Seven problems solved in production, across integration, contracts, data, Salesforce and AI. Each shows what was broken, how it was fixed and what changed. If one sounds like your company, start the conversation there.
Account teams could not see what customers did on the platform. Warehouse metrics now land on every Salesforce account each hour and match the BI team’s official figures.
Systems integration with Snowflake, Salesforce and BI
Rows arrive every hour and the count matches the warehouse.
Permission dropped
A warehouse permission is removed upstream. The sync stops receiving rows and still reports no errors.
Drift check fires
The hourly count stops matching. An alert reaches the owner the same day.
Caught up
The permission is restored and every missed row is loaded from where the sync stopped.
One set of numbers
BICRMStorefront
One agreed definitionThe CRM shows the official figures
Before the work, each system counted the same activity its own way.
How the sync signs in
Integration-only licenceNo person’s login behind the sync
Certificate sign-inNo stored password to leak
Access scoped to customer accountsAnd no delete rights at all
How it was done
The problem
Account teams were making renewal and upsell calls without seeing what customers did on the platform. The numbers lived in the warehouse and the people lived in Salesforce, while BI, the CRM and the storefront each produced a different count of the same activity.
What we did
Built a Snowflake-to-Salesforce pipeline on an integration-only licence with certificate-based authentication and sharing scoped to customer accounts, so the pipeline never held delete rights. Agreed every metric definition with the BI team first, then reproduced their official figures exactly before shipping. Added drift checks so a silent stall surfaces the same day.
Signed contracts turned into reportable data
Nobody could say which customers had signed which version of the terms. Every signed contract was read, and the terms now sit on each account as fields the business can report on.
Revenue & contract operations with DocuSign, Salesforce and Legal
From a clause in the PDF to a field on the account
Signed order form, page 2
Account record
Contract term
24 months
Platform fee
$1,200 a month
Notice period
60 days
Terms version
2026 standard
Example contract with invented terms. Pick a clause to see the field it becomes.
What changed
AreaBeforeAfter
Who signed which termsNobody could say without opening PDFsOne report answers it
New terms rolloutEvery customer assumed to need new paperOnly the accounts Legal’s rules require
Then the contract workflow
Clause library owned by LegalPre-approved wording, kept current by Legal
Approval routing by clauseEach clause carries its approver
Contracts generated from the dealSent for signature without retyping
How it was done
The problem
Legal changed the standard terms and nobody could say which customers had signed which version. The answer lived in years of signed PDFs, not in any field the business could report on.
What we did
Read every signed agreement at the document level with a reviewer-and-challenger process, encoded Legal’s rulings as rules and loaded the result as structured terms on each account. Then built a governed contract workflow: a Legal-owned clause library, approval routing by clause and DocuSign generation straight from deal data.
A new sales process for every live deal, in one evening
Leadership wanted five stages with exit criteria. Every open deal moved in one evening, and the reports leadership reads each week kept their history.
Salesforce engineering with Salesforce and Reporting
Solution valueExit criteria checked before moving on
ProposalExit criteria checked before moving on
Contract sentExit criteria checked before moving on
ClosedExit criteria checked before moving on
Pick an old stage. Salesforce’s default stage names are shown; the new names are generalised.
Before the move
Every report and dashboard checkedFor any filter or grouping on the old stages
The affected ones re-pointed firstSo leadership numbers stayed continuous
The migration evening
Rehearsed in a full-copy sandboxAt production volume
Pre-image savedEvery deal and its stage history
Every open deal moved in one eveningWith no failed records
History and dates preservedReports read the same the next morning
How it was done
The problem
Leadership wanted a five-stage process with exit criteria enforced at each step. Every open deal, a large report library and the numbers leadership reviewed each week all depended on the old stages.
What we did
Swept every report and dashboard for references to the old stages, rehearsed the move in a full sandbox and migrated with a pre-image of every deal and its stage history. Exit criteria went in as validation at the stage where the data is needed.
One record per customer, across two regions and three platforms
Customers existed twice, once per region. Each now has one record, with both regions’ storefront IDs, owners and terms side by side.
Data unification with Salesforce, Adobe Commerce and Snowflake
Every wave rehearsed at full volumeIn a sandbox, before production
A pre-image kept for every recordSo any merge could be reversed
Storefront IDs checked at the sourceAgainst the commerce platform’s own API
Caught by the premortem
Currency fields would have relabelled euros as dollars.
Middleware rewrote merge targets after the merge.
Other showstoppers, each fixed before the first wave.
How it was done
The problem
Customers existed twice, once for North America and once for Europe, with storefront IDs, owners and contract fields split between the two. Every regional report needed manual reconciliation.
What we did
Ran a premortem before any merge, which caught showstoppers including currency fields that would have relabelled euros as dollars and middleware that rewrote merge targets. Rehearsed every wave at full volume in a sandbox, kept a pre-image of every record, then reconciled regional storefront IDs against the commerce platform’s own API.
Revenue leaks closed between the deal and the bill
Money leaked between the signed deal and the bill. Each leak was traced to its cause on real records, fixed at the rule and backfilled.
Revenue protection with Salesforce, DocuSign and Billing
Surfaced for review, with a preview before anything closed.
Set-up fees dropped on the way to billing
Restored on every open deal and fixed at the rule.
Billing contacts wiped on save
Protected, and the save that wiped them fixed.
Renewal dates drifting
Reconciled to the signed contracts.
What changed
AreaBeforeAfter
PipelineInflated by deals nobody ownedOnly deals someone is working
Renewal datesDrifting away from the contractMatching the signed contract
How each leak was closed
Reproduce on real recordsFind the rule that causes the leak
Fix the ruleSo the leak cannot recur
Backfill with a change fileEvery changed value recorded
How it was done
The problem
Set-up fees vanished between the signed deal and the billing record. Renewal dates drifted. Billing contacts were silently wiped. Pipeline carried deals nobody had touched in years.
What we did
Traced each leak to its mechanism with a reproduction on real records, fixed it at the rule and backfilled with a file that records exactly what changed. Built an expiry engine that previews every deal before it closes anything.
AI prospecting a sales team can trust
An AI prospecting tool kept returning customers the CRM already had. The identity layer it checks against now covers every known prospect.
AI prospecting toolFinds companies that look like good prospects
Identity registryEvery known customer and prospect, across all their domains
Fail-fast guardStops the run if the registry ever loads empty
Sales booksOnly real net-new prospects, graded
What changed
AreaBeforeAfter
Leads called net-newMostly companies the CRM already hadExisting customers recognised, real prospects surfaced
Trust in the outputVendor accuracy taken on faithCost and accuracy measured on hand-checked samples
How it was rolled out
Check a sample by handBefore any rollout
Report cost and accuracyTo leadership, in plain terms
Roll out book by bookTo each sales team in turn
How it was done
The problem
An AI prospecting tool promised net-new leads. Most of them already existed in the CRM, because the registry it checked against covered only a small share of known prospects.
What we did
Rebuilt the identity layer the AI checked against, widened it to every prospect in the CRM across multiple domains per company and added a fail-fast guard for when the registry loads empty. Measured cost and accuracy by hand-checking samples before each rollout, and reported both to leadership.
A decade of org debt, retired without an outage
Ten years of automations, profiles and stored history had piled up. The org was simplified without an outage or a locked-out user.
AutomationsSeveral racing on the same objectOne governed pattern per object
Workflow rulesLegacy rules facing platform retirementRetired ahead of the deadline
Security profilesDozens, each a slight variationConsolidated, with every user checked first
Data storageOver the limit, failing integration insertsBack under the limit within a day
How it was done
The problem
Racing automations on core objects, legacy workflow rules facing retirement, dozens of security profiles and a storage limit that had started to fail integration inserts.
What we did
Mapped every automation per object and consolidated to one governed pattern, retired dead reports on usage evidence, checked every permission change against every user before it shipped, and archived history to bring storage back under the limit in a day.
Start with the problem that sounds most like yours.
Tell us which systems are involved and what is going wrong. We reply within one business day, and new engagements begin in January 2027.