Qeasy Cloud
Get Started

Syncing Customer Master Data from OKKICRM to DaoYi Cloud Qiqiao: A Practical Breakdown of One Strategy

· 尹春锐· Integration Solutions· 25 views· 4 min read
小满OKKICRM道一云·七巧客户主数据基础资料同步轻易云Incremental Sync编码映射集成策略

What This Strategy Solves (Scenario and Value)

Syncing customer master data from a CRM to a downstream business platform looks like the simplest integration chain: write the customer profile over by ID. But in one real project, three months after go-live the customer counts on the two sides did not match—some customers had their names changed in the CRM but not downstream; some had been manually added downstream, leaving ghost records in the CRM. The root cause was that nobody was watching encoding mapping, change detection, and initialization timing.

We used the Qeasy Data Integration Platform to host this strategy. The core value is one sentence: customer master data has only one source of truth, and everywhere else is always a mirror.

Data Flow and Field Mapping (Source → Middle Layer → Target)

Data flow: OKKICRM (source) → Qeasy Integration Platform (middle layer) → DaoYi Cloud Qiqiao (target).

The middle layer does three things: pull, map, push. We do not do point-to-point, because once encoding mapping is scattered across both systems it can never be reconciled.

Key fields (only the most problematic ones):

Business meaningOKKICRM (source)DaoYi Cloud Qiqiao (target)Mapping notes
Customer unique codeCustomer ID (auto-increment numeric)External code (string, within 30 chars)Use source code as downstream external code, stringify it; never let downstream re-number
Customer nameCustomer nameCustomer namePass through directly, no concatenation, no translation
Customer tierTier dictionary (A/B/C)Tier field (enum)Dictionary lookup table lives in the middle layer; add new values to the table before syncing
Contact personContact fieldContact header + detailStage header and body: write the customer profile first, then the contacts
StatusActive/InactiveActive/InactiveStatus changes must use the update API; do not delete and recreate the whole record

The last row of this table is where most projects get burned: to save effort, teams often just delete a customer when they deactivate them, which breaks every downstream workflow and historical sales order that referenced that customer.

How to Configure It on Qeasy

On Qeasy we usually split the configuration for this strategy into three blocks: data source, mapper, writer.

Data source side: use an "incremental pull by update time" interface. The starting point is the moment the strategy is first enabled; afterwards each run picks up from the last_sync_time of the previous successful run. If the source has a soft-delete flag, pull it too—do not filter to active only.

Mapper side: encoding mapping, dictionary mapping, and string cleansing all live in a single mapping component in the middle layer, not scattered across field transformation scripts. When the customer tier dictionary changes, only one place needs to be updated.

Writer side: writing to DaoYi Cloud Qiqiao is a two-step process: first call the customer profile create/update interface (use external code as the idempotency key), then call the contact interface. The idempotency key must be the source-side code—never use the target system's auto-increment ID, otherwise a rerun will create duplicates.

On Qeasy this strategy defaults to a "dual-track" pattern: incremental during the day, with a weekly or monthly full reconciliation run. When the two sides drift, the full volume is the source of truth for repair.

Implementation Steps (Phased Scheduling)

We usually break the go-live into three phases.

Phase 1: Initial full load. One-shot push every customer profile from the source to the downstream using external code. Schedule it during a business low-peak window, for example early Sunday morning. This step defines the "starting point" for subsequent incrementals, so keep thorough logs and record the exact time the full load finished.

Phase 2: Enable incremental scheduling. Start with a 15-minute cadence, then compress to 5 minutes once stable. Do not start at 1 minute—the source API cannot take it. The incremental starting point must be the full-load completion timestamp.

Phase 3: Run a full-volume reconciliation. One week after go-live, run another full comparison on purpose. The goal is to catch the tail—customers added during the full-volume window that incrementals have not yet caught up with—and to confirm the encoding mapping has no gaps. Only after this round passes is the system truly in steady state.

A small scheduling tip: incremental tasks and full-volume tasks must use different strategy IDs. Do not toggle them with the same switch. At one customer site, the full-volume task had not finished yet but incrementals fired anyway, and the data got corrupted.

Lessons Learned

Pitfall 1: Encoding mapping scattered across field transformations. The classic mistake is writing "tier A → VIP" directly inside every field's transformation script. One retail company did this, and when they later added an S tier they had to change over a dozen places. The reliable approach is centralized dictionary mapping, maintained in one place.

Pitfall 2: Writing header and body in one shot. Bundling the customer and contacts into a single request means that if contact validation fails, the entire customer profile fails to be created. Header and body must be staged: create the customer first, then write the contacts—failures should not pollute each other.

Pitfall 3: Treating soft delete as hard delete. When the source marks a customer as inactive, downstream should not delete; it should call the update API to set the status to inactive. Hard delete breaks references.

Pitfall 4: Wrong incremental starting point. Some teams set the incremental start to "today," so all the legacy customers that existed before that day get dropped on go-live. The incremental starting point must be the full-load completion timestamp.

Pitfall 5: Using the target system ID as the idempotency key. The first run fails, gets retried, and by then the target system has already auto-assigned a new ID, so the retry creates a duplicate. The idempotency key must be the source's stable business code.

Applicable and Non-applicable Scenarios

Applicable: one-way sync of customer master data from a CRM to a downstream business platform, where the source is the single source of truth and the downstream is read-only.

Not applicable: scenarios that need two-way sync, where customer profiles are allowed to be maintained on both sides, or where the source has no reliable update-time field. In all three cases, forcing a single-strategy sync will inevitably corrupt the data later.

Original content. Please credit the source when reposting: https://www.qeasy.cloud/insights/solutions/strat-okkicrm-pf4fe44-3644-ok-233cfa9a

Comments