
Bandwidth	~48 GB/s
Cost	Higher than storage drives

GPU VRAM

Memory optimized for throughput rather than low latency.
Role: Feeds large amounts of data to thousands of GPU cores simultaneously.

Characteristic	Value
Capacity	~24 GB
Latency	~250 ns
Bandwidth	Extremely high
Cost	Expensive

NVMe SSD

High-speed solid-state storage connected through PCIe.
Role: Fast persistent storage for operating systems, applications, and files.

Characteristic	Value
Capacity	~2 TB
Read Latency	~80 μs
Write Latency	~500 μs
Bandwidth	~5 GB/s
Cost	~7 cents per GB

SATA SSD

An older solid-state storage technology using the SATA interface.

Role: Affordable solid-state storage with lower performance than NVMe.

Characteristic	Value
Capacity	Up to ~4 TB
Latency	~120 μs
Bandwidth	Lower than NVMe
Cost	Similar to NVMe SSDs

Hard Disk Drive (HDD)

Mechanical storage using spinning magnetic platters.
Role: Lowest-cost local storage for large datasets and archives.

Characteristic	Value
Capacity	~8 TB
Latency	~8.3–10 ms
Bandwidth	Very low compared to SSDs
Cost	~1–2 cents per GB

The Fundamental Tradeoff

As we move down the memory hierarchy:

Capacity increases.
Cost per GB decreases.
Latency increases.
Access speed decreases.
Distance from the processor increases.

By leveraging this trade-off, modern computers use a hierarchy of registers, cache, memory, SSDs, HDDs, and cloud storage instead of relying on a single storage technology.

Any feedback or contributors are welcome! It’s online, source-available, and ready for anyone to use.

⭐ Star git-lrc on GitHub

Credits

Here are the sources that inspired me to write an article on this topic.

Why Don’t Computers Just Use One Type of Memory?

7 Enterprise Infrastructure Tools That Eliminate Months of Engineering Work

Sun, 26 Apr 2026 03:44:00 +0000

Engineering teams lose thousands of hours annually building infrastructure that already exists as battle-tested platforms. Authentication systems, AI security layers, internal developer portals, and database management consume months of development cycles that could go toward shipping product features.

This guide covers seven enterprise infrastructure tools that eliminate this engineering toil. Each tool addresses a specific infrastructure bottleneck, with verified pricing and deployment timelines based on current market data.

Quick Comparison: All 7 Tools at a Glance

Tool	Category	Free Tier	Paid From	Best For
MojoAuth	Passwordless Auth	25K MAUs	$50/month	B2C/B2B apps needing SSO
Gopher MCP	AI/MCP Security	1K tool calls	$179/month	AI agent deployments
Port.io	Developer Portal	15 seats	Custom	Platform engineering
Pulumi	Infrastructure as Code	Individual	$0.0005/credit	Multi-cloud IaC
Retool	Internal Tools	5 users	$12/user/mo	Admin panels, dashboards
LaunchDarkly	Feature Flags	Developer tier	$10/conn/mo	Progressive rollouts
Neon	Serverless Postgres	100 CU-hours	$5/month	Dev/test databases

1. MojoAuth: Passwordless Authentication Platform

Website: mojoauth.com

What it does: MojoAuth provides a unified API for passwordless authentication methods including Passkeys, Magic Links, OTPs (Email, SMS, WhatsApp), and enterprise SSO. It replaces months of custom authentication development with production-ready infrastructure that integrates in hours.

Key Features

Multiple authentication methods: Passkeys/WebAuthn, Magic Links, Email OTP, SMS OTP, WhatsApp OTP, TOTP/HOTP, and Social Login through a single unified API.
Enterprise SSO integration: SAML 2.0 and OIDC support for connecting to corporate identity providers including Okta, Azure AD, and Google Workspace.
Multi-factor authentication: Layer additional security with SMS, TOTP, or push notifications without rebuilding authentication flows.
Developer SDKs: Production-ready libraries for JavaScript/Node.js, Python, Go, .NET, and React with comprehensive documentation.
Self-enrollment and recovery: Users manage their own passwordless credentials without requiring support intervention.

Pricing (March 2026)

Free Plan: Up to 25,000 Monthly Active Users. Includes Email OTP, Magic Links, Social Login, and basic features.
Business Pro: Starting at $50/month. Scales to 500,000 MAUs at approximately $1,700/month. Includes all authentication methods, enterprise SSO, custom branding, and priority support.
Enterprise: Custom pricing. Includes dedicated infrastructure, SLA guarantees, and compliance certifications.

Engineering Time Saved

Building a secure, production-ready authentication system typically requires 3 to 6 months of engineering time. This includes handling token management, rate limiting, account recovery flows, and staying current with evolving security threats. MojoAuth claims integration in under a day with up to 75% faster time-to-market compared to building in-house. The platform reports 99.9999% uptime SLA and claims to reduce support tickets related to authentication by 80%.

2. Gopher MCP: AI Agent Security Gateway

Website: gopher.security

What it does: Gopher MCP provides enterprise-grade security infrastructure for Model Context Protocol (MCP) deployments. As AI agents gain access to production databases, APIs, and internal tools, Gopher acts as a security layer that inspects every tool call and enforces granular access policies.

Key Features

4D Security Framework: Deep inspection and threat detection, context-aware access control, granular policy enforcement, and post-quantum end-to-end encryption.
MCP threat protection: Defends against tool poisoning, puppet attacks, prompt injection, and malicious external resource loading.
API schema to MCP server: Upload Swagger, Postman, or OpenAPI schemas to deploy secure MCP servers in minutes.
Post-quantum encryption: P2P connections use quantum-resistant cryptographic algorithms to protect against both current and future threats.
Compliance support: Built-in audit logging and policy controls for SOC 2, GDPR, and HIPAA requirements.

Pricing (March 2026)

Free Plan: 1 MCP server, 1,000 tool calls/month, 5GB data transfer, community support.
Starter: $199/month ($179/month annual). 3 MCP servers, 1M tool calls/month, 5TB data transfer, multi-environment support.
Business: $999/month ($899/month annual). 10 MCP servers, 5M tool calls/month, 25TB data transfer, advanced IAM, SSO, and compliance reports.
Enterprise: Custom pricing. Unlimited servers, dedicated infrastructure, phone support, and custom integrations.

Engineering Time Saved

The MCP authorization specification has evolved rapidly through 2025 with major revisions adding OAuth 2.1, Cross App Access protocols, and machine-to-machine authentication. Building spec-compliant MCP authentication in-house typically requires 6 to 12 weeks of dedicated development. Gopher claims deployment in under 30 minutes with the platform handling ongoing spec compliance updates automatically.

3. Port.io: Internal Developer Platform

Website: port.io

What it does: Port.io is an internal developer portal that provides a unified view of your software catalog, self-service infrastructure provisioning, and developer experience scorecards. It replaces the fragmented tooling that forces developers to context-switch between AWS console, Datadog, PagerDuty, and dozens of other platforms.

Key Features

Dynamic software catalog: Define custom blueprints that model your specific SDLC, with relationships between services, teams, and resources.
Self-service actions: Developers provision infrastructure, create environments, and perform day-2 operations without filing tickets.
Scorecards and standards: Define quality, maturity, and production readiness standards with automated compliance tracking.
Workflow automation: Long-running and asynchronous actions with TTL support and run logs visible to developers.
Integrations: Native connectors for Terraform, ArgoCD, Kubernetes, GitHub, Datadog, PagerDuty, and most enterprise tooling.

Pricing (March 2026)

Free Plan: 15 seats, 10,000 entities, 500 automation runs/month. No time limit.
Paid Plans: Custom pricing based on team size and feature requirements. Contact sales for quotes.
Enterprise: Flexible deployment options including dedicated tenancy and Private Link connectivity.

Engineering Time Saved

Port.io reports customers reducing environment provisioning time from 30 minutes to 30 seconds. According to their State of Internal Developer Portals report, 75% of developers lose 6 to 15 hours weekly to tool sprawl. For a team of 50 engineers, this translates to nearly $1 million in lost productivity annually. Port raised $100M in December 2025 at an $800M valuation, signaling strong enterprise adoption.

4. Pulumi: Infrastructure as Code in Real Languages

Website: pulumi.com

What it does: Pulumi lets engineering teams define cloud infrastructure using TypeScript, Python, Go, C#, Java, or YAML instead of domain-specific languages. This means loops, conditions, functions, unit tests, and package managers work natively with infrastructure code.

Key Features

Real programming languages: Write infrastructure in TypeScript, Python, Go, C#, Java, or YAML with full IDE support, type checking, and debugging.
Pulumi Neo AI agent: Generate infrastructure from natural language requirements, review PRs, and debug deployments with organizational context.
120+ cloud providers: Deploy to AWS, Azure, Google Cloud, Kubernetes, and over 120 providers through a unified model.
Policy as code: CrossGuard lets you define security, compliance, and cost controls that run during previews and updates.
Secrets management: Pulumi ESC provides a single interface for all secrets with connections to HashiCorp Vault, AWS Secrets Manager, and Azure Key Vault.

Pricing (March 2026)

Individual (Free): Unlimited updates, automatic state management, and core features. Free forever for individuals.
Team: $0.0005 per Pulumi Credit (one resource managed for one hour). Includes staging environments and release management.
Enterprise: Starting at $32,850/year via AWS Marketplace. SAML SSO, self-hosted options, and source control integrations.
Business Critical: Starting at $50,000/year via Azure Marketplace. 150 concurrent deployments and volume pricing.

Engineering Time Saved

Werner Enterprises reported reducing provisioning time from 3 days to 4 hours using Pulumi. The platform claims one organization expanded infrastructure contributors from 1-2 people to over 40 active engineers because developers could use familiar programming languages instead of learning specialized DSLs like HCL.

5. Retool: Internal Tools Development Platform

Website: retool.com

What it does: Retool combines drag-and-drop UI building with AI assistance to create internal tools such as admin panels, dashboards, and workflow apps. It connects to databases and APIs natively, eliminating the need to build CRUD interfaces from scratch.

Key Features

100+ pre-built components: Tables, forms, charts, maps, and calendars that connect to data sources with minimal configuration.
Native database connectors: Direct connections to PostgreSQL, MySQL, MongoDB, Snowflake, and 40+ other data sources with SSH tunneling and SSL support.
AI-powered app building: Generate apps from prompts, then customize with code or visual editing.
Workflows and automation: Build automated processes with scheduling, webhooks, and conditional logic.
Mobile apps: Build native iOS and Android apps alongside web applications from the same platform.

Pricing (March 2026)

Free: 5 users, unlimited apps, 500 workflow runs/month, 5GB storage.
Team: $12/month per standard user, $7/month per end user. Includes staging environments and version history.
Business: $65/month per standard user, $18/month per end user. Adds audit logging, unlimited environments, and embedded apps.
Enterprise: Custom pricing. SAML/OIDC SSO, self-hosted deployment, custom SLAs.

Engineering Time Saved

Retool claims teams can build internal tools 10x faster than traditional development. Used by Amazon, DoorDash, OpenAI, and Mercedes-Benz for mission-critical internal applications. One customer reported building a complete dashboard with 10 data sources in 24 hours. The platform is particularly effective for replacing Google Sheets or legacy internal tools that have become unstable.

6. LaunchDarkly: Feature Flag Management

Website: launchdarkly.com

What it does: LaunchDarkly provides feature flags that act as on/off switches for features in production. Teams can deploy code continuously while controlling exactly who sees new features through gradual rollouts, A/B tests, and instant kill switches.

Key Features

Feature flags: Toggle features on or off for specific user segments without redeploying code.
Progressive rollouts: Gradually expose features to increasing percentages of users while monitoring for issues.
Experimentation: Run A/B tests and funnel optimization experiments with statistical analysis.
Release Guardian: Monitor critical flag changes in release pipelines and automatically detect regressions.
Mobile Lifecycle Assistant: Deliver personalized mobile experiences and fix bugs without waiting for app store review cycles.

Pricing (March 2026)

Developer (Free): 1 project, 3 environments. Feature flagging for individual projects.
Foundation: $10/month per service connection (annual) or $12/month (monthly). Plus $10 per 1,000 client-side MAUs.
Enterprise: Custom pricing. Typically $20,000 to $120,000/year depending on scale and feature requirements.

Engineering Time Saved

LaunchDarkly enables instant rollbacks when issues are detected. Instead of a full deployment rollback taking 15 to 30 minutes, a feature flag toggle takes seconds. The platform is trusted by over 5,500 customers. Users report that the ability to decouple deployment from release fundamentally changes how teams ship software, allowing continuous integration without continuous exposure to users until features are ready.

7. Neon: Serverless PostgreSQL

Website: neon.tech

What it does: Neon provides serverless PostgreSQL that separates compute from storage, enabling instant database branching, scale-to-zero for idle databases, and automatic scaling based on actual usage. It eliminates the operational overhead of managing database infrastructure.

Key Features

Database branching: Create instant copy-on-write branches for development, testing, and CI/CD pipelines without duplicating data.
Scale to zero: Databases automatically suspend after inactivity (5 minutes default), eliminating charges for idle resources.
Autoscaling compute: Automatically scale between minimum and maximum compute units based on query load.
Point-in-time restore: Instant recovery to any point in time within the configured retention window.
Full PostgreSQL compatibility: Standard Postgres with all extensions, no proprietary modifications.

Pricing (March 2026)

Following Databricks acquisition in May 2025, Neon reduced compute costs 15 to 25% and storage from $1.75 to $0.35/GB-month.

Free: 100 CU-hours per project (doubled from 50 in October 2025), 0.5GB per project, up to 100 projects.
Launch: $5/month minimum. $0.106 per CU-hour, $0.35 per GB-month storage. Up to 16 vCPU and 64GB RAM.
Scale: $700/month base. $0.222 per CU-hour, includes 1,000 CU-hours and 500GB storage. Up to 56 vCPU and 224GB RAM.

Engineering Time Saved

Database provisioning traditionally requires DevOps involvement for each new environment. Neon branches create isolated database copies in seconds using copy-on-write technology. For a multi-tenant SaaS with 100 customer databases, one analysis showed Neon costing approximately $4,380/month compared to $9,985/month for equivalent Azure Database for PostgreSQL Flexible Server instances. The serverless architecture eliminates infrastructure management overhead entirely.

Frequently Asked Questions

How do I decide between building infrastructure in-house versus buying?

Build in-house when the component is a core differentiator for your product. Buy when the component is commodity infrastructure that every company in your industry needs. Authentication, feature flags, internal tooling, and database management are rarely competitive differentiators. The engineering time spent building these could go toward features that actually distinguish your product.

What is the typical ROI calculation for enterprise infrastructure tools?

Calculate the fully-loaded cost of engineering time for the alternative. If building a passwordless authentication system requires 4 months of senior engineer time at $200,000/year fully loaded, that represents approximately $67,000 in engineering cost alone. Compare this to annual subscription costs. Most infrastructure tools pay for themselves within the first quarter through avoided engineering time.

Do third-party infrastructure tools create vendor lock-in risks?

Lock-in risk varies by tool category. Authentication providers using standard protocols (SAML, OIDC) offer relatively easy migration paths. Feature flag platforms have more switching costs since flags become embedded throughout your codebase. Infrastructure as code tools like Pulumi provide escape hatches through code generation and multi-provider support. Evaluate migration complexity during vendor selection.

How do these tools affect compliance requirements like SOC 2 and HIPAA?

Enterprise infrastructure tools often simplify compliance by providing pre-built audit logging, access controls, and encryption that would otherwise require custom implementation. When you build in-house, your team bears full responsibility for security and compliance of that code. With certified vendors, you inherit their compliance posture and can reference their certifications in your own security assessments.

What is the best way to evaluate multiple tools in the same category?

Amazon SNS for Cost Reduction and Message Delivery Assurance in Startups

Tue, 14 Apr 2026 20:48:42 +0000

Most startups need to send notifications — order confirmations, alerts, password resets — but don't want to run their own messaging infrastructure. Amazon Simple Notification Service (SNS) solves this: a fully managed pub/sub service where you pay per message, not per server.

Here's what makes it worth evaluating, where it falls short, and how it compares to alternatives.

The startup messaging problem

Early-stage teams face a specific tension: they need reliable message delivery across email, SMS, and push notifications, but can't justify the cost or operational overhead of self-hosted messaging systems. The requirements typically include:

Pay-as-you-go pricing — no monthly minimums, no long-term contracts
Multi-channel delivery — email, SMS, mobile push, HTTP webhooks, Lambda triggers
Automatic retries — temporary failures shouldn't lose messages
Global reach — users in multiple regions need low-latency delivery

SNS addresses all four. But so do other services, which is why the trade-offs matter.

SNS uses a pay-per-message model. The first million SNS API requests per month are free. After that, it's $0.50 per million requests. SMS and email have separate per-message pricing that varies by destination country.

For a startup sending 100K push notifications and 10K emails per month, the SNS cost is effectively zero (within the free tier). Compare that to a dedicated email service like SendGrid or Mailgun, which typically start at $15–20/month for similar volumes.

The catch: SNS doesn't do rich email templates, drip campaigns, or analytics. It's a delivery pipe, not a marketing platform.

Three mechanisms prevent message loss:

Retry logic. When a delivery attempt fails (subscriber endpoint is down, network timeout), SNS automatically retries with exponential backoff. The retry policy is configurable per delivery protocol.

Dead letter queues. Messages that exhaust all retry attempts are routed to an SQS dead letter queue instead of being silently dropped. You can inspect these later, replay them, or trigger alerts.

Cross-AZ replication. Messages are replicated across multiple availability zones within a region before SNS acknowledges the publish request. This protects against hardware failures in a single data center.

This is the decision most teams get wrong. All three are AWS messaging services, but they solve different problems:

Service	Pattern	Best for
SNS	Pub/sub (fan-out)	Broadcasting one event to many subscribers
SQS	Point-to-point queue	Decoupling a producer from a single consumer
EventBridge	Event bus with rules	Routing events to different targets based on content

The common pattern is SNS + SQS together: SNS fans out an event to multiple SQS queues, each consumed by a different microservice. This gives you both broadcast and buffering.

If you only need one consumer, skip SNS and use SQS directly. If you need content-based routing (e.g., "send order events to the billing service, send inventory events to the warehouse service"), EventBridge is the better choice.

Real-world usage

Netflix uses SNS to send push notifications about new content releases. When a new season drops, SNS fans out the notification to millions of subscriber endpoints simultaneously.

Amazon itself uses SNS for order lifecycle notifications — placed, shipped, delivered — routing events to email, SMS, and mobile push depending on customer preferences.

Walmart uses SNS for order fulfillment and in-store pickup notifications, integrating with their logistics systems to trigger real-time updates.

No rich content. SNS messages are plain text (or JSON for application-to-application). If you need HTML email templates, open/click tracking, or A/B testing, you need SES or a third-party email service.
No guaranteed ordering. Standard SNS topics don't guarantee message order. FIFO topics do, but they're limited to 300 messages/second per topic.
Vendor lock-in. SNS integrates deeply with AWS services (Lambda, SQS, CloudWatch). Migrating to a different cloud later means rewriting all your pub/sub logic.
SMS costs add up. International SMS delivery can be expensive ($0.02–0.15 per message depending on country), and you need to manage opt-in compliance yourself.

When to pick something else

Transactional email with templates: Use Amazon SES or SendGrid
Marketing automation (drip campaigns, segmentation): Use Brevo, Mailchimp, or Customer.io
Real-time chat or presence: Use WebSockets or a service like Ably/Pusher
Cross-cloud pub/sub: Use Google Cloud Pub/Sub, Confluent Kafka, or NATS

Getting started

If you decide SNS fits, the setup is straightforward:

Create an SNS topic in the AWS Console or via CloudFormation/CDK
Add subscribers (email, SQS queue, Lambda function, HTTP endpoint)
Publish messages via the AWS SDK from your application code
Configure a dead letter queue to catch failed deliveries
Set up CloudWatch alarms on NumberOfNotificationsFailed

The AWS SNS documentation covers each step with working examples in Python, Node.js, and Java.