HubSpot–Sage Intacct Integration Case Study: Project Margin at the Deal for a Consultancy

Summary

Ardenmoor Consulting delivers fixed-fee transformation projects, at 60 to 80 active engagements and an average project value of $186,000. Sage Intacct held revenue recognition, project accounting and the dimensional structure the firm reported on. HubSpot held the pipeline. Nothing connected them.

The engagement mapped Intacct dimensions to HubSpot objects, surfaced realised margin against the deal that sold each project, and left every recognition calculation in Intacct. Projects finishing below their target margin fell from 34 percent to 19 percent, the interval between a margin problem starting and someone seeing it fell from a median 61 days to 6, and quotes priced below the firm's floor fell from 12 percent to under 2.

Background: Margin Known Only in Arrears

Ardenmoor priced from a rate card and a scoping estimate. Whether a project made money was established at quarter close, by which point the engagement was frequently complete and the next one had been sold on the same assumptions.

The mechanism was simple and slow. Time posted to Intacct against a project dimension. Recognition ran on a schedule. Margin appeared in a report the delivery leads did not open, and the salespeople who set the price never saw it at all.

Karimi, Somers, and Bhattacherjee (2007) found that a system becomes a capability only where its outputs enter the working routine of the people the process runs through. Intacct was producing the right number on the right schedule for a finance function; nobody in the routine that set prices was reading it.

The Audit

Four weeks against Intacct, the HubSpot portal and two years of closed projects.

Margin variance. 34 percent of projects closed below their target margin. The median shortfall was 11 percentage points.

Detection lag. Median 61 days between a project's cost curve diverging from plan and any person recording awareness of it.

Pricing floor. 12 percent of quotes were priced below the firm's stated floor, ordinarily by a seller applying a discount the rate card did not accommodate.

Dimensional mismatch. Intacct carried five dimensions — project, customer, department, location and service line. HubSpot carried none of them as structured fields. Service line existed as free text on 71 percent of deals, spelled eleven ways.

Recognition Stays in Intacct

The firm asked whether margin could be calculated in HubSpot. The answer was no, and establishing why took the first week.

Recognition depends on rules about performance obligations, milestones and unbilled balances, and those rules are audited. Reproducing them in a CRM creates a second answer that may well diverge from the audited one, and the divergence would likely be discovered by whoever trusts the wrong figure. Pipino, Lee, and Wang (2002) set out how data quality can be assessed in practice and treat consistency as a metric to be computed rather than assumed; two implementations of an audited rule are two values to reconcile, and nobody had been assigned to reconcile them.

HubSpot therefore holds no recognition logic. It holds a realised margin figure computed by Intacct and stamped against the deal, with the date the figure was produced. Parnas (1972) compared two decompositions of one program and showed that modules organised around the decisions they conceal survive change, because neighbours depend on an interface rather than on internals. Recognition is such a decision.

The Build

Eight weeks, four phases, using the Intacct Web Services API, a nightly job in the client's tenancy, and a private HubSpot app.

Phase one, service line. The free-text field was replaced with a defined set matching Intacct's service-line dimension, and the 71 percent that had values were mapped by hand in review with delivery leads. Eleven spellings resolved to five values.

Phase two, dimension mapping. Project dimension maps to a HubSpot custom object; customer maps to the company; department and location map to properties on the deal. The mapping is written down and has not changed.

Phase three, margin surfacing. A nightly job writes realised margin, cost to date and percentage complete to the project object. The figures are read-only in HubSpot and carry the timestamp of the Intacct run that produced them.

Phase four, floor enforcement. A quote priced below the service line's floor cannot be submitted without an approval, and the approval is recorded against the deal rather than granted verbally.

Outcomes

Projects below target margin. 34 percent before, 19 percent after, across the two quarters following go-live.

Detection lag. Median 61 days before, 6 after. The improvement comes from surfacing rather than from any change to how cost is captured.

Sub-floor quoting. 12 percent of quotes before, 1.8 percent after, with each remaining instance carrying a recorded approval.

Service line data quality. Structured service line rose from 0 percent to 100 percent of new deals, against 71 percent carrying an unreliable free-text value before.

Margin visibility. Delivery leads opening a margin figure at least weekly rose from an estimated 2 of 9 to 9 of 9, measured by record views.

Estimate accuracy. Median variance between scoped and delivered hours narrowed from 23 percent to 14 percent, which the firm attributes to sellers seeing the outcome of prior estimates.

What Resisted

Percentage complete resisted. Intacct computes it from cost against budget, which is correct for recognition and misleading as a progress indicator to a seller: a project running over budget appears further along than it is.

The mechanism was to surface both the cost-based figure and the delivery lead's own assessment, side by side, and to label which is which. Reconciling them into one number was considered and rejected, because the disagreement between them is the signal rather than noise.

The second difficulty was cultural. Making margin visible to sellers exposed that some longstanding accounts were consistently unprofitable, and the first response was to question the figure. Haug, Zachariassen, and van Liempd (2011) separate the operational cost of working around defective data from the cost of decisions taken on it, and the useful move was to audit three disputed projects line by line rather than to defend the integration. Two of the three figures were correct; the third revealed a dimension mapping error, which was fixed.

Limits

This does not prove that surfacing margin improves it. Ardenmoor changed its approval process at the same time, and the two effects cannot be separated from these figures alone. The detection-lag improvement is the cleanest result here because it follows mechanically from the surfacing.

The design also assumes time is posted to Intacct promptly. Where timesheets run a fortnight behind, the margin figure on the deal is a fortnight stale and carries a timestamp saying so, which is honest but not useful. Fixing that would be a timesheet discipline engagement rather than an integration one.

Figures come from the client's own project accounting. No holdout was run, and the comparison periods differ in project mix.

Conclusion

The integration moved no calculation. It moved an existing number to the people whose decisions depended on it, and stamped it with the time it was produced so nobody would treat it as live.

The general form is that a figure produced accurately on a finance schedule is not thereby available to the commercial function. Availability is a separate problem, and it is ordinarily the one that matters.

References

Haug, Anders, Frederik Zachariassen, and Dennis van Liempd. 2011. "The Costs of Poor Data Quality." Journal of Industrial Engineering and Management 4 (2): 168–193. https://doi.org/10.3926/jiem.2011.v4n2.p168-193

Karimi, Jahangir, Toni M. Somers, and Anol Bhattacherjee. 2007. "The Role of Information Systems Resources in ERP Capability Building and Business Process Outcomes." Journal of Management Information Systems 24 (2): 221–260. https://doi.org/10.2753/MIS0742-1222240209

Parnas, David L. 1972. "On the Criteria To Be Used in Decomposing Systems into Modules." Communications of the ACM 15 (12): 1053–1058. https://doi.org/10.1145/361598.361623

Pipino, Leo L., Yang W. Lee, and Richard Y. Wang. 2002. "Data Quality Assessment." Communications of the ACM 45 (4): 211–218. https://doi.org/10.1145/505248.506010

Umble, Elisabeth J., Ronald R. Haft, and M. Michael Umble. 2003. "Enterprise Resource Planning: Implementation Procedures and Critical Success Factors." European Journal of Operational Research 146 (2): 241–257. https://doi.org/10.1016/S0377-2217(02)00547-7

Conflict of Interest

RevOps HQ is a HubSpot Solutions Partner and was engaged and paid by the client described.

Acknowledgments

Prepared by RevOps HQ.