How Much Does AI Data Analysis Software Cost to Build

How Much Does AI Data Analysis Software Cost to Build

Gartner has estimated that poor data quality costs the average organization $12.9 million a year, and its research found that 59 percent of organizations do not measure data quality at all. Many teams are making decisions from numbers nobody has checked.

AI data analysis software is meant to close that gap. A business connects its data once, and the software cleans it, spots patterns, forecasts demand and answers typed questions.

What it costs to build varies enormously. A simple tool that reads uploaded files can start near $20,000. A platform serving a large company across many systems can pass $350,000. The cost to build AI data analysis software depends on five areas:

•         Software development: screens, logins, dashboards and the server code behind them.

•         AI model work: connecting a ready-made language model or training your own.

•         Data infrastructure: pipelines that pull data in, clean it and store it.

•         Security: protecting business data and meeting privacy laws.

•         Maintenance: fixes, updates and accuracy checks after launch.

Which type of AI data analysis software are you building?

The category you pick sets the scope, and two AI analytics tools can differ in price fivefold.

AI-powered business intelligence software

Familiar dashboards with AI added: automated reports and written summaries. It is the cheapest category because the data is usually structured.

Predictive analytics software

It forecasts demand, estimates which customers may leave and scores risk. Models must be trained on past data and tested against real outcomes, which adds time.

Natural language data analysis software

Users type a question and get a chart. The software turns it into a database query (called text-to-SQL), runs it and explains the result. It looks simple and is hard to make dependable.

Data processing and analysis platforms

These focus on the data itself: cleaning records, sorting them, flagging odd values (anomalies) and finding patterns. Most of the money goes to data engineering.

Enterprise AI analytics platforms

These combine many data sources, advanced prediction, role-based access (each person sees only what they should) and real-time analytics, on top of everything above.

As a rule, AI data analytics software development gets more expensive as you move down this list.

What does an AI data analysis product cost to build in 2027?

These are indicative planning ranges, not fixed quotes. Final figures move with AI features, data systems, integrations, security needs, team location and scope.

Software type

Estimated development cost

Basic AI data analysis software

$20,000 to $40,000

Mid-level AI analytics platform

$40,000 to $90,000

Advanced AI data analysis software

$90,000 to $180,000

Enterprise AI analytics platform

$180,000 to $350,000+

What each range usually covers:

•         Basic: data uploads, automated analysis, standard charts and simple AI-written insights.

•         Mid-level: several data sources, predictive analytics, richer dashboards and natural language questions.

•         Advanced: custom AI models, real-time processing, advanced forecasting and recommendations.

•         Enterprise: large-scale data processing, complex integrations, advanced security and many users.

The jump between tiers comes from what sits behind the screens. A basic tool reads an uploaded file. An enterprise platform reads from eight systems that disagree with each other and still gives every user the same answer.

Treat any AI data analysis software development cost figure, including these, as a planning range until a developer has seen your actual data.

PRO TIP  Ask every vendor for a written list of what the quote excludes. Data cleaning, testing on real data and the first months of support are the items most often left out.

Where does the development budget actually go?

A $65,000 quote looks like one number. It is really seven jobs. This table shows how a mid-level platform budget is commonly divided. Your split will differ, but AI work, data engineering and backend usually take the largest shares.

Area

Share

Amount on $65,000

UI/UX design

8%

$5,200

Frontend development

17%

$11,050

Backend development

18%

$11,700

AI and machine learning

22%

$14,300

Data engineering

18%

$11,700

Third-party integrations

8%

$5,200

Testing and QA

9%

$5,850

Total

100%

$65,000

UI/UX design

Dashboards, chart layouts, the question box and phone layouts.

Frontend development

Dashboards, upload screens, report pages, filters and user management. Drawing huge numbers of data points can freeze a browser.

Backend development

Data processing, APIs (the connections between parts of the software), logins and report building. It also holds permission rules, so a sales rep never sees payroll.

AI and machine learning

Language model connections, prediction models, text-to-SQL, anomaly detection and recommendations. The AI data analysis tool development cost climbs when accuracy must be proven, because proof needs test data and many rounds of checking.

Data engineering

Collecting, cleaning and validating data, then moving it through pipelines (automated routes between systems) into a standard shape. It is the least visible and most underestimated line.

Integrations

Language model APIs, cloud warehouses, BI tools, payment gateways and notification services each have their own rules and limits.

Testing and QA

Beyond normal testing, teams compare outputs with known answers, then test speed under load and security.

Which AI features push the price up the most?

AI is not one expense. Each feature has its own build and running cost.

AI feature

What increases the effort

Relative cost

Natural language queries

NLP, query interpretation and text-to-SQL

Medium to high

Predictive analytics

Model development, training and validation

High

Automated data cleaning

Data quality rules and transformation logic

Medium

Anomaly detection

Statistical models and detection algorithms

Medium

AI-generated reports

Language model use and context

Low to medium

Recommendation engines

Data analysis and personalization algorithms

High

Real-time analytics

Streaming infrastructure and low-latency processing

High

Automated insights

Pattern detection and written explanations

Medium

Connecting to a language model takes days. Making it dependable takes far longer. The model does not know what "active customer" means in your company: someone who logged in, or someone who paid? Until it knows, it answers confidently and wrongly.

Take "what were sales last quarter?" The software must find the right tables, decide whether quarter means calendar or fiscal and handle returns. Teams build a layer of business definitions and test questions with known answers. That layer takes a large share of the cost to build AI data analysis software with natural language features.

Forecasting has its own trap: a model trained on past sales reads a stockout as weak demand. Real-time features add streaming infrastructure that never pauses, and each second of speed raises the bill. Hourly refresh is often enough.

KEY TAKEAWAY  Price features one at a time. Natural language queries with basic charts cost far less than a product that adds forecasting and live data, though both are sold as AI analytics.

A worked example: a retail analytics platform

Take a chain of 40 stores as an illustration. The platform collects sales and stock counts hourly, finds buying trends, forecasts demand per product and store, suggests reorders and answers staff questions.

Feature

Build cost

Main running cost

Sales and stock data pipeline

$9,000 to $14,000

Pipeline runs, storage

Dashboards and trend detection

$15,000 to $23,000

Hosting, computing

Demand forecasting

$14,000 to $22,000

Monthly retraining

Reorder recommendations

$8,000 to $13,000

Model upkeep

Natural language questions

$9,000 to $15,000

AI usage per question

Security, roles and testing

$8,000 to $13,000

Audits, monitoring

Total

$63,000 to $100,000

 

Forecasting costs most because its exceptions are costly. Zero sales on an out-of-stock product is unknown demand, not zero demand. A holiday closure is not a sales drop. A promotion spike should not repeat in next month's forecast.

Signals can conflict too. Searches for a product may rise while store sales fall, perhaps because shoppers buy online. The system should show both numbers and leave the decision to a person.

AI model development versus existing AI API integration

Three routes add AI, at very different price points.

•         Integrating existing AI APIs: you send data to a provider's model and get answers back. Upfront cost is lowest and launch is fastest, but you depend on the provider's prices and model changes.

•         Fine-tuning an existing model: you train it further on your own examples. That means data preparation, training rounds and hosting costs, often billed even when nobody is using it.

•         Developing custom AI/ML models: you design and train a model for a task nothing ready-made handles. That means data collection, training hardware (GPUs), deployment and retraining.

 

Factor

Existing API

Fine-tuned model

Custom model

Initial investment

Lowest

Medium

Highest

Recurring expenses

Per-use fees

Hosting and refresh

Hosting, GPUs, retraining

Customization

Prompts and settings

Language and format

Full control

Scalability

Provider handles it, within limits

Own capacity planning

Own capacity planning

Best for

MVPs and summaries

Specialized vocabulary

Unique predictions

Most products should start with an API. Custom AI data analysis software development makes sense when the prediction task is unusual, data cannot leave your servers, or volume is high enough that per-use fees exceed the cost of running your own model.

How data sources and data volume change the price

Clean, structured data from one database is a different job from millions of records across several enterprise systems. That gap is a major reason the AI data analysis software development cost varies so much.

Number of data sources

Each source needs its own connector. Excel and CSV files are easiest, then SQL and NoSQL databases. CRM and ERP platforms, cloud storage and third-party APIs cost more because each has its own login method and limits. A new source often adds one to three weeks.

Data volume

Thousands of rows fit in one database. Millions need a warehouse and tuned queries. Hundreds of millions need distributed processing, splitting work across many machines. Streaming data, such as clicks arriving every second, needs infrastructure that never stops.

Data quality and preparation

Real data has blank fields, duplicate customers, dates written three ways and products with two names. Each fix needs a business decision.

Two situations need their own budget. In a data gap, a store's sales feed fails for two days and the software reads two days of zero sales as collapsing demand. Good systems mark missing feeds as unknown. In conflicting signals, the CRM says a customer is active while billing says the account is closed. Someone must decide which system wins and write the rule down.

Structured and unstructured data

Analyzing an Excel sales report is simple because every number sits in a labeled column. Pulling insight from PDFs, reviews, emails and contracts is a different project. The software must read the text (sometimes from scanned images), decide what matters and turn it into something countable. That adds AI steps, testing and usage fees, and can move a project up a full pricing tier.

Integration complexity

Older systems often lack modern APIs, so developers work through exports or custom scripts. Real-time sync must handle late, duplicate and out-of-order records.

Technology stack and what it means for cost

Technology layer

Possible technologies

Frontend

React.js, Next.js, Angular

Backend

Python, Django, FastAPI, Node.js

Database

PostgreSQL, MongoDB, MySQL

Data processing

Apache Spark, Pandas, Apache Kafka

AI/ML

PyTorch, TensorFlow, Scikit-learn

LLM integration

OpenAI API, Anthropic API, open-source LLMs

Data visualization

D3.js, Chart.js, Plotly

Cloud infrastructure

AWS, Microsoft Azure, Google Cloud

Open-source tools such as Python, PostgreSQL and Apache Spark carry no license fee, but someone must host, update and secure them. Managed services cost more monthly and remove that work. Some charting libraries and connectors charge per user, and those fees grow with your customer base.

Compatibility matters more than fashion: a team strong in Python builds a Python backend faster. Cloud choice shapes the bill for years, since switching is costly, and widely used tools are easier to staff.

Monthly costs after launch

The software runs queries, calls AI models and stores data all month.

AI API and model inference

Providers charge by tokens, small chunks of text of roughly three-quarters of a word each. You pay for tokens sent in (the question, table descriptions, sample rows) and tokens returned (the answer). Costs rise with query count, model size and repeat checks.

The sample below uses Claude Sonnet 5 pricing from Anthropic's published API price list as of October 2026: $2 per million input tokens and $10 per million output tokens. A typical question sends about 6,000 tokens and gets back about 800, about two cents. A checking pass raises the average to three cents. A document-heavy question, with 40,000 tokens in and 1,500 out, costs about 9.5 cents.

Questions per month

Standard ($0.02 each)

With a check ($0.03 each)

Document-heavy ($0.095 each)

100

$2

$3

$9.50

1,000

$20

$30

$95

10,000

$200

$300

$950

At 10,000 questions the bill is small. At scale it is not: 100,000 users asking 20 questions each is two million questions, or $60,000 a month at the checked rate. Caching repeated answers and sending easy questions to a smaller model are the usual fixes.

Hosting, data processing and services

Cloud hosting covers servers, databases, storage and bandwidth, and idle capacity still bills. Data processing covers loading, transforming and querying data. Many cloud warehouses charge by the data each query scans, so one badly written query can cost more than a day of normal use.

Maintenance and technical support

Bug fixes, model updates, database tuning and security patches. A common planning rule is 15 to 20 percent of the build cost each year. Custom AI data analysis software development adds model monitoring, because accuracy drifts as data changes.

Monthly cost item

Small platform

Medium platform

Enterprise platform

AI API and inference

$200 to $800

$1,500 to $6,000

$12,000 to $50,000+

Cloud hosting and database

$300 to $1,200

$2,000 to $7,000

$15,000 to $60,000

Data processing

$200 to $900

$1,500 to $6,000

$10,000 to $45,000

Services and monitoring

$150 to $600

$800 to $2,500

$4,000 to $15,000

Maintenance and support

$1,200 to $3,000

$4,000 to $9,000

$20,000 to $50,000

Total

$2,050 to $6,500

$9,800 to $30,500

$61,000 to $220,000+

Development cost is not the total cost of ownership

A development quote covers building. Total cost of ownership (TCO) covers owning the product over time, including deployment, infrastructure, AI usage, maintenance, security and scaling. This example follows a $65,000 mid-level platform as usage grows.

Cost item

Year 1

Year 2

Year 3

Total

Development

$65,000

$0

$0

$65,000

Launch setup

$4,000

$0

$0

$4,000

Cloud, AI and data

$54,000

$78,000

$102,000

$234,000

Maintenance, support

$24,000

$30,000

$36,000

$90,000

Security, compliance

$6,000

$6,000

$8,000

$20,000

Retraining, upgrades

$10,000

$15,000

$15,000

$40,000

Total

$163,000

$129,000

$161,000

$453,000

Development is about 14 percent of the three-year total. Owning it costs roughly seven times its build price.

Team structure: how the hiring model changes the budget

Who builds the product changes the price as much as what gets built.

•         In-house team: AI/ML engineers, data engineers, developers, designers, QA engineers and a project manager. Control is highest, but salaries run every month and hiring can take months.

•         Freelance developers: hourly rates look low, but AI work, data engineering and design need different specialists, and someone must manage them. Once that time is counted, the AI data analysis tool development cost through freelancers can end up near an agency quote.

•         Software development company: a ready team across design, development, data engineering and QA. Fixed-price contracts suit a clear scope, time-and-material suits changing requirements, and many firms offer post-launch maintenance.

 

Factor

In-house

Freelancers

Development company

Cost predictability

Medium

Low to medium

High with fixed price

Control

Highest

High but split

Shared by contract

Expertise

Whoever you hire

Strong, with gaps

Broad

Scalability

Slow

Limited

Fast

Long-term support

Your own team

Uncertain

Often contracted

Hourly rates by country

Region

Estimated hourly rate

North America

$80 to $150+

Western Europe

$60 to $120

Eastern Europe

$35 to $75

India

$20 to $50

Southeast Asia

$25 to $60

Location is only one factor. AI specialists charge more than general web developers, and experience and complexity push rates higher, so compare like with like.

The lowest hourly rate does not give the lowest total. A cheaper team that takes much longer or rebuilds the pipeline twice can cost more than a pricier team that gets it right first time. The cost to build AI data analysis software is rate multiplied by hours, plus rework, and only the rate appears on a rate card.

How long does the build take?

Development phase

Estimated timeline

Requirement analysis and planning

1 to 2 weeks

UI/UX design

2 to 4 weeks

Backend and frontend development

4 to 10 weeks

Data engineering and integrations

3 to 8 weeks

AI model integration and development

3 to 8 weeks

Testing and optimization

2 to 4 weeks

Deployment

1 to 2 weeks

Phases overlap, so the total is not the sum. In a straight line they add up to 16 to 38 weeks. With parallel work, many projects take roughly three to seven months.

•         Custom models need training runs and repeated testing, which adds weeks compared with an API.

•         Complex pipelines and multiple integrations multiply hours, since each source needs its own cleaning rules and access approvals.

•         Changing requirements are the most common reason budgets grow. A new data source in month four can mean redoing months one to three.

 

Fix data sources and core features before development begins. AI data analytics software development with a clear scope holds its dates far better than one that starts vague.

Hidden costs that quotes often miss

Quotes cover building, not running. These items arrive later:

•         Data acquisition, licensing, cleaning and annotation.

•         GPU infrastructure for training or hosting your own models.

•         Unexpected AI API usage when a feature is used far more than planned.

•         Cloud storage growth and data warehouse query charges.

•         Third-party API upgrades and model retraining.

•         Security audits and compliance consultation.

•         Backup and disaster recovery.

•         Customer support, infrastructure monitoring and performance optimization.

 

A popular usage-based AI feature raises the bill in the same month it raises revenue. A common planning habit is to hold back 15 to 20 percent of the total as contingency.

Security, privacy and compliance

These products hold sales, payroll, customer records and sometimes health or financial data. IBM's 2025 Cost of a Data Breach Report put the global average breach at $4.44 million, with healthcare at $7.42 million and financial services at $5.56 million. It also found that 97 percent of organizations reporting a breach of an AI model or application lacked proper AI access controls.

Security work in a build includes:

•         Encryption of stored data and data in transit, and secure API communication.

•         Authentication (confirming who a user is) and authorization (deciding what they may see), including role-based access control.

•         Anonymization, which removes personal identifiers before data reaches a model or analyst.

•         Audit logs, backup and recovery.

•         GDPR for EU users, CCPA/CPRA for California residents, and rules such as HIPAA for health data.

•         Checks on what data goes to outside AI providers.

 

AI adds a risk older BI tools did not have. A natural language feature can be tricked into revealing data a user should not see, for example by a question worded to bypass filters. Permission checks belong in the database layer, not in the AI's instructions.

Custom AI data analysis software development for financial, healthcare or customer-data platforms needs more: independent audits, penetration testing, stricter logging and sometimes data stored in a specific country.

How to lower costs without hurting performance

Teams trying to cut the AI data analysis software development cost have these options:

•         Start with an MVP, the smallest version that proves people want it.

•         Use existing AI APIs first and build your own models only when volume or accuracy demands it.

•         Choose open-source processing tools such as Spark and Pandas.

•         Skip real-time processing where hourly or daily refresh is enough.

•         Send simple tasks to smaller, cheaper models.

•         Optimize database queries and cache repeated answers.

•         Automate testing and deployment, and reuse existing integrations.

•         Review API and cloud usage weekly, with alerts when spend passes a set level.

•         Add advanced features after real users ask for them.

 

PRO TIP  Log the questions people type most. Those logs show what to build next.

What happens to the cost as the platform scales?

Costs do not rise in a straight line. Some jump at thresholds, such as a database outgrowing one server, while others fall per user.

•         Early-stage platform: limited datasets, fewer users and basic AI processing.

•         Growing platform: multiple data sources, rising query volume and extra infrastructure.

•         Enterprise-scale platform: large datasets, concurrent users, real-time processing, high availability (the system stays up when parts fail) and advanced analytics.

 

The main drivers are active users, data volume, query count, AI inference, concurrent processing, database performance, cloud infrastructure, and security and monitoring.

The table assumes each active user asks 20 questions a month at the checked rate of $0.03 each. Other lines are planning estimates.

Monthly cost

1,000 users

10,000 users

100,000 users

AI usage (20,000 / 200,000 / 2,000,000 questions)

$600

$6,000

$60,000

Cloud infrastructure

$900

$7,000

$55,000

Data processing

$600

$5,500

$40,000

Monitoring and security

$400

$2,500

$15,000

Maintenance and support

$2,500

$9,000

$45,000

Total

$5,000

$30,000

$215,000

Cost per user

$5.00

$3.00

$2.15

Cost per user falls because fixed costs spread across more people. AI is the exception: it grows with usage and is the largest line at 100,000 users, over a quarter of the bill. Watch the AI data analysis tool development cost and the running cost together, because the design that is cheapest to build for 1,000 users is often not the cheapest to run for 100,000.

Pressure exposes weak spots. When many users ask at once, such as Monday reviews of weekend numbers, queries queue up, AI providers enforce rate limits and dashboards slow down. Teams answer with caching, queues and fallbacks, such as showing the last good result with a timestamp. An old number with a clear label beats a blank screen.

Real-time decisions raise a further question: should the system act on a figure the moment it arrives? A fraud alert may need that. A weekly sales report does not.

How can businesses earn revenue from the software?

Teams funding AI data analytics software development for resale usually pick from five pricing models.

•         Subscription (SaaS): monthly or yearly plans by features or users. Revenue is predictable, but heavy AI users can cost more than their plan brings in, so tiers need usage caps.

•         Usage-based pricing: charge by data volume, queries or AI processing. Revenue follows cost, but bills vary.

•         Enterprise licensing: a negotiated fee for custom setup and support. Deals are large, but sales cycles are long.

•         Freemium: basic analytics are free and advanced AI is paid. Every free user still costs hosting and AI calls.

•         White-label: other businesses license the platform under their own brand, opening sales channels but raising support demands.

 

Whatever the model, price against AI inference, infrastructure, maintenance and customer acquisition. A plan that covers only hosting looks profitable until the AI bill arrives.

How to calculate ROI and break-even time

Track these numbers before committing:

•         Customer acquisition cost (CAC): spending on sales and marketing to win one customer.

•         Customer lifetime value (LTV): the gross profit a customer brings before leaving.

•         Monthly recurring revenue (MRR) and average revenue per user (ARPU), which is MRR divided by customers.

•         Customer retention: the share of customers who stay each month.

•         Gross profit margin: revenue left after direct costs such as AI usage and hosting.

•         Monthly operating expenses, and the break-even point when cumulative profit equals the investment.

ROI = (Net Profit ÷ Total Investment) × 100

Net profit here means total profit minus the investment. Take a product that cost $90,000 to build, with 300 paying customers at an ARPU of $60. MRR is $18,000. At a 70 percent gross margin, gross profit is $12,600 a month. Support, maintenance and sales cost $7,600, so net profit is $5,000 a month and the investment is recovered in 18 months. Over 36 months, profit of $180,000 less the $90,000 investment leaves $90,000, an ROI of 100 percent.

AI costs move this quickly. If heavy usage cuts gross margin to 60 percent, net profit falls to $3,200 a month and payback stretches to about 28 months. With 3 percent monthly churn, a customer stays about 33 months, so LTV is about $1,400. Against a CAC of $300, that is a ratio near 4.7. The cost to build AI data analysis software is only the first number here, and the monthly figures decide how soon it pays back.

MVP or full platform: where to invest first

Consideration

MVP

Full-scale platform

Development investment

Lower

Higher

Data sources

Limited

Multiple

AI capabilities

Basic analysis

Advanced AI/ML

Analytics

Standard dashboards

Predictive and real-time

Integrations

Essential

Extensive

Scalability

Expandable foundation

Enterprise-ready architecture

Ideal users

Startups and small businesses

Growing and large enterprises

An MVP tests the one thing that matters most: do people use the output? If nobody does, the loss is $30,000 instead of $300,000. The AI data analysis tool development cost of the full platform is easier to approve once real usage supports it.

A good MVP does one job reliably on clean data from one or two sources. Avoid a throwaway version that must be rewritten when it succeeds.

Final thoughts: what to budget

Most builds fall between $40,000 and $180,000, with simple tools near $20,000 and enterprise platforms past $350,000. Your number depends on AI complexity, how messy your data is and how many systems must connect.

Running costs deserve equal attention. In the three-year example, development was about 14 percent of the total.

A tool that answers ten important questions reliably is worth more than fifty features nobody trusts. When you review any AI data analysis software development cost estimate, ask what it includes, what it leaves out and what it will cost to run in year two.

Prachi Singh

Prachi Singh

Prachi, our dedicated Digital Marketing Manager! With industry experience and expertise, she elevates our online presence and expands our reach. Prachi's eye for detail and data-driven insights help her formulate result-oriented marketing strategies. Her efforts consistently boost our business visibility and contribute significantly to our ongoing success.

Build Your Agile Team

We provide you with a top-performing extended team for all your development needs in any technology.

Hourly
$20
It Includes
Duration
Hourly Basis
Communication
Phone, Skype, Slack, Chat, Email
Hiring Period
25 Hours (MIN)
Project Trackers
Daily Reports, Basecamp, Jira, Redmime, etc
Methodology
Agile
Monthly
$2600
It Includes
Duration
160 Hours
Communication
Phone, Skype, Slack, Chat, Email
Hiring Period
1 Month
Project Trackers
Daily Reports, Basecamp, Jira, Redmime, etc
Methodology
Agile
Team
$13200
It Includes
Team Members
1 (PM), 1 (QA), 4 (Developers)
Communication
Phone, Skype, Slack, Chat, Email
Hiring Period
1 Month
Project Trackers
Daily Reports, Basecamp, Jira, Redmime, etc
Methodology
Agile

Frequently Asked Questions

How much does it cost to develop AI data analysis software in 2027?
Ask three vendors to price the same written list of data sources, features and security needs, with an itemized split across design, development, data work and testing. Differences between quotes then reflect real differences in approach.
How much does AI integration cost in data analysis software?
Connecting a ready-made model usually takes a few weeks for one or two developers. The larger cost sits around it: business definitions, test questions, permission checks and answer monitoring. Ask for these priced separately.
What is the monthly maintenance cost of AI data analysis software?
Ask for a support plan with fixed monthly hours and a stated price for extra hours. Review it after three months, once real usage is known.
Is custom AI model development necessary for data analysis software?
Run a two-week trial first. Send 100 real questions with known answers to an existing model and measure accuracy. If it meets your target, skip custom development. If not, check whether better data descriptions fix the gap first.
How long does it take to build AI-powered data analysis software?
Speed depends heavily on how fast your side provides database access and sample data. Late credentials often slip dates, so arrange access before kickoff.