Frequently asked questions

You hire a specific person — not a rotating pool. We match you with a vetted professional based on your stack, domain, and working style, you conduct a technical interview before committing, and the person is assigned exclusively to your project for the duration of the engagement. They work within your timezone overlap window, join your team's communication tools (Slack, Teams), and participate in your sprint ceremonies. We handle HR, payroll, equipment, and benefits on our side. You direct the work.

Yes — always. We present 1–2 matched candidates with their CV, portfolio, and a summary of why we think they fit your requirements. You interview them before any engagement begins. If the first candidates are not right, we keep searching at no extra cost until you find someone you want to work with.

Within the first 30 days, if the placement is not working — for any reason — we replace them at no extra cost. After the first 30 days, we require 2 weeks notice to transition to a replacement so there is no knowledge gap. Replacement sourcing is included in your engagement at no additional charge.

Pakistan Standard Time (PKT) is UTC+5. This creates a 3–5 hour overlap with UK/EU mornings, a 4–6 hour overlap with Middle East business hours, and an async-friendly relationship with US/Canada (with a 2–4 hour overlap possible with early starts). We set timezone expectations upfront and ensure a minimum 3-hour synchronous overlap per day with your team.

Not necessarily. A tech lead who is good at stakeholder communication and sprint management can often absorb the PM role for smaller teams. Where a dedicated PM adds value is when stakeholder reporting, scope change management, and cross-team coordination are consuming your tech lead's time that should be spent on architecture and code.

A data scientist focuses on statistical modelling, experimentation, and deriving insights from data — they live in notebooks and produce research outputs. An AI engineer builds production systems that use AI models — they write production-grade code, integrate LLM APIs, build reliable pipelines, and deploy to real environments. Most businesses building AI-powered products need an AI engineer, not a data scientist.

A data analyst works with existing data to answer business questions — building dashboards, writing SQL queries, producing reports, and identifying trends. A data scientist builds predictive models and runs experiments. Most SMBs need a data analyst. Data science is valuable when you have a specific prediction problem (churn, demand forecasting) with enough data to train a model reliably.

Use dedicated staffing when you have ongoing, evolving work that benefits from someone building deep context in your codebase, brand, or data. Use project-based engagements when you have a defined scope with a clear start and end — a new feature, a data analysis, a brand identity. Many clients start with a project and move to dedicated staffing once they see the quality.

Technical fixes (crawl errors, Core Web Vitals) often show indexation improvements within 4–8 weeks. Ranking gains for competitive keywords take 3–6 months of consistent on-page and link work. Traffic that compounds meaningfully typically builds from month 4 onward. Any agency promising first-page rankings in 30 days is misrepresenting how search engines work.

No — and you should be cautious of anyone who does. Search rankings are determined by Google's algorithm, which weighs hundreds of factors including the competitiveness of your niche, domain age and authority, content quality, and technical health. We guarantee the quality and execution of our work — audits, on-page optimisation, content production, and link acquisition — and we measure progress transparently.

Paid ads give you traffic immediately but stop the moment you stop paying. SEO builds an asset that compounds over time. For most businesses, paid ads are the right starting point to validate offers and generate early revenue — then SEO takes over as the long-term channel. We often recommend running both in parallel, with ads informing which keywords and landing pages to invest in for SEO.

Our technical SEO audit covers: crawlability (robots.txt, sitemap, redirect chains), indexation health (canonicals, noindex tags, duplicate content), Core Web Vitals (LCP, CLS, FID/INP), structured data implementation, internal linking structure, page speed analysis, mobile usability, and schema markup opportunities. We deliver a prioritised issue list with clear remediation steps.

We have active writing experience across legal, healthcare, ecommerce, education, real estate, logistics, and construction. For technical or regulated industries, we conduct a subject-matter interview with you or a relevant team member before writing begins, and we send drafts for factual review before considering them complete.

Yes. Every article includes: a target keyword mapped to the correct search intent, proper H1/H2/H3 hierarchy, meta title and description, internal linking recommendations, and structured formatting (lists, tables, callouts) designed to increase dwell time and featured snippet eligibility.

Up to two rounds of revisions are included per deliverable for content and copywriting projects. A revision round means structured feedback from you, not line-by-line edits in real time. For design projects, we include two revision rounds per concept stage (concept → refinement → final).

Yes, if you provide access. We have active experience publishing to WordPress, Webflow CMS, Contentful, Sanity, and Ghost. We follow your style guide for formatting, tags, categories, and featured images.

Content writing informs — it produces articles, guides, and educational material that builds authority over time. Copywriting converts — it writes the words on a homepage, landing page, or ad that move a specific reader from interest to action. Most businesses need both: content writing for organic traffic and authority, copywriting for the pages where conversion actually happens.

Yes. We start every engagement with a brand voice exercise — reviewing existing copy you like, existing copy you do not like, competitor examples, and a short questionnaire on tone attributes. We produce a concise tone-of-voice reference that every writer on the project follows, and it is delivered as part of your project documentation.

Yes — you receive full ownership of all source files (Figma, .ai, .psd, .indd) on completion of the final payment. We do not retain ownership or charge ongoing licensing fees for designs we produce for you. Font licensing is a separate matter — if commercial fonts are used, we will flag the licensing requirement and associated cost upfront.

We present 2–3 distinct directions at the concept stage — each with a rationale explaining the strategic thinking behind the visual choice. After concept approval, we refine the chosen direction through 2 revision rounds. We do not present 10 options and let you vote — more options produce more compromise, not more clarity.

Yes. Send us your existing brand guidelines (or source files if guidelines do not exist) and we will design within your established system. If your brand system has gaps or inconsistencies, we will flag them and optionally address them — but we will not change your brand identity without explicit instruction.

For brand and print: SVG, PDF (print-ready), PNG (transparent, @1x and @2x), EPS. For web/digital: SVG, PNG, WebP. For video: MP4 (H.264 and H.265), MOV. For presentations: PDF and native format (Figma, PowerPoint, or Keynote depending on the project). Source files are always included.

We work with whatever you have — phone footage, DSLR footage, screen recordings, or presentation slides. Higher quality input produces better output, but we can colour grade and enhance footage from almost any source. If you need help with a shoot plan, we can advise on equipment, framing, and lighting for self-shot content.

Yes. We use royalty-free music from Artlist, Epidemic Sound, or Musicbed (subscription cost passed through at cost) and stock footage from Pexels, Pixabay (free), or Shutterstock (licensed, cost passed through). We will confirm any licensing costs with you before using paid assets.

Yes — included in all video edits. We generate accurate captions (auto-generated then manually corrected), style them to your brand, and deliver both a burned-in version (captions embedded in the video) and a separate .SRT file for platform upload.

It depends on your audience. B2B: LinkedIn is non-negotiable, with X and YouTube as secondary. B2C / consumer: Instagram and TikTok for visual/short-form, Facebook for local community and older demographics. We recommend starting with 2 platforms you can execute consistently rather than spreading across 5 at low quality.

Our starter package covers 16 posts per month across 2 platforms (8 posts each). Our growth package covers 24 posts across 3 platforms. Frequency is balanced against quality — we do not pad calendars with low-effort content to hit a number.

Yes — as an add-on to organic management or as a standalone service. We manage Meta Ads (Facebook and Instagram), LinkedIn Ads, and TikTok Ads. Ad spend is billed separately directly to your ad account; we charge a management fee on top. Minimum recommended ad spend is USD 1,000/month for Meta and USD 2,000/month for LinkedIn.

Organic social takes time. Follower growth and engagement improvement are typically visible from month 2. Brand awareness compounds over 6–12 months of consistent posting. Paid social (ads) can produce traffic and leads within days. We set clear expectations at the start of each engagement and track the metrics that matter for your business goal.

We connect to any source with an API or export: PostgreSQL, MySQL, BigQuery, Supabase, Google Sheets, Xero, Salesforce, HubSpot, Shopify, WooCommerce, Google Analytics 4, Facebook Ads, and more. If it has an API or can export to CSV, we can work with it.

Not necessarily. For early-stage analysis, we can query your operational databases directly or work from exports. A data warehouse (BigQuery, Snowflake, Redshift) becomes worthwhile when: (a) your reporting queries are slowing down production databases, (b) you need to join data from 3+ different sources routinely, or (c) your data volume exceeds what Google Sheets can handle. We will advise based on your actual situation.

Accuracy depends on data quality, data volume, and the predictability of what you are trying to forecast. We always deliver accuracy metrics alongside the model so you understand what you are working with — and we build human review workflows into any process where model errors have serious consequences. We do not deploy models we cannot measure.

RAG (Retrieval-Augmented Generation) is a pattern where an LLM answers questions by first retrieving relevant content from your own documents or database, then generating a response grounded in that content — rather than relying on its training data alone. You need RAG if you want the AI to answer questions about your specific knowledge base (contracts, manuals, product catalogue, internal policies) accurately and without hallucination.

Accuracy depends on prompt engineering quality, retrieval precision (for RAG), and the inherent complexity of the task. We build evaluation frameworks that measure accuracy systematically — not just qualitatively. Every LLM feature ships with a defined accuracy baseline, and we monitor for drift in production. Features that cannot meet accuracy requirements that matter for your use case are flagged before deployment, not after.

Running costs depend on model choice, token volume, and caching strategy. GPT-4o at USD 0.0025/1K input tokens and USD 0.01/1K output tokens is typical for OpenAI. Claude 3.5 Sonnet is comparable. We configure token cost monitoring and budget alerts from day one, and design prompts to minimise token usage without sacrificing accuracy. For most SMB use cases, monthly API costs run USD 50–500.

OpenAI's API (as opposed to ChatGPT) does not train on your data by default. Anthropic has the same policy. However, we still recommend: (a) not sending PII or sensitive identifiers in prompts — use anonymised IDs, (b) for regulated industries (healthcare, legal), using Azure OpenAI or AWS Bedrock for data residency guarantees, (c) reviewing the API data processing agreements against your compliance requirements. We advise on this as part of every LLM integration scoping.

AI automation is reliable for: document classification, data extraction from structured or semi-structured documents (invoices, forms, contracts), email and ticket triage and routing, report generation from structured data, and content summarisation. It is less reliable for: tasks requiring judgment about edge cases, anything requiring real-world context the model does not have, and tasks where errors have serious legal or financial consequences without a human review step.

For well-scoped automation tasks with clear inputs and outputs, accuracy of 85–95% is typical on diverse real-world data. We always build a human review queue for outputs below a confidence threshold — so the automation handles the easy cases and surfaces the hard ones for human decision-making. This means your team does less work, not that the AI replaces human judgment for complex cases.

Every automation we build includes: (a) confidence scoring on each output, (b) a human review queue for low-confidence cases, (c) an audit log of every AI decision, and (d) an override mechanism so a human can correct any AI output. Mistakes in production are used to improve the system — they are not surprises if you have monitoring in place.

Zapier is rule-based automation: if X happens, do Y. It cannot understand meaning, classify content, or make judgment calls. AI automation handles work that requires interpretation — reading a contract and extracting key terms, triaging a support email by issue type, or classifying an invoice against a chart of accounts. If your process needs intelligence, not just a trigger and action, that is where AI automation adds value Zapier cannot.

Square and Lightspeed are excellent for standard retail or hospitality. A custom POS is worth the investment when: (a) your service pricing or workflow is non-standard enough that off-the-shelf forces workarounds, (b) per-terminal or transaction fees at your volume exceed the build cost within 12–18 months, (c) you need deep integration with a custom inventory, booking, or loyalty system. We help you model this before recommending either path.

Yes. We build POS systems as Progressive Web Applications with Service Workers and IndexedDB — meaning they can process transactions offline and sync to the server when connectivity is restored. We test this explicitly with simulated outage scenarios during development.

We integrate with Stripe Terminal (BBPOS WisePOS E, Stripe Reader M2) and Square hardware (Square Terminal, Square Reader). Terminal availability varies by country — we confirm hardware availability for your location during scoping. Stripe Terminal supports Australia, US, UK, Canada, Ireland, New Zealand, Singapore, and more.

Tableau and Power BI are powerful for self-service analytics by people who know how to use them. A custom dashboard is better when: (a) your stakeholders need a simple, opinionated view of specific metrics — not a full BI tool, (b) you need to embed the dashboard inside an existing web application, (c) you want to avoid per-user licensing fees at scale, or (d) the data model Tableau produces is not flexible enough for your reporting structure.

Any source with a query interface: PostgreSQL, MySQL, BigQuery, Supabase, MongoDB, Salesforce, HubSpot, Google Sheets, Xero, Shopify, and any REST API. We design the data layer so the dashboard is not tightly coupled to one data source — adding new sources later is straightforward.

We support real-time (WebSocket or polling every 30–60 seconds), scheduled (every 15 minutes, hourly, daily), and on-demand refresh depending on your data source and how fresh you need the data. For most operational dashboards, hourly or daily is sufficient and avoids excessive database load. Real-time is available for transactional use cases.

Probably not yet. Kubernetes adds significant operational complexity and is warranted when you need: container orchestration across many services at scale, sophisticated deployment strategies (blue/green, canary), or fine-grained resource management across a large microservice fleet. For most applications under USD 5M ARR, managed container services (ECS, Cloud Run, Railway) give you 90% of the benefit at 20% of the complexity.

AWS (primary), GCP, Azure, DigitalOcean, Hetzner, Linode, Vultr, and Cloudflare. For most projects, we recommend AWS for its depth of managed services, or Hetzner for cost-efficient European hosting. We choose based on your existing setup, compliance requirements, and budget.

Yes — especially for a small team. The smaller your team, the more damage an unstable deployment process causes. CI/CD removes the mental overhead of "is this safe to deploy?" and lets developers merge with confidence. A basic GitHub Actions pipeline that runs tests and deploys to staging takes 1–2 days to set up and pays for itself immediately.

Managed hosting (Vercel, Railway, Render) is the right answer for most web apps — it handles infrastructure so you can focus on code. DevOps engineering adds value when: (a) you need custom networking, VPC configuration, or compliance controls managed hosting cannot provide, (b) your application has complex multi-service orchestration needs, (c) infrastructure costs at scale justify self-managing compute. We help you make this call based on your actual requirements.

Linux VPS and dedicated servers on any cloud or bare-metal provider: AWS EC2, DigitalOcean Droplets, Hetzner Cloud/Dedicated, Linode/Akamai, Vultr, and physical servers. We manage Ubuntu 20.04/22.04, Debian, RHEL/CentOS, and AlmaLinux. We do not manage Windows servers.

Our standard retainer covers business hours support (Monday–Friday, 9am–6pm PKT) with on-call response for critical outages during extended hours. We define "critical" upfront — typically service-down or data-loss risk scenarios. For mission-critical applications requiring true 24/7 on-call coverage, we structure the engagement with an appropriate on-call rate.

Managed hosting (AWS RDS, Heroku, Render) handles infrastructure at the platform level — you do not touch the OS. Server management means we administer the underlying Linux server — OS patches, security hardening, Nginx config, database maintenance, and monitoring. You need server management when you are running workloads that require OS-level control: custom daemon services, specific software versions, port-level firewall rules, or compliance configurations managed hosting does not expose.

Included every month: security patch review and application, uptime monitoring with alerting, monthly health report (CPU/memory/disk/network trends), database maintenance (vacuuming, slow query log review, backup verification), SSL certificate renewal management, log rotation configuration, and response to alerts during covered hours. Excluded: major infrastructure changes, new software installation (billed separately), and application-level code debugging.