Data quality: the first thing users notice
Measure the basics: how many duplicate accounts and contacts exist, how many records are missing fields that your process depends on, how many are stale or orphaned, and how many have inconsistent values such as five spellings of the same country. Data problems are visible to users and undermine trust in every report, so they usually rank high in the plan. Look also at who creates records and how, since fixing the entry points prevents the problem from returning.
Quantify the findings. Telling leadership that fourteen percent of accounts are duplicates is more persuasive than saying the data is messy.
- Count duplicates and incomplete records by object.
- Check required-field rules, picklists, and validation for gaps.
- Identify the sources that create the worst data, such as imports or integrations.
Automation: overlapping flows, triggers and rules
Over time an org accumulates several kinds of automation, such as flows, process builders, workflow rules, triggers, and scheduled jobs, often firing on the same object. When they overlap, the order of execution matters and bugs become difficult to trace. Inventory the automation on each important object, note which are active, which are duplicates, and which no one can explain. Consolidating overlapping automations into a smaller, documented set reduces risk and speeds up changes.
Permissions, sharing and access
Review how access is granted: profiles, permission sets, roles, and sharing rules. Look for too many administrators, copied profiles that have drifted, users with broad access they no longer need, and inactive users who still hold licenses or credentials. The aim is least privilege with a design that a new administrator can understand. Document the access model, because access that nobody understands is access nobody can safely change.
Managed and advisory services →Technical debt and unused components
Orgs collect custom fields, page layouts, record types, reports, and code that no longer serve a purpose. Unused components add clutter, slow down administration, and sometimes affect performance. Identify items with no recent use, check dependencies before removing anything, and retire them in stages. For custom code, check test coverage and quality, whether it follows current practice, and whether it relies on retired features that will need replacing.
Limits, performance and integrations
Salesforce enforces limits on storage, API calls, and the processing allowed in a transaction. Review how close you are to them, and which integrations and jobs consume most of the capacity. Slow pages and reports often have identifiable causes such as inefficient queries or oversized layouts. Check that integrations handle failures, log errors, and do not depend on a single person's credentials.
Implementation and integration services →Adoption: are people actually using it?
Technical health means little if users avoid the system. Check login and activity data, how much work is happening in spreadsheets, and what users say about their daily experience. A short survey and a few observation sessions reveal friction that numbers hide. Adoption problems often point to specific fixes, such as simplifying a page or retraining a team, that cost little and help a lot.
Decisions to settle before the review
- Scope. Decide whether to review the whole org or the areas that cause the most pain.
- Access. Agree who can grant the reviewer read access and how sensitive data will be handled.
- Stakeholders. Name business owners to interview so findings reflect real use.
- Outcome. Define the output: a ranked remediation plan with effort and impact for each item.
A realistic timeline
- Week 1. Gain access, gather metadata and usage data, and interview key users and administrators.
- Weeks 2 and 3. Analyze data quality, automation, permissions, code, limits, and adoption, and validate findings with owners.
- Week 4. Deliver the ranked plan, agree priorities, and schedule the first fixes.
Pitfalls to avoid
- Fixing everything at once. Prioritize by risk and value, and deliver improvements in small, tested steps.
- Deleting without checking dependencies. A field that looks unused may feed a report or an integration. Check first.
- Ignoring the cause. Cleaning data without fixing how it enters guarantees a repeat.
- One-off exercise. Schedule a lighter review every year to keep the org healthy.
