Managers open the Monday report first. That is often the one that drags. A customer export, a branch comparison, a summary of last week’s calls: someone starts it, waits, and then asks whether it is time to replace the CRM. A replacement quote answers a different question. The problem is this report. A new product can build the same report in the same bad way: inside the click, row by row, on a table with no index, in a database shared by every branch.
This note is for the product owner. Keep three things in view: the report itself, the person who waits, and whether one customer can see another customer’s data. Queues and indexes address those points. They are not a reason to rewrite the platform. Time the report you already have. Then decide whether the wait belongs in a background job, an index, or a smaller response. A full rewrite comes last, and only after these checks give you a result.
1. Name the report people already wait on
A meeting that starts with “reporting is broken” and jumps straight to a vendor shortlist skips the only fact that matters. “Broken” does not name the report, the branch, or the wait in seconds. Two reports can live in the same product and have nothing in common. One is a short list. Another joins every call, invoice, and message for a month.
The person who feels the problem is not looking at the database. They see a spinner, a timeout, or a spreadsheet that arrives after the stand-up ends. Then the team proposes a new CRM, because that is a familiar decision. The report stays unnamed, so the new CRM inherits the same problem.
Write this on one page next to the report:
A platform-wide claim — “the CRM is slow” — will not tell a developer which query to inspect. The same rule applies on a public site: name the page before you redesign it. That wider pass is covered separately in what to check before a redesign.
2. Why a report crawls even when the product is not old
Age is a weak explanation. A new product slows down for the same reasons as an old one. The report hits the database once per row. The filtered column has no index. The response includes every field even though the page shows five.
The first version of a report is often a loop. For each customer, load calls. For each call, load a score. The logic is correct, and it becomes slow when the row count is real. The same pattern appears outside CRMs too. On one product, a query loop became one aggregating query, and that operation dropped from a bit over twelve seconds to a bit over two. That number belongs to that operation. It is not a forecast for your Monday export. What it shows is the shape of the work: one query instead of a loop, an index where search needs one, and a response no larger than the page needs.
Ask the person who can inspect the query:
3. When the report should leave the click
Some reports keep the browser open until every row is done. People click again. If the same heavy work starts again, the wait gets worse. If the job sends mail, writes a charge, or posts a score, a retry can also create the same side effect twice.
The report became a page because a page was the fastest way to ship it once. Monthly ranges, branch roll-ups, and file exports outgrow that page. The work can still be correct. It just no longer fits inside one request. A queue moves the run out of the click, stores the result, and lets the person return when it is ready. That is a much smaller change than replacing the CRM.
Move it off the click when these points are true:
On one multi-tenant CRM, call processing ran as queue jobs. A retry had to find the work already recorded and stop, so a second attempt did not create a second result. That workflow connected call handling, telephony, messaging, and accounting, then exposed the outcome through reporting. Your report may not score calls. The question is the same: if this run happens twice, do you get one row or two? Ask that even when the queue is a simple one you already operate.
4. Check branch and customer boundaries before you speed the report up
A faster report that shows the wrong customer’s rows is a worse report. Speed only makes the leak easier to notice. In a CRM or SaaS product, branches, tenants, or accounts can share one application, but they must not share each other’s records.
The convenient query is “all calls this month,” with the filter applied later on the page. Or there is one shared database, and some queries forget the column that says whose rows these are. An index can make that query faster. It cannot add the missing filter. On one product, each branch had its own database, and the migrations and indexes were built for that separation. The index supported the boundary. It did not replace the boundary.
Check the boundary before you celebrate a faster export:
5. A short plan that is not a rewrite
A proposal with two lines — “reporting is slow” and “replace the CRM” — is missing the third line you can write today. Immediate: the report people wait on, timed, with a suspected cause. Next: either one aggregating query and an index, or a background run that is safe to retry, plus a written answer on how accounts stay separate. Later: a new platform, if the product still fails the decision this report exists to support after those changes. Inherited production systems are usually steadied before anyone treats them as disposable. The real decision is what to steady first.
Keep the plan small enough to finish:
When to bring in a specialist
You can name the report, time it, and open it under two accounts yourself. Bring in help when the suspected cause is a query or an index and nobody inside can read it, when a retry might send a second email or a second charge, or when a replacement quote is already moving and it does not mention this report.
I do this kind of reporting work on an existing CRM or SaaS product, without starting from an empty repository. Services cover performance, SaaS and backend systems, and AI integration, including cases where a report sits on top of telephony or another system you already run. Selected work shows the same class of system: a report that had to stay correct under retry, and data that had to stay inside its own account. For one report that drags on a CRM you already run, performance and scalability means the query, the payload, and the repeat that must stay safe, with no rewrite of the product. If you want to walk through one report that keeps dragging, write to me with its name, who waits on it, and the seconds you wrote down.