HubSpot Deal Pipeline Case Study: Separating an Award from a Win at an Aggregate Supplier

Summary

Quarrystone Direct supplies and hauls aggregate to construction projects: site work, concrete, quarry and excavation scopes, across several regions, through a mix of its own fleet and subcontracted trucking.

Its commercial process is bidding, and bidding has a shape that standard customer relationship management pipelines get wrong in a specific way. A general contractor is awarded a project. That is one event. The contractor then selects a materials supplier. That is a second event, and it can be weeks later, or never.

A pipeline with one won stage forces those two events into one field. The result is a forecast containing revenue nobody has been given, which is worse than no forecast, because it is confidently wrong in a direction that flatters the business.

The build described here is eight stages, three terminal states, thirteen properties across three objects, and a rule that stages never move backward. It uses no custom objects at all, deliberately.

Client details are pseudonymised at the client's request. Figures are as measured.

Background: two events, one field

The sequence in construction supply runs through an ordinary bid.

A project appears at design stage. The supplier tracks it before anyone has asked for a price. A general contractor bids it. The contractor is awarded it, or is not. The awarded contractor then chooses a materials supplier, and only at that point does the supplier have the work.

Every stage before the last one is real, and none of them is a win. Moncrief and Marshall (2005) revisit the canonical seven steps of selling and find the linear model breaking down where the buying process is not controlled by the seller, which is the exact condition here: two of the decisive events happen inside a customer's organisation and a third happens inside a customer's customer's.

The failure this causes is not subtle once named. A deal marked won when the contractor is awarded the project produces a pipeline that contains work the supplier is still competing for. Lawrence, Goodwin, O'Connor and Önkal (2006) review judgmental forecasting and document optimism bias as a persistent finding across settings. A pipeline design that rewards early optimism does not correct that bias, it operationalises it.

Pre-engagement audit

Deal pipelines in use: none fit for bidding. Quote status lived across estimating spreadsheets and the memory of the people who produced them.

Distinct commercial outcomes recorded: one. Work was won or it was not. The distinction between losing to a competitor, losing on the supplier's own terms, and a project that was cancelled or delayed did not exist as data, so loss analysis was anecdote.

The company object's meaning: overloaded. Customers, subcontractors and fleet owners were all companies, with nothing distinguishing them, which makes any report grouped by company a report mixing three populations.

Margin data: authoritative, and held elsewhere. The operational system calculated margin correctly. Nothing carried it into a reporting surface a salesperson could see.

Rebids: indistinguishable from first bids. A project quoted twice looked like a project quoted once, and the win rate was computed over a denominator nobody could define.

The build

Phase one: eight stages that stop at the right places

The pipeline is single and linear: Tracked, Quoted, Bidding, Awarded, Won, Lost to Competition, Lost on Supplier Side, No Go.

Three decisions carry the design.

Awarded and Won are separate stages. Awarded means the contractor holds the project and the supplier is still undecided. Won means the supplier is supplying. A deal sitting at Awarded is a live opportunity with a named buyer and a real budget, and it is not revenue.

There are three terminal states, not one. Lost to Competition means a competitor took it. Lost on Supplier Side means the supplier's own price, capacity or decision ended it. No Go means the project was cancelled or delayed, and nobody lost anything. Collapsing the third into the first two would put dead projects into the competitive loss rate and make the sales team look worse than it is.

Stages do not move backward. A deal that is re-scoped does not return to Quoted. It closes, and a new deal opens with a deal type of Rebid.

Phase two: the rebid as a new record

The no-reverse rule is the part operators find strange and the part reporting depends on.

Moving a deal backward destroys the history of what happened the first time. Time in stage, the date quoted, the reason it stalled: all of it becomes unreadable once a record re-enters a stage it has already left. A rebid as a new deal keeps the first attempt intact and makes the second one countable.

The cost is a mental model change, and Speier and Venkatesh (2002) document exactly this kind of cost in sales force automation: the technology's failures are rarely technical and frequently arise where a system asks salespeople to work against a habit. The mitigation used here is that the deal type is set automatically where the project identifier already exists, so the salesperson is not asked to remember the rule, only to accept it.

Phase three: thirteen properties, and no custom objects

Eight on the deal: deal type, project name and identifier, scope of work, bid phase, calculated margin, close lost reason, public or private project, and region.

Three on the company: company type distinguishing customers and subcontractors from fleet owners, primary sales region, and an active quotes count maintained by workflow.

Two on the contact: role, distinguishing a decision maker from an estimator from operations, and optional project involvement.

No custom objects were created. The reasoning is that a deal is sufficient to represent a quote for a scope of work, and a custom project object earns its complexity only when reporting or lifecycle requirements exceed what a deal can carry. Holland and Light (1999) place scope discipline among the critical success factors in enterprise system implementation, and a deferred object is cheaper to add later than an unused one is to remove.

Phase four: the field the CRM is not allowed to calculate

Margin is synced from the operational system and is never recalculated inside the CRM.

That is a one-sentence rule with a large consequence. A margin computed in two places will disagree in two places, and the disagreement surfaces during a commercial conversation rather than during a data review. The operational system owns the number, the CRM displays it, and no workflow, calculated property or report derives it independently.

Redman (1998) treats poor data quality as a continuously borne cost, and duplicated derivation is one of its commonest sources. A single owner per value is the cheapest control available, and it costs nothing at implementation time and a great deal to retrofit.

Phase five: sequencing the integration behind the model

An integration through a workflow automation layer was planned and deliberately not started.

The stated precondition was that the deal stages exist and are final, the required properties are created, the company type property is defined, and the contact association rules are clear. An integration writing into a schema still under discussion binds itself to values that may change, and every subsequent schema change then becomes a migration.

Phase six: training the process, not the platform

Six sessions with stated durations: a sixty-minute executive orientation, ninety minutes of sales core training, forty-five to sixty minutes on quoting and margin awareness, sixty minutes for sales management and operations, forty-five minutes on the field service tool where it applies, and a sixty-minute technical handoff.

What was explicitly not trained is as deliberate as what was. Generic platform navigation, marketing features that were not activated, the automation logic behind the workflows, and custom report building for sales representatives were all excluded. A salesperson who understands the bid process as the system expresses it does not need to know how the workflow that stamps the quote date is constructed.

Outcomes

Deal pipelines: from 0 to 1. Single and linear by design, because parallel pipelines make a cross-pipeline report an exercise in reconciliation.

Pipeline stages: 8, of which two represent distinct events that a conventional pipeline merges.

Terminal states: from 1 to 3. Lost to competition, lost on the supplier's own terms, and no go.

Properties created: 13. Eight on the deal, three on the company, two on the contact.

Custom objects created: 0. Deferred by decision rather than by omission.

Fields the CRM recalculates from operational data: 0. Margin is displayed and never derived.

Company populations now separable: 3. Customers and subcontractors, and fleet owners, previously indistinguishable in any grouped report.

Training sessions defined: 6, with durations, audiences and an explicit exclusion list.

Lessons learned

The most valuable stage in this pipeline is the one that is not a win. Awarded exists so that a real, named, funded opportunity can be tracked without being counted. Every pipeline serving a business that sells into someone else's project has a version of this stage, and most of them do not have it.

Three loss reasons cost the same to build as one, although only the three can be acted on. The question a sales meeting should be able to answer is not how many deals were lost. It is how many were lost to a competitor, which is a pricing and relationship problem, against how many were lost because the project died, which is not a problem at all.

The no-reverse rule resisted, and it was worth the resistance. Salespeople move deals backward because it feels like correcting a record. It is overwriting one. The compromise that made it acceptable was automating the rebid classification so the rule costs nobody a decision.

Deferring custom objects made the first version shippable. A project object, a fulfilment object and a haulage object were all discussed. None of them was necessary to answer the questions the business could not previously answer, and each would have delayed the answers by weeks.

Limits

No business outcome is reported here. Win rate, quote-to-win ratio and margin by region are now answerable questions, and the engagement has not run long enough to produce a series worth publishing. Nothing here should be read as a claim that win rates improved, and the design does not establish that they will.

A project spanning several bids has no object of its own. Rebids relate to their predecessor through the project identifier carried as a property, not through an association. Grouping by project therefore depends on that identifier being entered consistently, which is a data-entry dependency rather than a structural guarantee, and it may fail quietly.

The integration had not begun at the point described. The CRM was being finished as a container first, by explicit sequencing. Everything written here describes the container.

Fulfilment and haulage execution are out of scope for this version. The pipeline models the commercial process up to the point of winning supply. What happens to the trucks afterwards is not in it.

The margin figure is only as good as its source. The design forbids the CRM from recalculating it, which removes one class of error although it removes none of the errors already present upstream.

This is a bid-driven business. A supplier selling from a catalogue to a buyer with authority to purchase has no award stage, because there is no third party deciding anything. The distinction this pipeline is built around would be dead weight there.

Conclusion

The question this build answers is which of the numbers on a sales dashboard is revenue.

Before, a supplier tracking construction projects had an estimating process, a set of spreadsheets and the judgement of the people running them. That judgement was good. It was also unauditable, and it could not be grouped by region, scope or reason.

Reinartz, Krafft and Hoyer (2004) find the strongest performance association in the maintenance stage of the customer relationship management process, and in bid-driven supply the maintenance stage is the long middle where a project is awarded and the supplier is waiting. A pipeline that has no stage for waiting either counts that time as a win or loses it entirely, and both make the same business look like a different one.

References

Holland, C. P., & Light, B. (1999). A critical success factors model for ERP implementation. IEEE Software, 16(3), 30–36. https://doi.org/10.1109/52.765784

Lawrence, M., Goodwin, P., O'Connor, M., & Önkal, D. (2006). Judgmental forecasting: A review of progress over the last 25 years. International Journal of Forecasting, 22(3), 493–518. https://doi.org/10.1016/j.ijforecast.2006.03.007

Moncrief, W. C., & Marshall, G. W. (2005). The evolution of the seven steps of selling. Industrial Marketing Management, 34(1), 13–22. https://doi.org/10.1016/j.indmarman.2004.06.001

Redman, T. C. (1998). The impact of poor data quality on the typical enterprise. Communications of the ACM, 41(2), 79–82. https://doi.org/10.1145/269012.269025

Reinartz, W., Krafft, M., & Hoyer, W. D. (2004). The customer relationship management process: Its measurement and impact on performance. Journal of Marketing Research, 41(3), 293–305. https://doi.org/10.1509/jmkr.41.3.293.35991

Speier, C., & Venkatesh, V. (2002). The hidden minefields in the adoption of sales force automation technologies. Journal of Marketing, 66(3), 98–111. https://doi.org/10.1509/jmkg.66.3.98.18510

Conflict of Interest Statement

RevOps HQ is a HubSpot Solutions Partner and was paid to design and build the pipeline described here, and to deliver the training listed. Stage names, property lists and training durations come from the approved architecture document. No business outcome is claimed, and the reporting questions the design makes answerable are described as answerable rather than as answered.

Acknowledgments

The bid phase taxonomy and the scope-of-work values come from the client's own estimating practice rather than from any reference model. Margin figures are calculated in the client's operational system and are displayed unchanged.