A directory of legal professionals lives on search traffic and on trust. Visitors browse listings in a hurry, often comparing several cities, and they leave quickly if a page will not load. When the platform behind such a directory stumbles, the owner needs more than good intentions. This audit turns the practices catalogued in the Ops Atlas under IncidentOps into questions you can score with evidence. It is an operations review for site owners, not a view on any lawyer, any listing or any legal question, and it makes no claim about the quality of professionals shown on a directory.
Running the review
Block out a couple of hours and bring in one engineer, one content owner and one person who handles visitor enquiries. For each item below, collect the named evidence rather than relying on memory. Score it as pass, partial or fail. At the end, choose the two weakest items and plan a fix within the month. The audit is deliberately analytical: a claim counts only when somebody can show a document, a log or a screenshot.
Section A: Declaring and classifying
A1. Are the triggers written?
Evidence: a one-page policy that lists what turns a fault into an incident. Typical triggers in public playbooks are visible customer impact, a need for a second team, or an unresolved problem after about an hour.
Pass signal: a new staff member can find it in under a minute and apply it to a hypothetical outage.
A2. Is there a short severity scale?
Evidence: two or three levels, each tied to a response: who is paged, how soon, and how often the public hears from you.
Pass signal: past incidents carry a level in their records.
A3. Do staff feel safe declaring?
Evidence: a count of declared incidents over the last two quarters, and a short staff poll. The Atlas warns that punishing the number of incidents suppresses honest reporting, so a suspiciously low count is itself a finding.
Pass signal: small incidents are declared and nobody is criticised for it.
Section B: Roles and rota
B1. Is the coordinator separate from the fixer?
Evidence: the last three incident timelines. Look for who made production changes and who coordinated.
Pass signal: different people, with the coordinator never appearing in the change log.
B2. Is on-call load sustainable?
Evidence: the rota and a simple calculation of each person’s share of slots. The Atlas works through a case of six people covering two slots around the clock every week, which lands near a third of their time and breaches a widely quoted ceiling of a quarter. Count the same for your own team.
Pass signal: nobody exceeds roughly a quarter of the cover.
B3. Are hand-offs explicit?
Evidence: timelines that show a spoken or written acceptance of command at each shift change.
Pass signal: no gap or overlap in command is visible.
Section C: Communication and independence
C1. Does the status notice survive an outage of the main site?
Evidence: the hosting arrangement for the status page and a test in which the main application is deliberately switched off in staging. The Atlas describes the architecture rule that the paging route and the status notice must live apart from the systems they describe.
Pass signal: the notice loads while the main site is unavailable.
C2. Is the update rhythm met?
Evidence: timestamps from recent serious incidents. Public playbooks suggest a first note within about five minutes of the response starting and new posts roughly every twenty minutes during the opening two hours.
Pass signal: the Atlas sets a maturity bar of keeping that rhythm in nine of ten of the most serious incidents.
C3. Do we avoid waiting for upstream confirmation?
Evidence: incidents where a hosting vendor’s dashboard stayed green while visitors saw errors. The Atlas observes that provider status pages can lag.
Pass signal: your own criteria, not the vendor’s colour, trigger your response.
C4. Is the working channel tidy?
Evidence: a screenshot of participant counts from a recent response.
Pass signal: a small working group and a separate read-only feed for everyone else. Emergency-management guidance of about five reports per supervisor is a useful yardstick.
Section D: Security of the tooling itself
D1. Are paging secrets protected?
Evidence: how the paging configuration is stored. If it is managed with Terraform, remember the Atlas’s caution that marking a value as sensitive hides it from plan output only, while state files and raw outputs can still leak the routing keys.
Pass signal: encrypted remote state with restricted read access.
D2. Is the incident record written once?
Evidence: a single timeline that feeds the customer note, the internal review and any formal report.
Pass signal: nobody retypes facts from memory.
Section E: Learning
E1. Is a review owner named in time?
Evidence: review dates against incident dates. Published schedules range from a few days for the gravest incidents to about a week for the next tier.
Pass signal: reviews happen inside the window you set.
E2. Are follow-up actions finished?
Evidence: the open list, with owners and dates. The Atlas proposes a bar of four in five priority actions finished on time across two quarters.
Pass signal: few items are vague, and none was closed merely by being downgraded.
E3. Are measurements honest?
Evidence: your latest incident report. Averages alone are fragile with small samples, because a single severe event can swing them; the Atlas prefers distributions, error-budget impact, pager load and closure rate.
Pass signal: every figure names its phase and its population.
Reading the results as a pattern
Individual scores matter less than clusters. If Section A is weak, your team probably hesitates to call incidents, and everything downstream starts late. If Section B is weak, you may have capable people but no shared picture of who does what, which shows up as duplicated effort and contradictory changes. A poor Section C usually means customers learn about outages from social media instead of from you. Weakness in Section D is the quiet risk, because it only becomes visible after a leak. A low Section E score means you are paying for each incident repeatedly because the lessons do not stick.
Look also at how the evidence was gathered. If most items needed an engineer to recall details from memory, record keeping itself is the first thing to improve. A good incident timeline answers most of this audit without anyone having to remember anything.
Scoring and next steps
Give two points for pass, one for partial, none for fail. Out of thirty, below ten suggests you are still relying on individuals, between ten and twenty indicates a defined but untested process, and above twenty suggests a well-practised team. Whatever the total, note your two weakest items, give each one an owner and a target date, and repeat the exercise in three months.
If you maintain city-by-city listing pages, for example a city page such as Guwahati’s roundup of IP advocates or the matching roundup for Madurai, remember that these are the pages that must stay reachable when visitors need them. The audit protects them.
For the reasoning, sources and disagreements behind each item, read the Atlas’s long IncidentOps reference. Pick the weakest section, fix one item, and run a short drill before the quarter ends.
Related posts:

Hardest-First Working in Legal Information: A State of Play

A Case Study: An Office Team Rereads Its Admin Goals Monthly

SecDataOps FAQ for Directory and Legal Information Platforms

Practice-Ready: Setup Habits That Help Lawyers Build Their Career

ThreatOps for an Advocate Directory: Four Problems and Their Fixes
