TL;DR
Data accessibility powers better decisions by giving the people and AI agents who decide trusted, governed data within minutes. It means authorized users can find, understand, trust and use the data they need in time to act. The seven gains are faster, consistent, frontline, cross-functional, confident, safer and AI-assisted decisions. Each one rests on a catalog, shared metric definitions and access that is governed rather than locked down. Open access without governance only creates new risk, so the goal is controlled self-service. Measure progress with time-to-data, the hours between asking a question and holding a trusted answer.
Key Takeaways Data accessibility goes beyond availability: the right people and systems can find, understand, trust and use data in time to act. Better decisions need data that is quick to reach and trusted at the same time. Shared metric definitions end the debate over whose number is right, so meetings move on to what the number means. Governed self-service, with least privilege, masking and audit trails, widens access without widening risk. Time-to-data is the clearest single measure of progress, backed by request turnaround, certified coverage and adoption. Copilots and AI agents need the same governed access as people, with their own scoped and auditable identities. Monday’s Allocation Call Ran on Last Month’s Numbers On a Thursday afternoon, the supply chain lead at a multi-plant manufacturer asks for on-time delivery figures by region. She needs them for Monday’s allocation call, where scarce inventory gets split across customers.
The numbers exist, because shipments sit in SAP, and the delivery history sits in the data warehouse. Yet nobody on her team can query either one, so the request joins the analytics queue.
Three days later the report finally arrives, after the call has already happened. The allocation went out on last month’s figures, and two regions that had slipped got the same share as before.
Nothing in this story is a technology failure. The data was available the whole time, but it was not accessible to the person who had to decide. That gap, between data that exists and data someone can act on, is what data accessibility is about.
Below are seven ways closing that gap makes decisions better. After them come the barriers that keep it open, a 90-day plan, the platforms that help and the metrics that show progress.
What Is Data Accessibility? Data accessibility is the ability of authorized people and systems to find, understand, trust and use the data they need in time to act. It is less about storage and more about the path between a question and an answer.
Each of the four verbs does real work. For example, find means a person can search for a dataset instead of knowing whom to email. Understand, similarly, means the columns carry business names and plain definitions.
Trust means the reader can see freshness, quality and ownership before relying on a figure. Use then means access arrives through a tool that person already works in, whether that is a report, a spreadsheet or a SQL editor.
Systems belong in the definition too. A forecasting model, a Power BI report or an AI agent also consumes data. So each one needs the same governed path as a human analyst.
The phrase has a second meaning as well. In design, an accessible dashboard is one that people with disabilities can read, following web content accessibility standards such as WCAG. That matters, but it is a separate topic, covered briefly in the FAQ below.
Accessibility vs Availability, Democratization and Security Four neighboring terms tend to blur together in planning meetings, so the table separates them by the question each one answers.
Term Question it answers Example Data availability Is the data stored, online and recoverable? The sales warehouse is running and its backups restore cleanly. Data accessibility Can the right person find, understand and use it in time? A regional manager opens a certified sales dataset in Power BI without filing a ticket. Data democratization How widely is data access spread across the company? Self-service tools and training reach every business unit; see these data democratization tools . Data security Who must be kept out, and how is that enforced? Salary columns are masked for everyone outside HR.
Availability is the floor, while security sets the boundaries. Accessibility is the work in between, and democratization is how far you spread that work once it runs well.
7 Ways Data Accessibility Powers Better Decisions A decision gets better when it happens sooner, rests on the right number, sits with the right person or carries less risk. Data accessibility moves each of those levers, so the seven ways below map to them one by one.
For each way you will find the decision it improves, what makes it possible, a short example and the sign that it is working. A closing table then puts all seven on one page.
1. Faster Decisions: Time-to-Data Drops From Days to Minutes Speed comes first, because waiting is the default. When every question becomes a report request, the answer tends to arrive after the moment to act has passed.
Two changes shift the timing. A searchable catalog lets people find an existing dataset instead of asking for a new one, as our guide to the enterprise data catalog explains. Certified datasets then tell them which version to use, so they also skip the round of checking with the data team.
Consider, for instance, a replenishment planner who sees demand jump for one product family. With a certified sales and inventory dataset in reach, she adjusts reorder points the same afternoon. Under the old process, that change would have waited for next week’s report.
The sign it is working is simple: routine questions get answered without a ticket, and the analytics queue shrinks to genuinely new work.
2. Consistent Decisions: One Definition for Every Metric Consistency matters as soon as two teams decide from the same number. If finance and sales calculate revenue differently, every review meeting opens with a debate about whose figure is right.
The fix is a shared semantic model, where each metric has one definition, one owner and one calculation. Reports then reuse that model instead of rebuilding the logic each time. As a result, a change to the formula reaches every report at once.
The problem is easiest to see at scale. A global packaging company Kanerika worked with had more than 10,000 measures spread across 400+ reports, and definitions drifted between business units. An automated metadata repository in a Fabric Lakehouse cut measure duplication across those reports.
A semantic layer also helps AI tools , because an assistant that reads governed definitions is less likely to invent its own. The sign this way is working: meetings spend their time on what the number means, not on reconciling it.
Watch on YouTube
LLM Hallucination in Enterprise Data: The Semantic Layer Fix
Kanerika explains why AI assistants give wrong answers on enterprise data and how a governed semantic layer, with one definition per metric, keeps their answers consistent.
3. Frontline Decisions: Data Reaches the People Closest to the Problem The best person to act on a problem is often the one standing next to it, because proximity shows problems early. A plant manager notices a line slowing down long before a central analyst does, but only if the plant’s data actually reaches him.
Role-scoped self-service is what makes that safe. With row-level security, one report serves every plant while each manager sees only their own site’s rows, as Microsoft’s row-level security guidance describes. The setup steps appear in the trade-offs section below.
A store manager gives a second example. When she compares her shrink rate against the regional average, she can adjust weekend staffing on Friday morning. No analyst touches the request, and no other store’s data appears in her view either.
Our walkthrough of Power BI row-level security covers the setup in detail, and our review of self-service analytics software compares the tools. Either way, the sign it is working is that frontline managers open their reports every week without asking for help.
4. Cross-Functional Decisions: Joined-Up Data Across Finance, Supply Chain and Sales Some of the most expensive decisions also cross department lines. Sales and operations planning, for instance, needs demand from sales, capacity from operations and margin from finance in a single view.
When each function keeps its own copy of the data, every planning cycle begins with stitching spreadsheets together. Breaking those silos means landing the data in one governed store and modeling shared dimensions such as product, customer and region once.
In the packaging case, a Center of Excellence model and OneLake let Finance, HR and Supply Chain work from unified and consistent KPIs. Larger groups sometimes go further with domain-owned data, a pattern our guide to data mesh principles explains.
The sign it is working: a margin decision can be traced through sales, cost and supply data without anyone exporting a file. Planning meetings also get shorter, because the numbers already agree before anyone walks in. For the supply side in more depth, see our guide to supply chain analytics .
Case Study
60% More Data Accessibility with Microsoft Fabric
Kanerika cut duplication across 10,000+ measures in 400+ reports for a global packaging leader on Microsoft Fabric and Purview, delivering a 60% increase in data accessibility and a 45% improvement in decision-making.
Read the Case Study → 5. Confident Decisions: Freshness, Quality and Lineage Sit Next to the Number A number is only useful if the reader knows whether to trust it. Without that context, cautious managers re-check everything, while hasty ones act on stale data instead.
Accessible data, by contrast, carries its context with it. For example, endorsement labels show which datasets are certified, refresh times show how current they are, and data lineage shows where each figure came from. Microsoft’s OneLake catalog , for example, is the central place in Fabric to discover data and check its governance status, through its Explore and Govern tabs.
Picture a finance lead about to approve a price change who sees that the margin dataset last refreshed four days ago. Because she spots it before acting, she waits for the overnight load instead of deciding on an outdated view.
Quality checks also belong in the same picture. Data observability tooling flags broken pipelines before a dashboard shows a wrong number, so trust problems surface early. The sign it is working: people stop asking “is this right?” and start asking “what should we do?”
6. Safer Decisions: Governed Access That Does Not Slow Anyone Down Wider access does raise risk when governance is bolted on afterwards. That is why accessibility and security have to be designed together rather than traded against each other.
The stakes are concrete, too. IBM’s Cost of a Data Breach Report 2026 puts the global average breach at USD 4.99 million. That is a record high and 12% more than the prior year.
Governed access answers with least privilege, masking for sensitive columns and an audit trail on every read. These map to the Access Control and the Audit and Accountability families in NIST SP 800-53 Rev. 5 , a catalog many security teams already follow. When done well, the controls run inside the platform, so a business user never waits on them.
Role-based and attribute-based models, and the review cycle behind them, are covered in our data access governance framework guide. For masking and encryption, see these data security best practices . The sign it is working: access requests close in hours, and audit questions are answered from logs rather than interviews.
7. AI-Assisted Decisions: Copilots and Data Agents Need the Same Governed Access Some of the newest decision-makers are not people at all. Copilots and data agents answer questions, draft analyses and sometimes trigger actions, so they need data access just as people do.
The Microsoft Fabric data agent is one example. It is generally available and builds conversational Q&A over an organization’s data, so a manager can ask a question in plain language. Purview risk and compliance controls for data agents are still in preview, which shows governance catching up with the capability.
IBM’s 2026 report offers practical advice here. It recommends securing agent identities with dynamic, identity-based access controls and tightly scoped permissions enforced at runtime. In addition, it calls for human attribution and auditability, so that every agent action traces back to a person.
In practice, an agent should inherit the same certified datasets, row-level rules and audit trail as the human it serves. The sign it is working: agent answers match the governed dashboard, and every agent query lands in the same logs as everyone else’s.
Once access is governed, the next step is forecasting on the same data; the episode below covers seven ways predictive analytics builds on business intelligence.
The Seven Ways at a Glance The seven ways follow one pattern. Accessible data shortens the wait, aligns the numbers and widens who can act, while governance keeps all of it safe at the same time. The table below sums up each way.
Way Decision it improves What enables it Sign it is working 1. Faster Pricing, replenishment and scheduling calls Catalog search and certified datasets Routine questions answered without a ticket 2. Consistent Any decision built on a shared KPI Semantic model with one owner per metric Meetings stop reconciling numbers 3. Frontline Plant, store and branch actions Row-level security and self-service reports Weekly use by frontline managers 4. Cross-functional S&OP, margin and workforce planning One governed store with shared dimensions No spreadsheet stitching before reviews 5. Confident Approvals that depend on current data Endorsement, refresh time, lineage and quality checks Fewer “is this right?” questions 6. Safer Any decision touching sensitive data Least privilege, masking and audit trails Access requests close in hours 7. AI-assisted Questions asked through copilots and agents Scoped agent identities on the same governed data Agent answers match the dashboard
What Stops Data From Reaching Decision-Makers? If accessible data helps this much, why do so many decisions still wait on data? Usually because several small barriers stack up between the source system and the person deciding.
None of them is exotic, though. Silos keep data in departmental systems, missing metadata hides what exists, and approval queues stretch a simple request into weeks. Our wider list of data management challenges covers the full set. The infographic shows the five that come up most often; the table adds raw data overload and cost.
Each barrier also leaves a recognizable symptom, which makes diagnosis easier than it sounds. If people keep asking analysts where data lives, the catalog is the gap. When they find the data but argue about it, the gap is definitions and quality instead.
The table below pairs each barrier with its effect on decisions and the fix that usually works.
Barrier Effect on decisions Fix Data silos Each team decides from a partial view Land data in one governed store with shared dimensions Poor data quality Leaders distrust the figures and delay action Quality rules and freshness shown beside the data No metadata or catalog People rebuild datasets or give up looking Searchable catalog with business definitions and owners Permission bottlenecks Requests wait days or weeks Self-service requests with policy-based approval Skills gaps Data is reachable but misread Role-based training and report templates Raw data overload Users face tables they cannot interpret Curated, certified data products per domain Cost Programs stall before they reach users Start with the decisions that wait longest, then expand
Notice that most of these fixes are design and ownership choices rather than purchases, which is why the plan below starts with the decisions themselves.
How to Improve Data Accessibility Without Losing Control Improving data accessibility works best as a sequence rather than a single project. The seven steps move from choosing what to fix, through building trust, to measuring the result.
Start from the decisions that wait on data. List ten recurring decisions and note how long each one waits for data. The longest waits on the decisions that matter most then become your first scope.Catalog and certify trusted data. Register the datasets behind those decisions, give each one an owner and certify the versions people should use. Our guide to data governance best practices covers ownership and stewardship, and this comparison of data catalog tools helps with selection.Agree on shared definitions. Build a semantic model so revenue, margin and on-time delivery each have one formula. Then publish the definitions where people read the reports.Make access requests self-service with least privilege. Let users request a data product, with policies deciding who is approved. In Microsoft Purview’s Unified Catalog , users subscribe to data products, and access policies pair those self-service requests with compliance requirements.Match delivery to the user. A business user needs a report, while an engineer needs an API. Pair each method with data literacy training so people read the data correctly.Add freshness, quality and lineage checks. Show refresh time and quality status beside each certified dataset, and alert its owner when a check fails.Measure and review every quarter. Track the metrics in the measurement section below, then review them with business owners each quarter.Steps two to four carry most of the effort, mainly because they turn scattered tables into products people can trust. The Data Analytics Checklist below turns all seven steps into a working list for your team.
Checklist
Data Analytics Checklist
Kanerika’s Data Analytics Checklist helps teams plan an analytics program step by step, a practical companion to the seven steps above.
Get the Checklist → Match the Access Method to the User Step five is where many programs stall, since they hand everyone the same tool. For business users, self-service business intelligence is usually the right start, but other roles need other routes. The table matches each type of user to an access method.
User What they need Best access method Business user Answers to recurring questions Certified Power BI reports and dashboards Analyst Room to explore within guardrails Semantic models, governed datasets and SQL endpoints Data scientist Long history and model features Notebooks on governed lakehouse tables Data engineer Pipelines and integrations APIs, catalog-registered sources and CI/CD Executive A short, trusted summary Scorecards showing endorsement and refresh time AI agent Scoped, auditable access Its own identity, least-privilege permissions and logged queries
A 30/60/90-Day Plan for One Business Domain A first phase fits into about 90 days if it stays narrow. The plan also assumes one business domain, such as supply chain, and a data platform that is already in place.
First 30 days: pick the domain and its ten decisions, measure baseline time-to-data and register the datasets behind them.By day 60: certify the top datasets, publish shared metric definitions and switch on self-service access requests.By day 90: roll out role-based reports with row-level security, train users and report the first movement in the metrics.Once the first domain shows results, repeat the same 90 days for the next one. Each repetition goes faster, because the catalog, the access policies and the training material already exist.
Four Common Mistakes The same four mistakes show up in many first attempts, although each is cheap to avoid at the start.
Buying a catalog before deciding which decisions it should serve. Opening access without shared definitions, so more people argue about more numbers. Treating security as a final gate instead of a policy built into every request. Counting logins instead of time-to-data and decision speed. All four are much harder to unwind later, since habits and reports soon form around them.
Which Platforms and Tools Support Data Accessibility? Most enterprises run on one of three data platforms, and each can deliver governed access when it is configured for it. The comparison below stays at the capability level; for tool-by-tool reviews, see our guide to data democratization tools .
Capability Microsoft Fabric and Purview Databricks Snowflake Discovery and catalog OneLake catalog (Explore, Govern and Secure tabs); Purview Unified Catalog for data products Unity Catalog as one governance layer for data and AI Horizon Catalog, described as an agentic catalog for data inside or outside Snowflake Access control Purview access policies for self-service requests; Power BI row-level security for report viewers Access control enforced on queries and model calls Sensitive data protection within Horizon governance Lineage and audit Governance status in the OneLake catalog; Purview for labels and audit Lineage tracking and activity logs for auditing End-to-end lineage and data quality in Horizon AI and agent access Fabric data agent (generally available); Purview controls for agents in preview The same Unity Catalog rules apply to model calls AI guardrails within Horizon governance Setup note RLS roles in Power BI Desktop, members in the service, then Test as role Unity Catalog is on by default for workspaces created after November 8, 2023 Horizon covers data both inside and outside Snowflake Trade-off to plan for RLS restricts only Viewer permission, not workspace roles Partners outside Databricks need OpenSharing, a separate sharing setup Horizon also spans data outside Snowflake, so agree early which catalog is the system of record
Sources: Microsoft Learn on the OneLake catalog , Databricks documentation on Unity Catalog and Snowflake documentation on Horizon Catalog .
Kanerika Service
Microsoft Purview Consulting and Implementation
Kanerika implements Microsoft Purview catalogs, sensitivity labels, access policies and audit trails alongside Fabric, so wider data access stays governed from the first request.
Explore Purview Services Two Trade-Offs to Plan For Two trade-offs deserve a closer look before you commit. The first is the scope of row-level security in Power BI. RLS restricts only users with Viewer permission, so workspace Admins, Members and Contributors still see every row. Keep report consumers on Viewer access, and use one dynamic rule rather than a role per site:
// Role "Plant Managers", created in Power BI Desktop (Modeling > Manage roles)
// Table filter on 'Plant': each signed-in manager sees only their own plant
'Plant'[ManagerEmail] = USERPRINCIPALNAME()After publishing, add the managers as role members in the Power BI service and check the result with Test as role. As a result, one role covers every plant, and a new manager needs only a row in the Plant table.
The second trade-off is sharing beyond your own platform. Every external share is a second access path, with its own policies, expiry and audit outside your internal rules. Databricks offers OpenSharing , an open protocol it developed for secure data sharing with other organizations, regardless of their computing platform. For the governance side on each platform, see our guides to Databricks Unity Catalog , Snowflake data sharing and data governance with Microsoft Purview .
How Do You Measure Data Accessibility? Accessibility is measurable, and the cleanest unit is time. Start with time-to-data, then add five supporting metrics that each protect a type of decision.
First, set a baseline in the first 30 days, before anything changes. Then give each metric an owner and a target agreed with the business, not with IT alone.
Time-to-data leads the list because everyone understands it. It is the elapsed time between someone asking a question and holding a trusted answer. You can take it from ticket timestamps, or from a short monthly survey when tickets do not capture the wait.
The other five then explain why time-to-data moves. A short request turnaround with low adoption, for example, points to a skills problem rather than an access problem.
Metric How to calculate Decision it protects Owner Time-to-data Median hours from question to trusted answer Every time-sensitive decision Analytics lead Access-request turnaround Median hours from request to grant Frontline and analyst decisions Data governance lead Certified coverage Share of critical datasets that are cataloged and certified Consistent decisions Data owners Self-service adoption Share of target users active each month without analyst help Frontline decisions Business unit heads Freshness SLA hit rate Share of scheduled refreshes that land on time Confident decisions Data engineering lead Decision cycle time Days from issue raised to decision made, for tracked decisions The business outcome itself Executive sponsor
Finally, review all six together each quarter. If time-to-data falls while adoption stays flat, the data got faster but people did not follow, so training becomes the next investment.
How Kanerika Improves Data Accessibility Kanerika builds governed self-service on Microsoft Fabric, Power BI and Purview. Its work for a global packaging leader shows how that raises data accessibility from start to finish.
The client, a packaging solutions provider present in more than 100 countries, had 10,000+ measures across 400+ reports and 600 production dataflows. Its data also came from SAP, Azure Synapse, SQL Server and Marketo. Definitions differed across those systems, and business units lacked curated, governed datasets.
Kanerika made OneLake the unified store and built an automated metadata repository in a Fabric Lakehouse to manage the measures. Next, some legacy dataflows moved to Dataflow Gen2 and PySpark with CI/CD. Microsoft Purview then supplied naming conventions, RBAC, audit trails and sensitivity labels. Finally, Power BI templates, training and a Community of Practice helped people adopt the new model.
The results, as published in the case study:
Outcome Result Data accessibility 60% increase Decision-making 45% improvement ETL processing time 30% reduction
Read the full packaging case study for the detail. The same approach also runs through Kanerika’s data analytics services , data governance services and Microsoft Fabric practice . Each covers the path from catalog design to row-level security and user adoption.
Talk to Kanerika
Where Does Your Data Wait Longest?
In a short working session, Kanerika’s data team maps the decisions that wait on data in one domain, measures today’s time-to-data and outlines a 90-day plan to cut it.
Book a Working Session → Bottom Line: Accessible Data Turns Waiting Into Deciding All seven ways share one mechanism. When people and agents can find, trust and use governed data on their own, decisions get faster and more consistent. They also get safer at the same time.
So start with the decisions that wait longest, measure time-to-data and widen access only as fast as governance keeps up. Accessible, trusted and governed data powers better decisions because it turns time spent waiting into time spent deciding.
Frequently Asked Questions
What does data accessibility mean? Data accessibility means authorized people and systems can find, understand, trust and use the data they need in time to act on it. A dataset that exists but sits behind an analyst queue, carries no definitions or cannot be trusted is available but not accessible. Data access is the narrower act of reading or querying it.
What is the difference between data availability and data accessibility? Data availability asks whether data is stored, online and recoverable when systems need it. Data accessibility asks whether the right person can actually find, understand and use that data in time to decide. A warehouse can run with perfect uptime while business users still wait days for a report, which is high availability with poor accessibility.
How do you measure data accessibility? Lead with time-to-data, the elapsed time between asking a question and holding a trusted answer. Support it with access-request turnaround, the share of critical datasets cataloged and certified, monthly self-service adoption, freshness SLA hit rate and decision cycle time. Set a baseline in the first 30 days and review all six with business owners quarterly.
What is the difference between data accessibility and data democratization? Data accessibility is about whether the right person can find, trust and use specific data in time. Data democratization is about how widely that access spreads across the organization, usually through self-service tools and training. Accessibility comes first, because spreading access to data people cannot understand or trust only multiplies confusion and risk.
How can data accessibility be improved without compromising security? Design access and security together. Use least-privilege roles, row-level security and masking for sensitive columns, and route requests through policies that approve the routine cases automatically. Keep an audit trail of every read so reviews rely on logs. Governance then runs inside the platform, and business users get speed without the organization taking on extra risk.
How does data accessibility support AI agents? AI agents and copilots answer questions from enterprise data, so they need the same governed access as people. Give each agent its own identity with tightly scoped permissions enforced at runtime, point it at certified datasets and log every query. Its answers then match the governed dashboards, and each action can be traced back to a person.
How much does improving data accessibility cost? Cost depends on how many domains are in scope, whether a modern data platform already exists, how many sources need integration and how much catalog, governance and training work is required. Starting with one domain and its ten slowest decisions keeps the first phase contained, shows measurable results and builds the case for wider investment later.
How can companies improve data accessibility? Start from the decisions that wait longest for data, then catalog and certify the datasets behind them. Agree on shared metric definitions in a semantic model, make access requests self-service with least privilege and match delivery to each type of user. Add freshness and quality checks, then measure time-to-data every quarter to confirm progress.
What are the types of data access? Most enterprises offer five routes to data. Business users read certified reports and dashboards, analysts query semantic models or SQL endpoints, and data scientists work in notebooks on governed tables. Engineers connect through APIs and pipelines, while partners receive data through governed sharing protocols. AI agents now form a sixth route that needs its own scoped identity.
What is accessibility in data quality? In data quality frameworks, accessibility is one dimension alongside accuracy, completeness, consistency and timeliness. It measures whether authorized users can retrieve the data easily and in a usable form. Data can be accurate and complete yet still fail on accessibility if it is locked in a system nobody can query or lacks the definitions needed to read it.
What is accessibility in data visualization? In visualization, accessibility means charts and dashboards can be read by people with disabilities, including those using screen readers or with color vision differences. Good practice follows web accessibility standards such as WCAG, with text alternatives, sufficient contrast, keyboard navigation and patterns beyond color alone. This is a different meaning from governed access to enterprise data.
Why is data accessibility important in business? Decisions are only as timely and accurate as the data behind them. When data is hard to reach, teams decide on old numbers, argue over conflicting figures or wait for analysts. Accessible, governed data shortens that wait, aligns everyone on one version of each metric and lets frontline staff act without escalating every question to a central team.
How does data accessibility improve decision-making? It improves decisions in seven ways. Choices become faster, consistent across teams, possible for frontline staff, joined up across functions, more confident because freshness and lineage are visible, safer through governed access and better supported by AI agents. The common thread is less time waiting for data and more time acting on figures people trust.
What are the main barriers to data accessibility? The most common barriers are data silos, poor data quality, missing metadata or catalogs and slow permission processes. Skills gaps, raw data that users cannot interpret and project cost also hold programs back. Most of these are fixed through ownership, shared definitions and self-service access policies rather than by buying another tool or platform.
How does a data catalog improve data accessibility? A data catalog lets people search for datasets instead of asking colleagues where data lives. It shows business definitions, owners, certification status, freshness and lineage beside each asset, so users can judge whether to trust it. Modern catalogs also handle access requests, which turns a slow email chain into a governed, trackable workflow with clear approvals.
